The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Dollarization: A Scorecard R O B E R T O C H A N G The author is an associate professor at Rutgers University. At the time this article was written, he was a research officer in the macropolicy section of the Atlanta Fed’s research department. He thanks Marco Espinosa, Frank King, Will Roberds, George Selgin, and Ellis Tallman for useful comments and suggestions and Michael Chriszt for valuable assistance. I N J ANUARY OF THIS YEAR , J AMIL M AHUAD , THEN PRESIDENT OF E CUADOR , STARTLED HIS COMPATRIOTS BY PROPOSING TO ELIMINATE THE NATIONAL CURRENCY, THE SUCRE . I NSTEAD , M AHUAD S UCH ADVANCED , THE U.S. DOLLAR WOULD REPLACE THE SUCRE FOR ALL PURPOSES . A CHANGE WOULD STABILIZE E CUADOR ’ S and help return the country to growth and prosperity. Mahuad’s announcement was received with skepticism, and a popular uprising forced him out of office a week later. But the succeeding government has actually implemented Mahuad’s proposal and recently announced that U.S. dollars will have completely replaced the sucre by September 2000. The question remains as to whether the Ecuadorian plan will be successful and, more generally, whether other countries will follow Ecuador’s lead. But one thing is clear: proposals to replace the domestic currency with the U.S. dollar, or to “dollarize” the economy, have taken center stage in Latin America and other developing regions. The idea is the subject of hot debates in academic and policy circles. And where candidates stand with respect to dollarization has become a key factor in several political contests. The ascent of dollarization from being a relatively neglected issue to the role of powerful economic medicine is mysterious and fascinating. It implies, for example, that as monetary policy in advanced economies has steadily gained respect, developing nations have come to believe it is better to get rid of their own currencies. To illustrate why, SINKING ECONOMY AFTER A DISMAL 1999 this article will discuss the currently fashionable proposals for dollarization in Latin America and other developing regions. After describing what dollarization is about, the article places special emphasis on identifying and analyzing various arguments for and against dollarization in the light of existing economic theory and evidence. By embracing dollarization, a developing country would accept at least three costly consequences: • Its government would give up the revenue it enjoys from creating money. • Its central bank would no longer serve as a lender of last resort to domestic banks. • It would no longer control domestic monetary policy. On the other side of the ledger, dollarization may result in at least two benefits: • Dollarization may lower the country’s cost of foreign credit. • It may enhance the credibility of government policy. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 1 Are the benefits from dollarization worth its costs? The answer is unclear. Comparing the relative significance of the costs and benefits listed involves many difficulties. The article discusses these issues and, as it moves down the list, it finds that measurement problems become considerably larger. The conclusion is that, remarkably, the popular belief that dollarization is a desirable reform has been reached in spite of widespread uncertainty about its economic benefits. tries? In contrast to many other economic reforms, recent dollarization proposals are very simple. Essentially, two exchanges would have to be enacted: • The domestic monetary base, that is, all local currency (coins and notes) in circulation plus the vault cash reserves of banks, would be redeemed for U.S. dollar bills at some predetermined conversion rate and then destroyed. In Ecuador, for example, the central bank is currently paying one dollar for each twenty-fivethousand-sucre coin and note presented to it. • All contracts denominated in local currency would be transformed into contracts in U.S. dollars, also at predetermined conversion rates (which may, but do not necessarily, equal the rate for coins and bills). In particular, local currency bank deposits would become deposits denominated in and payable in dollars. Official Dollarization: What It Is and What It Is Not ost nations have their own currencies, but in a number of countries the U.S. dollar is also used widely as a means of payment, store of value, and unit of account. This phenomenon, which may be called unofficial dollarization, became As inflation rates in manifest in the 1980s in some economies Mexico, Israel, Peru, and experiencing rapid other countries reached inflation. As inflation triple digits, domestic resirates in Mexico, Israel, Peru, and other dents learned to protect countries reached themselves against the loss triple digits, domestic of purchasing power of residents learned to protect themselves their national currencies against the loss of by switching to the dollar. purchasing power of their national currencies by switching to the dollar. Over time, the governments of these countries validated unofficial dollarization by allowing residents to open bank accounts denominated in dollars and by including dollars in circulation and dollar deposits in their own monetary statistics. In spite of the growing importance of dollarization, attention to it remained confined to a relatively small and mostly academic literature.1 This state of affairs changed, however, with the sequence of crises in emerging markets that started with the Mexican one in 1994–95 and continued with the 1997–98 crises in Asia, Russia, and Brazil. The observation that dramatic and costly devaluations of national currencies were common to most crises led to a renewed search for alternative exchange rate arrangements. Official dollarization emerged then as a feasible and intriguing possibility. But its popularity ballooned after President Carlos Menem of Argentina announced at the beginning of 1999 that he was seriously considering a dollarization plan. What would official dollarization actually involve for Argentina, Ecuador, and similar developing coun- M 2 Three features of this scheme deserve particular emphasis. First, dollarization would be unilateral. Ecuador did not ask the U.S. government for permission to implement its dollarization scheme, nor did it have to. Second, and as a consequence, under this plan a government would give up any power to conduct independent monetary policy and would implicitly accept the monetary policy decisions of the U.S. Federal Reserve. Ecuador’s decision to dollarize gave it neither a voice nor a vote in the Federal Open Market Committee. Third, the local currency would be completely replaced by the dollar, not just by a dollar equivalent. These aspects are noteworthy because they set dollarization apart from some alternatives.2 For instance, instead of pursuing dollarization unilaterally, a developing country could try to negotiate a Treaty of Monetary Association with the United States. Such a treaty would entitle the dollarizing country to some transfers from the U.S. government (and, ultimately, from the U.S. taxpayer) as compensation for the loss of monetary policy independence. A more encompassing alternative would be to negotiate a monetary union similar to the current European Monetary System. In a monetary union, the United States and the developing country would agree to have a common currency, which could be the U.S. dollar, and a common monetary authority. Hence the developing country would presumably gain some control over monetary policy decisions. It has become evident that neither a monetary association treaty nor a monetary union would be feasible without prolonged and complex negotiations between the U.S. government and the developing countries involved. As a consequence, some developing nations, Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 led by Argentina, have moved almost all the way toward dollarization and established a currency board system. In a “pure” currency board, the central bank stands ready to buy or sell U.S. dollars for domestic currency at a fixed exchange rate; in addition, the currency board does not issue domestic currency in exchange for local currency assets. This commitment is, in turn, guaranteed to be permanent because the amount of domestic currency in circulation is fully backed by U.S. dollars held by the central bank. This arrangement does not eliminate the local currency but makes it, in principle, completely equivalent to the U.S. dollar. That would indeed be the outcome in Argentina if holders of Argentinean pesos were 100 percent sure that the central bank of Argentina would always be there to pay one dollar for each peso in circulation.3 If in fact a government can make the national currency a perfect substitute for the U.S. dollar by establishing a currency board, what are the further gains from official dollarization? On the other hand, if dollarizing the economy is so straightforward, why not just do it? The answer to both questions, of course, is that dollarization may have costs as well as benefits, to which the discussion now turns. The Seigniorage Question country engaging in official dollarization would naturally have to destroy its national currency. This fact has been used by some to argue against dollarization on the basis of national pride, symbols, traditions, and the like. While such arguments can be politically effective, their economic significance is virtually impossible to evaluate. Perhaps as a consequence, economists have avoided debating these political factors. Instead, economic theory identifies other reasons why it is costly for a nation to eliminate its national currency. Some of the losses can be readily quantified, and they turn out to be significant in practice. The first source of losses from dollarization is what is called seigniorage. The right to create domestic currency is valuable for a government because newly printed currency can be issued in exchange for goods and services. For example, newly printed pesos, which cost essentially nothing to produce, allow the Bank of Mexico to pay for purchases of goods or for foreign exchange as well as A to grant credits to domestic banks. This is what economists mean when they say that a government collects revenue when it prints domestic currency or, in other words, that it extracts seigniorage revenues from the economy. A dollarizing country would give up its seigniorage revenues, which would accrue instead to the U.S. government. For developing countries, the loss can be large. To illustrate, the significance of annual seigniorage revenues has been calculated for Argentina, Brazil, and Mexico, over the 1993–97 period for Argentina and Mexico and 1995–97 for Brazil (the difference being that for Brazil previous years are abnormally turbulent.)4 Chart 1 displays the actual flow of revenue from printing domestic currency, given by the increase in the domestic monetary base in a year, both in terms of U.S. dollars and as a percentage of the country’s gross national product (GDP). Accordingly, the chart shows that seigniorage revenue per year has been about one-third of 1 percent of GDP for Argentina, or U.S.$1.2 billion per year. In contrast, seigniorage revenues in Brazil have been about 1.3 percent of GDP, or U.S.$10 billion per year. In other words, after correcting for the relative size of the Argentinean and the Brazilian economies, the seigniorage transfer from Brazil to the United States under dollarization would be more than three times larger than the corresponding transfer from Argentina to the United States. These figures explain, perhaps, why the dollarization idea has been clearly more popular in Argentina than in Brazil. Mexico turns out to be a middle case: annual seigniorage revenues amount to nine-tenths of 1 percent of GDP, about U.S.$3.8 billion. The different magnitudes of seigniorage in Argentina, Brazil, and Mexico reflect the different degrees to which their governments have relied on money creation to finance their budgets. Chart 2 shows that money creation is only 1.7 percent of total government revenue in Argentina. Therefore losing seigniorage would not require a major fiscal adjustment in that country. But in Brazil, domestic money creation is responsible for almost 9 percent of government revenue. As a consequence, dollarization would require Brazil to find new and significant sources of tax revenues or to drastically reduce government expenditures. While the fiscal effects 1. For a taste of the literature prior to the 1994 Mexico crisis, see Chang (1994) and the references therein. 2. See Berg and Borensztein (2000) for a more detailed discussion. 3. Argentina’s system, called the Convertibility Law, departs from a pure currency board in a number of ways. For example, up to one-third of the central bank’s reserves that back base money can be held in the form of Argentinean government bonds. For a full discussion, see Hanke and Schuler (1999), who also argue for a move toward official dollarization. 4. The method for calculating seigniorage revenues follows Fischer (1982). Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 3 C H A R T 1 Flow of Revenue from Seigniorage as a Percentage of GDP 1.4 U.S.$10.0 billion Percent 1.0 U.S.$3.8 billion 0.6 U.S.$1.2 billion 0.2 Argentina 1993–97 average Brazil 1995–97 average Mexico 1993–97 average Source: International Financial Statistics C H A R T 2 Flow of Revenue from Seigniorage as a Percentage of Total Government Revenue 10 U.S.$10.0 billion Percent 8 6 U.S.$3.8 billion 4 U.S.$1.2 billion 2 0 Argentina 1993–97 average Brazil 1995–97 average Mexico 1993–97 average Source: International Financial Statistics would not be as large for Mexico, seigniorage is about 4.7 percent of total government revenue. The numbers just quoted refer to what is called flow seigniorage, the amount of resources that Argentina, Brazil, and Mexico would have to transfer to the United States each year in order to obtain dollar bills. In addition to these annual transfers, dollarization would require exchanging the existing amounts of domestic currency in circulation for U.S. dollars, entailing an additional, one-time startup cost equal to the dollar value of the domestic currency in circulation. 4 Chart 3 shows the magnitude of this startup cost. For the three countries under analysis, the cost is substantial, between 2 percent and 4 percent of GDP. The relatively smaller cost for Brazil reflects a smaller demand for domestic currency, which is natural because hyperinflation was perceived as a major threat until very recently. Note that the cost, when expressed in absolute dollar amounts, is very similar in the three cases (U.S.$14 billion–15 billion). The conclusion is that dollarization would imply a fairly large seigniorage transfer from a dollarizing country to the United States. For Argentina, the Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 C H A R T 3 Cost of Replacing Domestic Currency in Circulation with U.S. Dollars U.S.$13.6 billion 4 Percent of GDP U.S.$13.9 billion 3 U.S.$15.8 billion 2 1 0 Argentina 1993–97 average Brazil 1995–97 average Mexico 1993–97 average Source: International Financial Statistics transfer would be paid mostly up front while for Brazil the transfer would be paid mostly over time. Mexico seems to be an intermediate case. The amounts involved are clearly identifiable and relatively straightforward to calculate.5 Even the most ardent proponents of dollarization do admit that the loss of seigniorage is an important drawback. To mitigate the effects, it has been suggested that a dollarizing nation should engage in negotiations with the United States to recover some of the seigniorage revenue involved. But it is fair to say that there is little reason to be optimistic about prospects for such negotiations in the short run. There is an important caveat. The discussion has assumed, implicitly, that the loss of seigniorage revenue is costly. That assumption is usually accepted even by dollarization proponents, but is a debatable one. In particular, it has been argued that the loss of seigniorage may be beneficial if it forces an otherwise irresponsible government to choose sound policies. This key contention is taken up later in this article. For the time being, it suffices to note that the loss of seigniorage would be costly even if the government were completely responsible. In such a case, however, the calculation of the seigniorage lost with dollarization may have to be based on the inflation rate that would be chosen by such a government, which may be lower than past inflation (see Chang and Velasco 2000a). The Lender of Last Resort second issue with official dollarization is that a country’s central bank would no longer serve as the lender of last resort to the domestic banking system. A lender of last resort is an institution that stands ready to provide credit to banks in the event that they experience a sudden demand for liquidity, as when bank runs occur. Such an institution is crucial in a system of banks with fractional reserves in order to reassure bank depositors and short-term creditors that their claims on the banks will be honored if they attempt to liquidate them. If there is no lender of last resort, confidence crises and bank runs become more likely and more damaging if they occur.6 In most countries, the role of lender of last resort has traditionally been played by the central bank. This role is natural because the central bank can create credit quickly and at a negligible cost simply by issuing domestic currency. Since the ability to print currency would disappear under official dollarization, the central bank would no longer be able to serve as the lender of last resort. A 5. Incidentally, the amounts in Chart 3 may be taken as not only the startup costs of dollarization but also the potential profit of reintroducing a national currency from the viewpoint of a government in a dollarized economy. In other words, these amounts are a measure of the temptation to renege on the implicit promise that dollarization is a permanent reform. 6. U.S. history, in fact, provides a good example of the importance of the lender of last resort. Miron (1986) and others have documented that bank panics became less frequent after the establishment of the Federal Reserve. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 5 The associated cost is hard to quantify, however. For one thing, the cost depends on what system dollarization is replacing. In Argentina, for instance, dollarization would replace a currency board system. But, as shown formally by Chang and Velasco (2000c), a central bank cannot effectively act as a lender of last resort if it is fixing the exchange rate, as in a currency board system. In other words, moving from a currency board to dollarization would not change the ability of the central bank to act as a lender of last resort: that ability does not exist in the first place. On the other hand, in countries with flexible exchange rates, such as Brazil and Mexico, the central bank can act as a lender of last resort. Dollarization may therefore be more costly in these cases because of the need to implement an alterUnder dollarization a govnative arrangement. ernment would give up any In addition, dollarpower to conduct indepenization (or, for that matter, a currency dent monetary policy and board or a fixed exwould implicitly accept the change rate system) monetary policy decisions would not imply the complete absence of a of the U.S. Federal Reserve. lender of last resort, only that the central bank cannot perform that role. There are a number of alternatives open to a dollarized country. The government could set aside a liquid fund to be lent to banks in a crisis. Another possibility would be for the government to secure lines of credit from abroad that could be drawn upon in the event, and only in the event, of a banking crisis. Argentina, in fact, has been implementing the latter strategy since the Tequila Crisis of 1995. However, both alternatives are more costly than having the central bank serve as the lender of last resort. And there is considerable debate about how to calculate that cost in practice. Argentina provides an excellent illustration of these points. In 1995, following the Mexican crisis of December 1994, the Argentinean banking system came under intense pressure; in particular, there was a massive liquidation of domestic deposits, caused by growing expectations of bank failures. Because Argentina was committed to a currency board system, the central bank could not assist domestic banks with emergency credit. In other words, the rules of the currency board did not then (and do not now) allow the central bank to act freely as a lender of last resort. In 1995 the situation was saved thanks to a rescue package orchestrated by the International Monetary 6 Fund and the World Bank. After that scare, the Argentinean government secured lines of credit from foreign private banks to be used in case of an emergency. This arrangement clearly helps replace the lender of last resort, but it is not free. Argentina must pay a premium for what is essentially an option to borrow. In addition, it is not clear whether the credit lines are large enough to replace what the central bank might be able to do as a lender of last resort. In particular, in a crisis, banks may need enough liquidity assistance to meet all of their demandable obligations, as stressed by Chang and Velasco (2000c). To be effective, the Argentinean lines of credit may therefore need to be much larger than they actually are. It may be instructive to attempt at least a very rough calculation of the cost of the Argentinean strategy. In 1996 the Argentinean private line of credit reached U.S.$6.1 billion, at a cost of about U.S.$18 million a year—about 0.3 percent. Assuming that this rate remains the same, the total cost of the strategy would depend on how large a credit line is “enough.” At the end of 1999, Argentina’s M2 (the sum of its banking system monetary and quasimonetary liabilities) was U.S.$88.2 billion. Since it had U.S.$26.5 billion in international reserves, the banking system’s net liquid liabilities were arguably as large as U.S.$88.2 billion – U.S.$26.5 billion = U.S.$51.7 billion. At 0.3 percent a year, a line of credit large enough to cover that amount in whole would cost somewhat more than U.S.$150 million per year. While this is not a negligible figure, it is only a small fraction of Argentina’s GDP. Of course, the calculation just performed could be refined in several ways. However, the point is twofold. First, replacing the domestic central bank with an Argentina-style line-of-credit approach would be a negative consequence of dollarization. Second, the cost can be calculated explicitly. For Argentina, the cost turns out to be relatively minor, and it is much smaller than that associated with the loss of seigniorage. To end the discussion of the Argentinean strategy, it should be noted that significant enforcement questions remain unsettled. Will the foreign banks involved be reliable enough to deliver on their contractual obligations in a time of crisis instead of defaulting? Who would enforce the contracts between Argentina and the creditor banks? And who would ensure that the creditors themselves have access to the necessary liquidity? Losing Independent Monetary Policy t is widely accepted that, in modern economies, conditions in the domestic market for money are crucial determinants of macroeconomic outcomes. I Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 There is debate about what concept of “money” is relevant for analysis. However, whatever that concept is, it must include at least the monetary base, that is, local currency in circulation and in bank vaults. By controlling the monetary base, a central bank affects the domestic market for money and, consequently, may have an impact on real (inflationadjusted) activity and inflation. With official dollarization, the domestic central bank would no longer have access to its key policy tool, the monetary base. In this sense, there would simply be no sovereign monetary policy. Whether the loss of monetary autonomy is costly in practice is controversial. This fact may come as a surprise, particularly to readers who believe in the effectiveness of monetary policy. Nevertheless, as the discussion will show, it has been argued that the loss of independent monetary policy would entail essentially no cost to developing economies. There are many sides to this contention. One depends, as with the question of the lender of last resort, on what system dollarization is compared with. If dollarization is considered as the alternative to a currency board, as in Argentina, then the loss of monetary independence is not an issue: in a currency board, there is no such independence anyway. The issue of independence emerges only if dollarization is considered against exchange rate flexibility. Even then, there are those who argue that the issue is moot. For instance, Hausmann and others (1999) have observed that, in response to the 1997–98 emerging markets crises, interest rates were least variable in countries with more rigid exchange rate systems. At the same time, exchange rates moved very little in countries with flexible exchange rate systems. Hausmann and others attribute this combination to the fact that, in this period, countries with flexible exchange rates raised interest rates aggressively to defend their currencies. Hence they state that “Latin American Central Banks used their exchange rate flexibility very sparingly, even when they formally float or have wide bands” (1999, 7). The implication is that whatever policy leverage is lost with dollarization is unimportant as it would not be used even if exchange rates were flexible. However, the fact that central banks try to smooth exchange rate fluctuations is not an argument in favor of fixing exchange rates, currency boards, or dollarization. Such a behavior is neither inconsistent with floating exchange rates nor suboptimal in principle. Further, it can be shown that even if exchange rate stabilization were not a primary goal for the central bank, monetary instruments—and consequently interest rates—would still have to react to exchange rates. This reaction would occur because, as discussed by Chang and Velasco (2000b), current exchange rate movements carry information that is useful for predicting future goal variables, such as future inflation or employment. Perhaps most importantly, a recent study by Ghosh and others (1997) provides convincing evidence that more rigid exchange rate systems result in larger fluctuations in output and employment and in perhaps less growth. Table 1 summarizes the implications for developing countries. The table compares developing countries under pegged, floating, and intermediate exchange rate systems in terms of their economic performance. The latter is measured by the level and volatility of inflation, growth, and employment relative to the average for all countries. Hence, the leftmost number of the first row, 0.00, indicates that developing countries with pegged exchange rates grew, on average, at the same rate as a typical country. In contrast, the third number of the first row, 0.50, indicates that developing countries with floating exchange rates grew, on average, half a percentage point faster per year than the typical country. From the table, it is apparent that floating exchange rate regimes have enjoyed faster growth T A B L E 1 Developing Countries: Exchange Rate Regimes and Macroeconomic Performance Average for Various Exchange Rate Regimesa Pegged Intermediate Floating Output Growth Level Volatility 0.00 0.08 0.70 –0.80 0.50 –0.52 Employment Volatility 0.05 0.01 –0.32 –2.90 –1.74 –0.10 0.53 3.80 1.67 Inflation Rate Level Volatility a For the countries with each type of exchange rate regime, the figures are the average, in percent, of the deviation from the average of all countries. Source: Goldfajn and Olivares (2000), based on Ghosh and others (1997) Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 7 and less volatility in output and employment than fixed rate regimes. The cost of this accomplishment has been substantially higher and more variable inflation. These results are consistent with the view that flexible exchange rates allow a government to cushion the impact of exogenous shocks on output and employment, even at the cost of generating inflation. But the fact that flexible exchange rates allow for the existence of a trade-off seems indisputable.7 The remaining question is whether that is a tradeoff worth having. The net cost associated with the loss of monetary independence depends, at the end, on the value of a more stable price level that dollarization brings about relative to the more satisfactory behavior of output and employment associated with flexible exchange rates. The answer is unclear and likely to depend on the specific characteristics of individual countries. Would Dollarization Lower the Cost of Credit? p to this point, the discussion has mentioned a number of costly consequences of official dollarization. But dollarization has some favorable aspects, as one should expect. One of them is that the risk of currency devaluation would not exist because the domestic currency would disappear. As a consequence, dollarization proponents argue, the cost of foreign credit for a dollarizing country would come down, stimulating investment and economic growth.8 However, that dollarization would result in a lower cost of credit is not as straightforward as it might sound. Consider what determines the interest rate that a resident of a developing economy must pay to borrow in the world market. If the loan is denominated in U.S. dollars, there is often a difference or spread between the rate charged to that borrower and the rate lenders would charge to otherwise similar American borrowers. That difference is what is called default risk or sovereign risk. It reflects the possibility of a developing country’s default on its foreign debt. Such a possibility increases the cost of credit for all domestic agents. Even if default can be declared only by the government on official debt, domestic residents are likely to also stop or have difficulties servicing their individual obligations.9 If the loan is denominated not in dollars but in domestic currency, there may be an additional spread component that compensates lenders for the possibility of a devaluation of the domestic currency. In other words, the spread must also incorporate devaluation risk. Since dollarization would imply the elimination of the national currency, there would no longer be domestic currency loans, and in this sense devaluation U 8 risk would disappear. However, it does not necessarily follow that the cost of credit would be lower for domestic residents. Instead, those that had the option of borrowing domestic currency before dollarization would be forced to take dollar loans instead. Thus, other things being equal, dollarization would be detrimental to those borrowers because, before dollarization, they presumably could have borrowed in dollars but chose not to. The hope of dollarization advocates, however, is that the disappearance of devaluation risk would also reduce sovereign risk and hence result in lower dollar interest rates. This development may occur for at least two reasons. First, domestic residents or the government often have foreign currency liabilities but revenues that depend on the value of the domestic currency. A devaluation of the currency then increases the relative value of the liabilities, causing domestic bankruptcies. This situation, implied by the currency mismatch of assets and liabilities, would presumably be avoided with dollarization. Second, a government that is committed to defending the value of the national currency may resort to capital controls in order to fend off speculative attacks. Capital controls may, in turn, force domestic borrowers into default on their foreign debts. However, it is also possible that eliminating devaluation risk may increase sovereign risk instead of reducing it. In particular, Chang and Velasco (2000c) showed that, in a theoretical model, a currency board or official dollarization may “succeed” in eliminating devaluations. However, such a success comes at the cost of preventing the central bank from acting as lender of last resort to domestic banks and implies the possibility of bank runs. In contrast, a flexible exchange rate system allows the central bank to create domestic credit to help banks in case of trouble and eliminates costly speculative runs. The implication, if domestic banks have foreign debts, is that flexible exchange rates may help reduce the risk of default on those debts. A different mechanism, emphasized by Berg and Borensztein (2000), is that a devaluation may have expansionary economic effects, which may improve fiscal revenues and reduce pressures for default. Unfortunately, the empirical evidence is not very helpful in clarifying whether dollarization would reduce sovereign risk. To illustrate why, Chart 4 displays measures of sovereign risk for Argentina, Brazil, Mexico, and Peru during the second half of 1998 and all of 1999. Sovereign risk is measured by the difference between the yield of each country’s Brady bonds and the yield on five-year U.S. Treasury bonds. The chart shows no discernible relation between sovereign risk and whether Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 C H A R T 4 Sovereign Risk 3000 1000 0 July 1998 Argentina ––––– ––––– 2000 ––––– Basis points – – – – – Brazil Peru January 1999 Mexico July Note: Each country’s graph is its Brady bond discount stripped yield minus the five-year U.S. Treasury yield, in basis points. Source: Calculated by the Federal Reserve Bank of Atlanta using data from BancBoston Robertson Stephens, Inc., and the Board of Governors of the Federal Reserve System. exchange rates are fixed or flexible. Argentina maintained a currency board and for most of the period had to pay a sovereign spread higher than that of Mexico or Peru, both of which maintained flexible exchange rates. On the other hand, the Argentinean spread was lower than Brazil’s, even after Brazil switched to a flexible rate regime at the beginning of 1999. Also, the chart shows that the sovereign spreads of the four countries moved together after the Russian crisis of August 1998 and the Brazilian devaluation of January 1999. The chart thus underscores the fact that, to a large extent, sovereign spread movements reflect shocks that affect developing countries as a whole, independently of their exchange rate regimes. The above remarks apply even for the case of Panama, which has been officially dollarized since 1904. Berg and Borensztein (2000), in particular, point out that Panama’s sovereign spread has been very similar to Argentina’s and was strongly affected by the Asian, Russian, and Brazilian crises. And Goldfajn and Olivares (2000) document that in 1998 and 1999 Panama’s sovereign spread was considerably higher than that paid by Costa Rica, which maintained a flexible exchange rate system. One can dispute the evidence just presented on a number of counts. The most important one may be that the international comparison of sovereign spreads should take into account not only their exchange rate regimes but also the influence of other, country-specific characteristics. But the fact remains that evidence that dollarization is likely to reduce sovereign risk premia and dollar interest rates is still missing. Dollarization and Policy Credibility inally, one often hears that official dollarization would be beneficial because it would enhance the credibility of domestic policy. This contention is very difficult to evaluate, partly because the word “credibility” has been employed in many different senses and partly because there has been virtually no success at quantifying the size of the potential credibility gains. F 7. One may argue that the data in Table 1 are consistent with a different view: that economic characteristics determine exchange rate regimes and not vice versa. This case would hold if, in particular, low-inflation countries are better able to maintain pegged exchange rates. However, Ghosh and others (1997) investigate this “reverse causation” possibility and conclude that it makes little difference to their results. 8. For instance, Hanke and Schuler observe, “The major benefit of dollarization [in Argentina] would be reduced interest rates” (1999, 407). 9. This would be the case if, in particular, a government’s default is coupled with exchange restrictions and capital controls, limiting the private sector’s access to hard currency. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 9 One sense in which policy would be more credible is that official dollarization would be more difficult to reverse than other fixed exchange rate schemes. This contention has been most prominent in Argentina. There, the prevailing currency board system is widely regarded as having played a key role in lowering inflation in the 1990s. In spite of that success and repeated official announcements to the contrary, there remains a market expectation that the currency board may be abandoned. This market perception is evident from comparing dollar interest rates against Argentinean peso interest rates on otherwise similar financial securities. The spread between peso- and The costs of dollarization dollar-denominated eurobonds, for inare by and large identifiable stance, averaged 2.5 and can be quantified. On percentage points the other hand, the benefits during the 1997–98 period, indicating from dollarization remain to that international be demonstrated. investors demanded compensation for the risk of the government’s possibly abandoning the currency board and a subsequent devaluation of the peso. While the Argentinean government could end its currency board system virtually overnight if it wished to, it may be much more difficult to reverse official dollarization. Such a reversal would entail reintroducing national currency and, presumably, convincing domestic residents to turn in their holdings of dollars. Seen in that light, official dollarization would be a more “credible” arrangement than maintaining a currency board with an uncertain future. But a further question emerges: are there clear benefits from such marginal gains in credibility? Dollarization enthusiasts hope for a positive answer in at least three respects. The first is that, by eliminating the possibility of devaluation, enhanced credibility will help reduce interest rates. It has already been argued, however, that it is unclear whether such a promise will materialize. The second benefit from the irreversibility of dollarization is, in fact, that monetary policy will be taken out of the hands of the domestic central bank. Calling this development a benefit may sound paradoxical, but if the central bank cannot fully commit to its policy announcements then there is a benefit in taking control away. Specifically, it may be the case that the central bank has incentives to promise 10 low inflation ex ante but to engineer unexpectedly high inflation ex post. By generating surprise inflation, the central bank may stimulate the economy and increase employment and output. But private agents will soon understand if there is a systematic attempt to act in this way and will adjust their behavior accordingly, possibly not only offsetting the effectiveness of surprise inflation but also leading to a situation in which average inflation is suboptimally high. The importance of the argument just stated remains to be elucidated. While dollarization may reduce the inflation bias associated with the central bank’s lack of commitment, it can do so only at the cost of reduced policy flexibility. Policy flexibility is advantageous under some circumstances, as when the economy is hit by exogenous shocks. In addition, taking monetary policy away from domestic authorities, as dollarization does, is not the only way to overcome the commitment problem. An alternative is to structure the contracts of central bankers so as to eliminate their incentives to generate surprise inflation. Finally, it should be observed that in the last decade average inflation came down in most developing countries, regardless of their exchange rate regimes. Annual inflation in Peru was over 7,000 percent in 1990, the same year in which important economic reforms were enacted, including flexible exchange rates. Inflation then fell steadily and has remained in the single digits since the mid1990s. While other cases may not be as dramatic, the evidence shows that fixing exchange rates has not been necessary to reduce average inflation. A third reason that the irreversibility of dollarization may be beneficial is that it may enhance fiscal discipline. If a government is prone to lax fiscal behavior, the argument goes, dollarization may impose some discipline by making it more difficult for the government to finance excess fiscal behavior. It might do so by eliminating seigniorage revenues and inflationary finance and by forcing the government to issue only foreign currency debt. The fiscal discipline argument for dollarization has its own problems. Tornell and Velasco (1995) have pointed out that its theoretical underpinnings are myopic. Regarding fiscal discipline, they argue, dollarization differs from flexible exchange rates not in preventing lax fiscal behavior but in shifting its costs to the future. A government can always finance its expenditures today by borrowing. The cost of such a move would be a higher interest rate if exchange rates were flexible. Under dollarization, the interest cost may be lower but the fiscal expenditures have to be paid in some other way; in most cases, doing so means lowering Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 expenditures or raising taxes tomorrow. Which alternative acts as a better deterrent for a fiscally irresponsible government, therefore, depends largely on the rate at which the government discounts the future. An impatient government would, in fact, be more disciplined with flexible exchange rates, which impose immediate costs on lax fiscal behavior, than in a dollarized regime, which postpones the time of reckoning. In addition, the evidence is not supportive of the idea that more rigid exchange systems enhance fiscal discipline. As noted by Goldfajn and Olivares (2000), Panama’s annual fiscal deficit averaged 3.8 percent of GDP between 1970 and 1998, considerably higher than the deficits of Chile, Costa Rica, and Peru, where exchange rates were more flexible. And, in spite of its currency board, Argentina has had recent difficulties controlling its fiscal deficit: its fiscal deficit in 1999 is estimated at 3.8 percent of GDP. In sum, dollarization advocates are right in arguing that dollarization would be harder to reverse and in that sense more “credible” than a currency board or other exchange rate regimes. But whether that irreversibility would translate into actual benefits to the economy is more uncertain. Further, there has been virtually no attempt to quantify the net benefits associated with the credibility argument. Conclusion here is an old movie in which the leader of the world’s smallest country asks his advisers how to handle an economic crisis. One minister suggests, “Declare a war on the United States.” When the perplexed leader asks how that would help, the minister points out that nations like Germany and Japan fought a war against the United States, lost, and then became world powers. At first, the leader seems satisfied with this explanation. But then he has a doubt and asks, “And what happens if we win?” Like declaring war on the United States, official dollarization would impose considerable costs on a developing country. And as the discussion has shown, the costs are by and large identifiable and can be quantified. On the other hand, the benefits from dollarization remain to be demonstrated. As in the war story, it is not even clear what would happen if dollarization won. T REFERENCES BERG, ANDREW, AND EDUARDO BORENSZTEIN. 2000. “The Pros and Cons of Full Dollarization.” International Monetary Fund Working Paper WP/00/50, March. GOLDFAJN, ILAN, AND GINO OLIVARES. 2000. “Is Adopting Full Dollarization the Solution? A Look at the Evidence.” Pontifícia Universidade Católica do Rio de Janeiro. Unpublished paper. CHANG, ROBERTO. 1994. “Endogenous Currency Substitution, Inflationary Finance, and Welfare.” Journal of Money, Credit, and Banking 26 (November): 903–16. HANKE, STEVE H., AND KURT SCHULER. 1999. “A Monetary Constitution for Argentina: Rules for Dollarization.” CATO Journal 18 (Winter): 405–20. CHANG, ROBERTO, AND ANDRÉS VELASCO. 2000a. “Dollarization: Analytical Aspects.” New York University. Unpublished paper. HAUSMANN, RICARDO, MICHAEL GAVIN, CARMEN PAGÉS-SERRA, ERNESTO STEIN. 1999. “Financial Turmoil and the Choice of Exchange Rate Regime.” Inter-American Development Bank Working Paper 400. AND ———. 2000b. “Exchange Rate Policy for Developing Countries.” American Economic Review 90 (May): 71–75. MIRON, JEFFREY A. 1986. “Financial Panics, the Seasonality of the Nominal Interest Rate, and the Founding of the Fed.” American Economic Review 76 (March): 125–40. ———. 2000c. “Financial Fragility and the Exchange Rate Regime.” Journal of Economic Theory 92 (May): 1–34. TORNELL, AARON, AND ANDRÉS VELASCO. 1995. “Fiscal Discipline and the Choice of Exchange Rate Regime.” European Economic Review 39 (April): 759–70. FISCHER, STANLEY. 1982. “Seigniorage and the Case for a National Money.” Journal of Political Economy 90 (April): 295–313. GHOSH, ATISH R., ANNE-MARIE GULDE, JONATHAN D. OSTRY, AND HOLGER C. WOLF. 1997. “Does the Nominal Exchange Rate Regime Matter?” National Bureau for Economic Research Working Paper 5874, January. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 11 Credit Crunch or What? Australian Banks during the 1986–93 Credit Cycle E L L I S W. TA L L M A N N A R G I S B H A R U C H A A N D Tallman is a research officer in charge of the macropolicy section of the Atlanta Fed’s research department. Bharucha is a graduate student at the Woodrow Wilson School of Public and International Affairs, Princeton University. The authors thank Graham Anderson, Marianne Gizycki, Brian Gray, Frank King, Les Phelps, and Larry Wall for helpful comments and suggestions. They are especially grateful for extensive comments from Philip Lowe.1 C YCLES IN CREDIT HAVE IMPORTANT IMPLICATIONS FOR THE EFFICIENCY OF RESOURCE ALLOCATION AS WELL AS FOR DEVELOPMENTS IN THE MACROECONOMY. DURING THE UPSWING OF A CYCLE, RAPID CREDIT GROWTH CAN SUPPORT INCREASING ECONOMIC ACTIVITY AND RISING ASSET PRICES. CONVERSELY, boom can exacerbate an economic downturn if financial intermediaries become less willing to lend, thereby imposing further financial constraints on firms. When bank loan growth slows or contracts, is this change due to a reduction of loan supply or loan demand, or both? Assuming that the typical answer is “both,” what are the relative contributions of supply and demand factors for explaining a contraction in bank loan volume? Finding evidence that credit-supply constraints reduced loan issuance noticeably in addition to contractions in credit demand would suggest that the credit-supply constraints could add to the macroeconomic costs of an economic downturn. These questions have motivated a plethora of empirical studies using U.S. banking data to investigate them. The empirical literature has experienced a major resurgence especially since the observation of the U.S. credit cycle starting in the mid-1980s. LOAN LOSSES AND THE UNWINDING OF A CREDIT Specifically, many researchers have asked whether there was a “credit crunch” following the institution of the Basel risk-based capital standards—that is, whether bank capital supply-induced financial constraints led to observable restrictions on bank loan growth in the United States. Despite abundant empirical literature using U.S. banking data, these basic questions remain unsettled. Most of the research into supply-based lending contractions relates to the experience of U.S. banks. These studies seek to measure a relationship between indicators of bank financial condition and bank lending. Hancock, Laing, and Wilcox (1995) assert that large losses experienced by U.S. banks in the early 1990s implied a negative shock to bank capital. They find evidence that such shocks were a major factor underlying the observed reduction in bank lending. Bernanke and Lown (1991) also Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 13 demonstrate a link between bank capital and lending by conducting a cross-section study examining the effect of the ratio of capital to assets on banks’ subsequent loan growth. The study detects a significant relationship, indicating that a fall in the capital ratio reduces loan growth.2 Hancock, Laing, and Wilcox (1995) provide evidence not only that lending declines in response to an economic contraction but also that banks alter their portfolio composition.3 Kaufman (1992) supports this view, providing evidence that in periods of capital constraint banks tend to shift away from commercial lending and into residential mortgages. Yet Berger and Udell (1994) offer evidence Given that developments in and reasonable alterthe real economy can affect native hypotheses to suggest that the evithe financial sector, and dence supporting the that developments in the existence of important financial sector can affect supply-based forces on bank lending is not the real economy, monetary significant. policy must account for In the pursuit of changes in the pattern of more evidence to bear on the subject, recent financial intermediation. empirical work has examined banking data from other countries. Kang and Stulz (2000) examine whether banking shocks affected the performance of borrowing firms in Japan in the early 1990s. The results show that the firms with a higher proportion of bank loans performed worse than other firms. In contrast, Ongena, Smith, and Michalsen (1999) examine the Norwegian banking crisis (1988–91) and find evidence suggesting that bank distress had only a small impact on the real economy. More international evidence can be brought to bear on these questions. This article investigates the credit cycle during the late 1980s in Australia as additional evidence on whether supply factors are important to bank loan behavior. Along with other studies that use banking data from foreign countries, this article examines bank loan behavior as if it were an additional creditcycle observation, following the implementation of the Basel risk-based capital standards.4 There are key differences and similarities between the U.S. and Australian banking systems that can be evaluated up front, and the comparison allows a useful analysis of the Australian experience as it relates to the general economic issue of supply-based loan contraction. Specifically, this article analyzes the 1986–93 credit cycle in Australia, paying particular attention to the lending behavior of banks during the down14 swing of the cycle. While demand-side factors account for much of the credit cycle, evidence is presented consistent with the argument that supply-side elements also played a role. The study focuses on the examination of reporting data as collected by the Reserve Bank of Australia (RBA), in conjunction with information contained in banks’ annual reports. Analysis of the data suggests that banks subject to relatively high levels of impaired assets during the early 1990s experienced a relatively sharp decline in loan growth. The results are consistent with the proposition that losses weakened the condition of banks’ balance sheets and added a supply-based element to the contraction in lending. Australian banks dramatically altered the composition of their portfolios during this credit cycle. One interpretation of this observation is that, in response to strong demand for housing loans and weak demand for commercial loans, banks did so to boost capital ratios and improve the risk profile of lending. This portfolio shift was particularly noticeable for banks with weaker balance sheets. In part the shift was made possible by a reduction in nominal interest rates, which allowed the household sector to increase its borrowing. In addition, the concessional risk weighting of housing loans in the risk-based capital standards introduced in 1988 gave reasons for banks with weak balance sheets to move toward more housing loans.5 The improvement in the condition of banks’ balance sheets was also aided by the widening of margins on housing loans. However, this widening of margins also dampened credit growth because the high margins reduced the demand for loans. Overall, while evidence in the article indicates only the existence of a credit-supply channel, it constitutes a circumstantial case that the loan losses of the early 1990s played some role in retarding lending growth, particularly of commercial loans in Australia. Hence, supply-based contractions in aggregate loan growth in Australia may have contributed nontrivially to the length and macroeconomic cost of the observed credit contraction associated with the 1990–91 recession. Bank Condition and Lending: The Australian Experience eregulation of the Australian financial system in the 1980s reduced constraints on banks’ access to depositor funds and gave banks the discretion to make price and quantity decisions with regard to credit allocation.6 There was a view that access to a broader funding base and the ability to participate in a wider range of lending D Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 activities would eliminate the practice of creditrationing among banks (Committee of Inquiry 1981; Grenville 1991). The credit expansion of the late 1980s supported this view. Blundell-Wignall and Gizycki (1992) estimate supply and demand for business loans, finding no evidence of credit rationing during the 1980s. An alternative view emerged during the downturn of the credit cycle. Some business commentators argued that even in the deregulated environment of the 1990s, banks were restricting the supply of credit to creditworthy firms. The gist of the argument was that losses incurred by banks as a result of ill-fated loans made in the 1980s had forced banks to restrict lending to borrowers with otherwise viable investment opportunities. Despite the hypotheses offered in the popular press, there has been little research examining the relationship between bank lending and indicators of bank condition among Australian banks. This article is motivated by the notion that credit contractions initiated by economic downturns are exacerbated by banks’ subsequent reluctance to extend credit— that is, economic downturns can increase borrower defaults and impose large losses on banks’ consolidated operations. To the extent that banks must rely on capital to absorb these loan losses, and to the extent that raising additional capital is costly, they are forced to curtail future lending and to reassess the composition of their loan portfolios. The supply response of banks is mixed, differing across banks according to their respective capital positions. The techniques used in U.S. studies cannot be directly replicated using Australian banking sector data. First, although information on capital adequacy is available beginning in 1985, data on impaired assets for Australian banks are available only from TABLE 1 Shares in Bank Lending Markets (Percentage of Total; Average over 1990–96) Total loans Commercial loans Housing loans Personal loans Major Banks Regional Banks Foreign Banks 68.48 61.98 71.60 80.61 22.79 21.69 25.12 15.85 8.73 16.33 3.27 3.54 Source: Reserve Bank of Australia Bulletin 1990 onward, leaving few data points for timeseries analysis. Second, the Australian banking sector is highly concentrated relative to that of the United States; only thirty-one Australian banks operated continuously during the downturn in the 1986–93 credit cycle, compared with many thousands of banks in the United States (see Berger, Kashyap, and Scalise 1995, Appendix A, Table A1). Australia’s four major banks, Australia and New Zealand Banking Group (ANZ), Commonwealth Bank of Australia (CBA), National Australia Bank (NAB), and Westpac Banking Corporation (WBC), account for around 70 percent of total bank lending (see Table 1); the six U.S. money-center banks hold only around one-quarter of U.S. commercial bank assets. Therefore, data constraints for Australian banks, along with the dominance of the four major banks, limit the ability to replicate U.S. studies meaningfully. Although there is only limited scope for econometric analysis of the credit-supply channel in Australia, direct analysis of bank reporting data shows that Australian banks were subject to significant loan losses during the early 1990s and that this experience weakened their capital position. 1. This work was initiated while Tallman was a visiting senior research economist at the Reserve Bank of Australia. The views expressed are those of the authors and should not be attributed to the Reserve Bank of Australia, the Federal Reserve Bank of Atlanta, or the Federal Reserve System. 2. The effect is stronger for those banks that are capital-constrained, with capital ratios approaching the regulatory minimum. The regulatory minimum is determined by capital adequacy guidelines, discussed below in the section on impaired assets and capital ratios. Berger, Herring, and Szego (1995) also participate in the debate by asserting that the safety net associated with deposit insurance weakens the relationship between capital and lending. This assertion implies the existence of moral hazard problems, whereby banks covered by deposit insurance have less incentive to control risk exposures. 3. In addition, Hancock and Wilcox (1998) produce evidence that small banks in the United States shrank their loan portfolios more than large banks in response to declines in their bank capital. By examining explicitly the transmission mechanism linking small bank loans to small firms, this research aims more directly at the question of how the supply-based credit contractions affect the real economy. 4. Basel risk-based capital standards were instituted in Australia on August 23, 1988 (see Thompson 1991, 141). 5. The Basel risk-based capital standards recommend that a housing loan has a risk weight that is half the risk of a commercial loan. Hence, a bank faced with a choice between a commercial loan and a housing loan of the same amount would have to hold only half as much capital for the housing loan as for the commercial loan. 6. Fisher and Kent (1999) show that the financial market in Australia changed dramatically prior to the 1890 financial crisis. These observations suggest that intermediaries must learn to adjust to new powers and new environments. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 15 Extrapolating from U.S. results, these losses may have reduced subsequent loan growth and extended the contraction in credit beyond that associated purely with a fall in demand for credit. A subsequent section discusses this issue more fully. The 1986–93 Credit Cycle he 1986–93 credit cycle was Australia’s first cycle in a deregulated environment. The key features of this cycle, examined briefly in this section, include T • a large increase, and then fall, in the ratio of credit to gross domestic product (GDP); • the increasing importance of banks relative to other financial institutions; • large losses by banks in the downswing of the cycle; • a narrowing, and then widening, of lending margins; and • a fall in loan growth and changes in portfolio composition. The Ratio of Credit to GDP. The ratio of credit to GDP is a standard measure of intermediated credit as a proportion of the aggregate economy. In Australia, the ratio of intermediated credit almost doubled over the 1980s, reaching a peak for the 1986–93 credit cycle of 0.9 in June 1990. Chart 1 shows that the credit expansion occurred primarily in the second half of the 1980s following deregulation of the financial sector. The deregulation and innovation in the financial market may have lowered the price of intermediated credit. The expansion coincided with an asset price boom and relatively high inflation (Macfarlane 1991).7 Rising asset prices increased collateral values, thus improving the balance sheet position of borrowers and enabling an increased demand for credit, to which a deregulated Australian financial system was able to respond. In 1990–91, however, the economy slid into recession, and a sharp decline in credit followed. The ratio of credit to GDP fell for three years, and it took almost five years to reach its previous peak. Chart 1 includes subcategory measures of business, housing, and personal loans, each of which includes loans made by foreign subsidiary banks as well as domestic Australian banks. The breakdown shows that the 1986–93 credit cycle was driven primarily by the cycle in business credit. The ratio of business credit to GDP more than doubled over the 1980s, reaching a peak of 0.57 in June 1990. The downturn in aggregate credit also coincided with that of business credit. In contrast, the ratio of housing credit to GDP remained fairly stable over the 1980s but increased steadily during the 1990s as Australian households began to increase their comparatively low levels of debt (Stevens 1997). Although the downturn in the ratio of credit to GDP began in September 1991, the pace of credit expansion had begun to slow a couple of years earlier. Loan commitments are sometimes viewed as a signal for future lending growth. This pattern is observable in Australian data as well. Chart 2 shows that banks’ net lending commitments fell substantially in 1989 and were weak in nominal terms for the following two years, anticipating the weakness in bank loan growth that took place in 1990.8 The Importance of Banks. Although the economic expansion of the late 1980s stimulated strong lending growth across all financial intermediaries, the banking sector contributed most to the expansion in credit. Chart 3 shows real (inflation-adjusted) growth in lending by financial intermediaries.9 Real growth in lending by banks increased sharply between 1988 and 1990, briefly reaching annual rates of almost 30 percent and averaging 21.5 percent over the second half of the 1980s. This pace was considerably faster than the rate of growth in lending by all financial intermediaries and faster than the growth rate of M3 (which averaged 14 percent over the second C H A R T 1 Ratio of Credit to Nominal GDP 1.0 Total Loans Business Loans 0.6 Housing Loans 0.2 Personal Loans 1970 1980 Source: Credit measures, Reserve Bank of Australia Bulletin; GDP, Australia Bureau of Statistics 16 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 1990 C H A R T 2 Banks’ Net Lending Commitments Billions of Australian Dollars 50 Total Loans 30 Commercial Loans 10 1985 1989 1993 1997 Source: Reserve Bank of Australia Bulletin C H A R T 3 Real Growth in Lending by Financial Intermediaries Banks Percent 20 All Financial Intermediaries 0 1980 1990 1985 1995 Source: Reserve Bank of Australia Bulletin half of the 1980s), as banks financed some of their lending growth from abroad.10 Edey and Gray (1996) argue that deregulation led to a one-shot expansion in the financial sector and allowed banks to reassert their dominance as financial intermediaries. Prior to the mid-1980s, the assets of nonbank financial institutions grew rapidly relative to those of banks.11 Controls on bank deposit and lending rates as well as on asset composition had left banks at a competitive disadvantage (Battellino and McMillan 1989). The gradual removal of those controls saw this disadvantage diminish, with bank intermediation gaining strength over the 1980s. Between 1985 and 1990, total assets of banking institutions as a share of GDP rose from 57 percent to 88 percent (Edey and Gray 1996). The rapid growth in bank lending following deregulation has some historical antecedents. For Australia, Fisher and Kent (1999) describe the Banking Crisis of 1893; they note that in the years 7. It is notable that the unequivocal policy recommendation from Boyd and others (2000) is that banking crises tend to occur in environments of high inflation. They note that even predictable inflation is unhealthy for the financial system. 8. Net lending commitments are defined as bank offers to provide finance, minus cancellations of commitments; the duration of commitments differs across borrowers. Between 1986 and 1993, banks’ net lending commitments averaged around 40 percent of their loans outstanding to the private sector. 9. Lending by financial intermediaries comprises loans, advances, and bills held with the private nonfinancial sector. 10. Growth in bank lending and M3 is adjusted for the conversion of building societies (institutions similar to U.S. savings and loans prior to 1989) to banks. M3 is currency plus bank deposits of the private nonbank sector, excluding commonwealth and state government deposits and interbank deposits. 11. In many cases banks established nonbank subsidiaries in order to bypass regulatory constraints. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 17 prior to the crisis there was a notable growth in the assets of intermediaries that could effectively compete with trading (note-issuing) banks. In response, they argue, banks then began acquiring a riskier portfolio to maintain profitability. Given the seriousness of the banking collapse of 1893 in Australia, it seems clear that banks were relatively unprepared for the rigorous credit assessment necessary for maintaining risky portfolios profitably. More recently, the savings and loan institutions in the United States grew rapidly upon gaining new powers in the early 1980s, and most Americans are well aware of the requirements needed to recover from the loan losses of savings and loan institutions. The rapid growth in the loans among both U.S. savings and loan institutions and Australian banks in the 1980s exposed the underdevelopment of their respective credit-assessment skills at that time. In both cases, the institutions had operated in a constrained environment for many years and hence had devoted insufficient resources to credit assessment and the pricing of risk (see Ullmer 1997 for a discussion of the Australian case). As a result, when deregulation relaxed constraints, Australian banks were not well positioned to manage credit risk. With pressure to regain market share, banks may have extended loans that under other circumstances they would not have made. Some anecdotal evidence in support of this view is contained in ANZ’s 1992 annual report: “Undoubtedly, there was some imprudent lending during the boom period of the late 1980s, particularly in the small and medium business sectors, where the battle for market share following deregulation was hardest fought” (1992, 2). The end result was rapid growth in bank lending and increasingly risky loan portfolios. Large Losses for Banks in the Downturn of the Credit Cycle. When the downturn in economic activity occurred, the previous rapid loan growth came home to roost, so to speak, and indicated, ex post, the inadequacy of existing credit assessment. There was a substantial increase in the level of banks’ nonperforming loans accompanied by sharply falling profits and a sharp deterioration in the average return on shareholders’ funds, as shown in Chart 4. In fact, the average return on shareholders’ funds in 1991–92 was negative, at –1.8 percent. One indicator of Australian banks’ exposure to nonperforming loans is the ratio of impaired assets to capital. The data presented here are for net impaired assets (total impaired assets less provisions held against specific loans; see Appendix A for more details). This is a standard measure of the vulnerability of banks’ capital to problem assets. 18 Chart 5 presents this measure for major, regional, and foreign banks.12 The early 1990s recession in Australia contributed to the sharp increase in impaired assets across all bank categories. For foreign and regional banks, net impaired assets during 1991 were equivalent to their entire capital base. More importantly, the net impaired assets of the major banks also reached high levels in 1991, covering more than 60 percent of their capital base. Major bank impaired assets remained high for a relatively long period; it was not until late 1994 that the ratio of net impaired assets relative to capital fell below 0.2. Bank Lending Margins. Amid sizable loan losses and a falling cash rate in the early 1990s, the margin between bank lending rates and the cash rate (the cost of funds) increased. The top panel of Chart 6 shows the difference between the mortgage rate and the cash rate and the difference between the business indicator rate and the cash rate.13 Some background on the institutional structure of the Australian financial market will help clarify the significance of these figures. The cash rate in Australia is analogous to the federal funds rate in the United States; the cash rate represents the marginal cost of funds to the Australian banking system. Also, the cash rate is the rate of interest set by the Reserve Bank of Australia. During the 1986–93 period, the typical mortgage in Australia did not have a fixed interest rate. Instead, banks typically adjusted the mortgage rate based on market conditions, namely, some premium over the cash rate. In essence, during the time period for the data in this study, the mortgage rate paid in Australia was set periodically by the bank holding the mortgage. Only somewhat later did competitive pressure arising from new mortgage providers reduce the power of banks to dictate the mortgage rate. The differential between lending rates and the cash rate narrowed in the late 1980s as the cash rate was increased and competitive pressures within the financial sector also increased. Based on the business indicator rate, the differential associated with commercial loans averaged around 2.5 percent through the late 1980s but then widened to around 4 percent in 1991 and remained at this level for almost three years. The housing-loan interest rate spread followed a similar pattern over the 1990s, also peaking at a little over 4 percent. Part of the explanation for this widening of interest rate margins (measured relative to the cash rate) is that while the cash rate fell substantially over this period, banks’ average cost of funds did not fall to the same extent (see Lowe 1995 and Reserve Bank of Australia 1992). To maintain average margins, the Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 CHART 4 Average Return on Bank Shareholders’ Funds (Major Banks) Percent 20 10 0 1980 1990 Source: Reserve Bank of Australia Bulletin; annual reports of major banks, various issues C H A R T 5 Ratio of Net Impaired Assets to Capital Regional Banks 1.0 0.6 Majors Banks Foreign Banks 0.2 1990 1995 Source: Internal reports to Reserve Bank of Australia (Prudential Statements) difference between lending rates and the cash rate widened. One effect of this development was that it made the writing of new loans, particularly housing loans, more profitable.14 Undoubtedly, this increased profitability was a factor underlying the rebound in average returns on shareholders’ funds, which returned to almost 15 percent only two years after reaching negative levels in 1991–92. Contraction in Lending and Large Portfolio Shifts. Although the asset portfolios of major, regional, and foreign banks exhibit the same general response to losses in the early 1990s, there are differences in the timing and magnitude of lending responses across these categories. Chart 7 presents growth in total loans and commercial loans across bank categories, adjusted for breaks in lending 12. The sample consists of the thirty-one banks that traded continuously during the downturn in the 1986–93 credit cycle. There are four major banks, eleven regional banks, and sixteen foreign banking subsidiaries; their shares of total bank lending are 70 percent, 20 percent, and 10 percent, respectively. For more details, see Table 1 and Appendix B. 13. The standard variable rate on bank housing loans is taken as a measure of the mortgage rate on housing loans. The commercial lending rate is given by the business indicator rate on banks’ large, variable-rate business loans. 14. High levels of profitability also encouraged mortgage managers to enter the market and contributed to the narrowing of margins evident in the mid-1990s. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 19 C H A R T 6 Bank Lending Margins (Major Banks) Business Indicator Rate Less Cash Rate Per cent 4 2 Mortgage Rate Less Cash Rate 0 1990 1995 Per cent 18 14 Cash Rate 10 6 1990 1995 Source: Reserve Bank of Australia Bulletin series. These breaks are the result of bank reclassifications and the transfer of loans to or from intermediaries outside the category under consideration. Foreign banks suffered the most rapid and extreme decline in total lending, with loan growth becoming negative in 1992.15 Growth in regional banks’ total loans also fell during 1991 but fluctuated at positive rates of between 10 percent and 15 percent. Despite the fact that growth in major banks’ total loans remained positive, reaching a trough of 2 percent in 1992, major bank lending only began a sustained recovery in 1994. The chart shows that the decline in commercial loan growth was more uniform across bank categories. Growth in commercial lending suffered a more extreme fall than growth in total loans. During 1991, growth in commercial loans fell from 25 percent to around 10 percent for both regional and foreign banks. Foreign banks’ commercial loan growth became negative in 1992, with the decline in lending continuing until 1994. Major banks also reduced commercial lending in the early 1990s, with growth becoming negative in late 1992. Growth in commercial lending remained weak for an extended period, only showing signs of recovery in 1995. 20 Sizable changes in bank loan growth over the 1986–93 credit cycle were accompanied by substantial shifts in the composition of banks’ lending portfolios. Chart 8 presents the loan portfolio composition of major and regional banks, respectively. Portfolio shares in commercial, housing, and personal loans combined account for almost 100 percent of total loans, the residual being lending to government. Sharp changes in portfolio shares, such as that of commercial loans in the regional bank portfolio in 1992, can be explained by breaks in lending series. These breaks are listed in Appendix C. For major and regional banks, the most notable trend is the steady decline in the share of commercial lending and the corresponding increase in the share of housing loans. By 1995, housing loans accounted for more than 50 percent of major bank lending, compared with 40 percent in 1992. The shift toward housing loans was even more marked for regional banks, with the share of housing loans increasing from around 40 percent in 1992 to a share of 65 percent by the end of 1996. This shift from commercial loans to housing loans following large losses is consistent with the analysis of U.S. banks presented in Kaufman (1992) and Hancock, Laing, and Wilcox (1995). But, as will be discussed Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 CHART 7 Loan Growth across Bank Categories (Percentage Change over Four Quarters) Commercial Loans To t a l L o a n s Per centage Change 40 Foreign Regional 20 Regional Major 0 Major Foreign 1991 1995 1991 1995 Source: Reserve Bank of Australia Bulletin CHART 8 Portfolio Composition of Major Banks and Regional Banks (Percentage of Total Loans) Major Banks Regional Banks Per centage 70 Housing 50 Housing 30 Personal Commercial 10 Commercial Personal 1990 1990 1995 1995 Source: Reserve Bank of Australia Bulletin below, the shift toward housing loans was further reinforced by the lower risk-weighting given these loans by the Basel standards. The 1986–93 credit cycle was one of the more pronounced cycles experienced in Australia’s history. In part, it can be explained as a by-product of the transition from a highly regulated financial system to a deregulated system. The cycle saw banks contribute significantly to the increase in lending during the late 1980s but incur large losses in the downswing of the cycle. There was a subsequent widening of lending margins and a decline in banks’ loan growth during the early 1990s. Furthermore, there was a distinct shift in banks’ loan portfolio composition away from commercial lending and into housing loans. The analysis in the next section suggests that the characteristics of the 1986–93 credit cycle are consistent with the existence of a credit-supply channel. The Credit-Supply Channel hile the downturn in the credit cycle saw a significant fall in banks’ loan growth and major changes in the structure of Australian bank portfolios, it is not immediately clear whether these changes were driven by the demand for credit or a combination of demand and supply factors. The discussion now turns to the relationship between loan losses and capital positions across individual banks and the subsequent changes in those banks’ portfolios. Analysis is predicated on the assumption that all banks face broadly similar demand conditions so that differences in the size and composition of bank assets reflect decisions W 15. Some observers viewed the losses among these banks as a transfer from overseas banks to Australian borrowers. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 21 made by individual banks. Factors underlying these decisions are loosely referred to as supplyside influences. It is important to acknowledge the difficulty in identifying credit-demand versus credit-supply channels. For example, the fall in banks’ net lending commitments beginning in 1988 no doubt reflects a fall in demand for credit due to high interest rates. However, reduced commitments may in part also be a supply-side initiative, with banks recognizing that the prevailing growth in lending was not sustainable and that high interest rates would impose financial constraints on borrowers. Similarly, alternative explanations can be offered for the extended weakness in commercial lending commitments during the first half the 1990s (see Chart 2). Restructuring of business and substitution away from debt financing undoubtedly reduced demand for commercial credit, but substantial loan losses may also have reduced banks’ willingness to make relatively high-risk commercial loans. The Commonwealth Bank describes the contribution of demand and supply factors: “It is important to stress that the Bank (CBA) remains willing to lend—it is the lack of demand and, to a lesser extent, the absence of viable proposals, that is determining current lending levels” (CBA 1992, 7). The existence of the creditdemand channel is widely accepted, and, although far from conclusive, evidence does suggest that the credit-supply channel was active during the downswing of the 1986–93 credit cycle. This article analyzes primarily the data for Australia’s four major banks because these banks account for the bulk of bank-intermediated lending, provide relatively consistent data, and have comparable market shares (see Table 2). Also, the assumption that demand conditions are broadly similar across banks is probably more accurate for the major banks, each having national coverage and extensive branch networks, than for a more diverse grouping of financial institutions. Even so, the assumption is unlikely to hold exactly. Large loan losses are often the result of borrowers who take out large loans and default on loan repayments. To the extent that customer markets exist, demand for loans from a bank with large loan losses might be significantly reduced because defaulting borrowers are unlikely to be borrowing additional funds. On the other hand, major banks have a long-established history in financial intermediation and they have banking relationships with a wide range of clients, so if one group of customers has reduced demand for loans, new customers can be attracted from other banks. Facts on the Major Banks. In order to examine the relationship between loan losses and subsequent lending behavior, measures of bank condition—namely, impaired assets and capital—are obtained from the major banks’ annual reports while lending data are those collected by the RBA.16 Impaired Assets and Capital Ratios. Data on impaired assets reflect the actual and potential losses of a bank. Up until 1994 each bank tended to disclose a slightly different measure of impaired assets in its annual report.17 Although the figures are not strictly comparable across banks and may over- or understate the true level of problem assets, they are presented here as an indication of the relative size of problem loans across major banks and the change through time in problem assets within each of the major banks. Chart 9 shows that for each of the major banks, the ratio of net impaired assets to capital peaked in 1991 or 1992. Each of these banks had nonbank subsidiaries that contributed to losses.18 WBC experienced the most marked deterioration in asset quality with net impaired assets in 1992 being equivalent to around 70 percent of its capital base. These problem loans contributed to a net after-tax loss of $1,562 million in 1992. ANZ also suffered high levels of net impaired assets, which in 1991 amounted to almost 60 percent of capital. The peak in ANZ’s impaired assets was also associated with a net aftertax loss of $579 million in 1992. Although CBA’s net impaired assets also covered a substantial proportion of its capital base in 1991, CBA managed to maintain positive after-tax profits. More notable, however, is the performance of NAB, whose net TABLE 2 Shares in Major Bank Lending Markets (Percentage of Total; Average over 1990–96) Total loans Commercial loans Housing loans Personal loans ANZ CBA NAB WBC 20.69 25.71 17.60 18.79 28.84 24.59 34.58 23.75 25.92 29.65 20.27 32.94 24.55 20.06 25.55 24.51 Source: Internal reports to Reserve Bank of Australia (Prudential Statements) 22 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 CHART 9 Measure of Major Bank Condition ANZ (as of September 30) CBA (as of June 30) Ratio of Net Impaired Assets to Capital (RHS) 0.5 Ratio Per cent 15 10 0.1 Capital Ratio (LHS) 1990 1990 1995 NAB (as of September 30) 1995 WBC (as of September 30) 0.5 Ratio Per cent 15 10 0.1 1990 1995 1990 1995 Note: Capital ratio is percent of total risk-weighted assets. Source: Annual reports of major banks impaired assets were modest in comparison, peaking at 30 percent of capital in 1991. The high levels of impaired assets experienced in 1991 and 1992 reflect a concerted effort on the part of banks to undertake a one-time write-off of problem loans in order to improve their balance sheet positions. This action would in part explain the rapid decline in net impaired assets after 1992, particularly for WBC and ANZ, whose ratio of net impaired assets to capital was halved within two years of its peak. In addition to impaired asset data, Chart 9 presents capital ratios across the major banks as another measure of bank condition. Conforming with international supervisory arrangements established by the Basel Accord, Australian banks have been subject to minimum capital requirements since August 1988. Capital adequacy guidelines stipulate that each Australian bank is expected to maintain a minimum ratio of capital to risk-weighted assets of 8 percent. Capital base is defined as tier 1 capital plus tier 2 capital, less goodwill and future income tax benefits. For detailed definitions see Appendix A. Risk-weighted assets are calculated by applying a 0 percent weight on gold and cash balances with the RBA, a 10 percent weight on federal and state 16. For confidentiality reasons, present capital adequacy and impaired asset data on individual banks as collected by the RBA are not presented. The measures taken from bank annual reports are year-end figures and do not reflect withinyear variability. 17. In September 1994 the RBA provided banks with a set of guidelines for the definition of impaired assets, facilitating consistent reporting across banks. Impaired assets are defined as the sum of nonaccrual items, restructured items, other real estate–owned items, and other assets acquired through enforcement. See Appendix A for further details. 18. These losses are attributed to major banks’ large exposures in the commercial property market. Conroy (1997) argues that in many cases subsidiaries failed to properly consult the lead bank when increasing their stake in property development. WBC subsidiaries Australian Guarantee Corporation (AGC) and Partnership Pacific Limited (PPL) made losses of $107 million and $146 million, respectively, in 1992. PPL also made losses of this magnitude in 1990 and 1991. ANZ’s finance company, Esanda, made losses of $139 million in 1992. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 23 government securities and claims on governments and central banks of Organisation for Economic Cooperation and Development (OECD) countries, a 20 percent weight on local government securities and claims held against Australian and OECD banks, a 50 percent weight on mortgage-backed lending, and a 100 percent weight on commercial and foreign assets. Personal loans that are not mortgage-backed also attract a 100 percent risk weight. Lower risk weights are associated with assets that are typically assumed to be less subject to credit risk. Chart 9 shows that the peak in impaired assets in 1991 and 1992 generally coincided with reductions in capital ratios. The coincident peaks are most apparent for both WBC and ANZ, whose comparatively large losses were associated with a fall in their capital ratios of almost 1 percentage point, to 9.7 percent and 9.0 percent, respectively. The fall in CBA’s capital ratio was of lesser magnitude while NAB maintained a fairly constant capital ratio of around 11.5 percent in 1991 and 1992. After 1992, capital ratios increased quite quickly. In the case of WBC, capital relative to risk-weighted assets increased from 9.7 percent in 1992 to 13.8 percent in 1994. ANZ, being subject to the second-largest losses among the major banks, also significantly increased its capital ratio between 1992 and 1994, from 9.0 percent to 11.3 percent. As the following equation shows, changes in a bank’s capital ratio can be decomposed into three elements: changes in the bank’s capital, changes in the bank’s total assets, and changes in the composition of those assets. The capital ratio is defined as k= K , A* (1) where K is the capital base and A* is risk-weighted assets. For simplicity it is assumed that there are two assets and that one of the assets has a concessional risk weight of θ while the other has a risk weight of one. Therefore, A* = θ A1 + A2 0 ≤ θ < 1, (2) where A1 and A2 are the two assets. This calculation can also be expressed as A* = [1 – w(1 – θ)] A, (3) where A is total assets ( A1 + A2 ) and w is the portfolio share of the asset attracting the lower risk weight. Substituting (3) into (1) and totally differentiating yields 24 dk = ( ) 1−θ . dk − k A + k A* 1 − w 1 − θ ( ) dw , (4) where a dot (·) above a variable denotes a percentage change. Equation (4) implies that the capital ratio will increase if capital increases, if total assets decline, or if there is a portfolio shift toward the asset with the concessional risk weight. The lower the concessional risk weight (θ) is, the larger the effect will be on the capital ratio of a given change in the structure of a bank’s portfolio. As will be discussed below, portfolio reallocation played an important role in the improvement of capital ratios among Australian banks in the 1990s. Lending Behavior. Chart 10 shows total loan growth and commercial loan growth for each of the major banks during the 1990s. Consistent with data published by the RBA, total loans are defined as lending to the nonfinancial sector. The lending figures presented here are those of the lead bank, and all growth rates are break-adjusted. The most notable adjustments occur for CBA in January 1991, given its acquisition of State Bank of Victoria, and for WBC in October 1996 when its subsidiary, Challenge Bank, transferred the bulk of its loans to the lead bank. As Chart 10 indicates, following the peak in impaired assets in 1991 and 1992, each of the major banks experienced substantially slower loan growth. The contraction in commercial lending was generally more extreme and relatively extended compared with the contraction in total loans. Slow growth in major bank lending can in part be explained by reductions in net lending commitments, which occurred prior to the realization of loan losses.19 Having absorbed the largest losses among the major banks, WBC suffered the deepest and most extended contraction in commercial lending. The contraction in lending, however, was preceded by a brief period of rapid loan growth in 1992. This growth spike largely reflects changes in WBC’s holding of bank bills in 1991 and 1992. Excluding the “bills held” component of lending shows that commercial loan growth was negative in 1992 (see Appendix A for more details). Consistent with the established definition of commercial lending (including bills held), Chart 10 shows that growth in WBC commercial loans was consistently negative between 1993 and 1995, averaging around –8 percent. The contraction in ANZ commercial lending was less severe but followed the same broad pattern as that of WBC. However, growth in ANZ total loans was weaker than that of Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 CHART 10 Growth in Major Bank Lending (Percentage Change over Twelve Months) CBA ANZ Percentage change 30 Total Loans 10 Commercial –10 1991 1995 1991 WBC NAB Per centage change 1995 30 10 –10 1991 1995 1991 1995 Source: Reserve Bank of Australia collected data WBC. Total lending by CBA recovered slowly from 1993 onward. The contraction in CBA commercial lending was of a smaller scale than that experienced by WBC and ANZ, but the weakness in loan growth was relatively extended. Compared with the other major banks, NAB lending emerged strongly from the economic downturn, with a moderate slowdown in commercial lending. NAB total loan growth fell in 1991 but rebounded in late 1993, reaching 20 percent within a year; NAB commercial loan growth was positive and increasing during 1994, despite the contraction in commercial lending experienced by the other major banks. Charts 9 and 10 combined suggest that the banks that suffered most acutely from large loan losses in the early 1990s, namely, WBC and ANZ, were subject to deeper and more extended contractions in commercial lending as they were rebuilding capital ratios. Composition of Loan Portfolios. Australian banks made a distinct substitution out of commercial lending and into housing loans in the aftermath of the recession of 1990–91. Consistent with the aggregates in Chart 8, Chart 11 shows a marked shift in portfolio composition among individual major banks.20 Those banks subject to the highest impaired asset levels made an aggressive shift out of commercial lending and into housing loans. For example, housing loans accounted for almost 60 percent of the WBC lending portfolio by 1995, compared with 40 percent in 1992; the share of commercial loans fell from almost 40 percent to 25 percent over this period. Similarly, by 1995 housing loans accounted for 50 percent of the ANZ loan portfolio, compared with 35 percent in 1992, with the share of commercial loans falling from around 50 percent to 40 percent. CBA and NAB have also 19. Net lending commitments across the major banks are consistent with trends shown in Chart 2. Commitments turned down in 1988 and remained relatively weak for commercial lending in particular. Commitments data also confirm that weak loan growth was not simply the result of nonperforming loans being written off. Net lending commitments made by banks in weak condition were relatively low between 1991 and 1995, especially in the market for commercial loans. In contrast, NAB commitments recovered steadily from 1992 onward. 20. Once again, the sum of portfolio shares in commercial, housing, and personal loans accounts for almost 100 percent of total loans. Sharp changes in portfolio shares can be explained by breaks in lending series. These breaks are listed in Appendix C. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 25 CHART 11 Portfolio Composition of Major Banks (Percentage of Total Loans) CBA Per centage ANZ 50 Commercial Housing 10 Personal 1990 1995 Per centage 1995 1990 NAB WBC 50 10 1990 1990 1995 1995 Source: Reserve Bank of Australia collected data devoted an increasing share of their portfolios to housing loans over the 1990s, with a moderate substitution away from commercial lending. Some Evidence of a Credit-Supply Channel. Chart 12 relates impaired assets across the four major banks to average growth in risk-weighted assets in the two years after the peak in impaired assets and average growth in total loans over the same period. Average growth in risk-weighted assets is calculated based on annual report data, that is, publicly available information, while growth in total loans is based on data reported to the RBA. Impaired assets, rather than net impaired assets, are taken as the point of reference given that exposure to problem loans (provisioning aside) is likely to prompt an internal review of lending policies. Chart 12 makes it apparent that the banks with a higher ratio of impaired assets to capital generally experienced relatively large contractions in riskweighted assets and loan growth, observations generally interpreted and viewed as a shift by banks out of risky assets. The left-hand panel of Chart 12 shows that WBC, being most exposed to impaired assets, experienced an annual change in risk-weighted assets of around –12 percent in the two years following the peak in impaired assets. Only NAB experienced a significant increase in risk-weighted assets, with 26 growth averaging almost 10 percent per annum following the peak in impaired assets. The right-hand panel of Chart 12 shows that total loan growth remained positive for each of the major banks but was particularly strong for NAB, it having been the least exposed to impaired assets. Given that average growth in total loans for each of the major banks was positive, any reduction in riskweighted assets was due to changes in banks’ balance sheet structure—a shift toward assets that attract a concessional weight in the calculation of risk-weighted assets. Thus it is not surprising that banks with the largest losses, and subsequent contraction in risk-weighted assets, exhibited the largest shift toward assets with a concessional risk weight. Chart 13 relates impaired assets across the major banks to the percentage point change in the ratio of housing loans to total loans two years after the peak in impaired assets. As suspected, those banks with the largest losses (ANZ, WBC) shift most aggressively toward assets with concessional risk weights. The chart shows that WBC, with the most sizable reduction in risk-weighted assets, undertook the most significant shift in portfolio composition. WBC, closely followed by ANZ, substituted most strongly into housing loans, which attract a risk weight of 50 percent as opposed to 100 percent on commercial loans. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 CHART 12 Major–Bank Asset Growth Two Years after the Peak in Impaired Assets (Average Annual Growth) R i s k - We i g h t e d A s s e t s To t a l L o a n s 10 10 NAB Per cent NAB 0 6 ANZ CBA WBC ANZ CBA 2 –10 WBC 0.4 0.8 0.6 1.0 0.6 0.4 1.0 0.8 Peak in Impaired Assets Relative to Capital Source: Annual reports of major banks; Reserve Bank of Australia collected data CHART 13 Percentage Point Change Major–Bank Portfolio Shifts Two Years after Peak in Impaired Assets (Ratio of Housing Loans to Total Loans) 14 WBC ANZ 10 NAB 6 CBA 2 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Peak in Impaired Assets Relative to Capital Source: Annual reports of major banks; Reserve Bank of Australia collected data WBC and ANZ also experienced the largest corresponding shifts away from commercial lending, with the portfolio share of commercial loans falling around 9 percentage points and 8 percentage points, respectively (see Chart 11). NAB made a strong shift toward housing loans while the portfolio share of commercial loans declined by 4 percentage points.21 Using the simple model discussed above, one can obtain a rough guide to the importance of the shift from commercial lending into housing loans in improving capital ratios. It is assumed that housing loans initially account for 40 percent of total loans (that is, w = 0.4), rising to 60 percent (so that dw = 0.2). The concessional weight on housing loans is known to be 50 percent (that is, θ = 0.5). Furthermore, the initial capital ratio (k) is assumed to be 0.1, with the capital . base and total assets held constant (dK = 0 and A= 0). The resulting change in risk-weighted assets is . . 1−θ A* = A − 1 − w 1 − θ 1 =− dw 2−w ( ) dw (5) = −0.125 21. CBA’s impaired assets peaked in 1991, but it is not until 1994 that the shift toward housing loans became apparent. As such, Chart 13, which is based on lending two years after the peak in impaired assets, does not capture this shift. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 27 Thus, the assumed change in portfolio composition reduces risk-weighted assets by 12.5 percent. With no change in capital base or total assets, equation (5) implies that an increase in the portfolio share of housing loans from 40 percent to 60 percent will increase the capital ratio from 10 percent to 11.25 percent. For banks that suffered extreme loan losses and weakened capital condition, the more marked substitution toward housing loans served to reduce risk-weighted assets, thereby boosting capital ratios. More fundamentally, however, loan losses caused banks to reassess portfolio allocations among assets in different risk classes. Banks were apparently less willing to absorb the heterogenous risks associated with commercial loans when lowrisk housing loans could be issued at similar margins. This approach was reinforced by the fact that mortgage-backed loans attract a concessional risk weighting in the calculation of risk-weighted assets. Given that the risk-based capital standards were introduced in 1988, it does not appear that the introduction of those standards alone generated the dramatic portfolio shift, but it was clearly a contributor. In essence, the question becomes one of whether banks altered risk-weighted assets to maintain capital ratios or whether large losses motivated banks to become more active risk managers when making loans. Both reflect supplybased credit responses. Interpretation of the Evidence ollowing the economic downturn and the collapse of the credit boom in the early 1990s, there was an extended contraction in bank lending among Australian banks. In large part this reaction reflected the unwinding of rapid increases in corporate leverage during the 1980s. High interest rates, high leverage, and weak economic growth saw a significant decline in the demand for loans. However, there is some indication that the contraction in lending was reinforced by supply-based constraints. There is evidence of a relationship between measures of bank condition and bank lending, with those banks subject to the largest loan losses having experienced larger-than-average declines in lending growth and substantial changes in the composition of their portfolios. This outcome is consistent with the view that the decline in loan growth in part reflected banks’ reluctance to lend in light of the deterioration in the quality of their balance sheets. Although WBC, for example, stresses that it in no way sought to restrict lending, its 1991 annual report states, “Following the rapid asset growth of the 1980s, we have managed our balance sheet very F 28 tightly, a process more recently facilitated by the reduction in credit demand in Australia” (1991, 5). Reduced loan growth may also be a decision on the part of banks to control risk-weighted assets in order to lift capital ratios. Weak loan growth was combined with a change in the composition of banks’ loan portfolios away from commercial lending and into housing loans. In large part this shift was made possible by strong demand for housing loans as the economy emerged from recession. The portfolio shift was, however, more marked for banks in relatively weak balance sheet condition. The ability of banks to undertake large changes in portfolio composition was an important element in the recovery of banks’ balance sheets. This portfolio shift had two effects. First, it allowed banks to increase capital ratios by reducing risk-weighted assets given that housing loans attract a lower risk weight of 50 percent compared with the 100 percent risk weight on commercial loans. Second, portfolio substitution allowed banks to reduce their exposure to commercial loans, which are inherently riskier than housing loans. More specifically, however, the shift toward housing loans also stimulated a recovery in the profitability of Australian banks. Increased margins on lending helped restore profitability following the loan losses of the early 1990s. Between 1992 and 1994, margins on housing loans reached their highest level at around 4 percent, equal to the margin on commercial loans. In part, the portfolio reallocation undertaken by banks was made possible by falling nominal interest rates. The consequent reduction in mortgage lending rates, to interest rate levels well below those of the 1980s, encouraged households to increase their indebtedness and hence augmented their demand for housing finance (see Stevens 1997).22 Australian banks were fortunate that there was a ready demand for low-risk, high-yielding housing loans. Meeting this demand meant that banks were able to generate higher returns with lower risk while simultaneously improving their capital ratios. Banks in other countries, such as the United States and Japan, also faced large loan losses during the 1990s. However, banks in these countries were not in the same fortunate position as the Australian banks. In the United States, banks were forced to shift into lower-yielding Treasury securities for some time before their balance sheets recovered. Japanese banks are still working to recover. In contrast, Australian banks enjoyed a strong demand for a concessionally weighted asset that was also relatively high-yielding, thereby speeding the recovery of Australian banking. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 Conclusion his article analyzes the relationship between measures of bank condition and bank lending during the downswing in the Australian credit cycle of 1986–93. The concentrated nature of the Australian banking sector and a short data history mean that conclusions are based on direct observation of the data rather than econometric tests. While changes in the demand for loans clearly account for much of the cycle in credit growth, the analysis describes evidence consistent with the hypothesis that the sizable losses Australian banks incurred in the early 1990s played at least some role in constraining the availability of funds for commercial lending. The basic conclusion from the Australian experience is that there is a relationship, albeit a relatively weak one, between the loan loss experience of the early 1990s and subsequent lending behavior. In essence, the conclusions imply that the larger the losses of a particular bank are, the slower its subsequent loan growth tends to be and the larger the change in its balance sheet structure toward housing loans tends to be. This description of the Australian data is consistent with the view that a bank’s recent profitability (and its capital position) can affect its lending decisions. The bank losses also contributed to an increase in interest rate margins, which in turn constrained the demand for loans. The fall in lending, combined with the portfolio reallocation away from commercial lending and into housing loans, served to increase capital ratios by reducing risk-weighted assets. The shift toward housing loans also gave banks the opportunity to restore profits with a relatively low-risk, high-yielding asset, given T that margins on housing loans were similar to those on commercial loans. What is less clear is whether the reduced supply of loans from banks with recent weak performance was compensated for by additional loans from banks with stronger profit performance. However, even if other banks did compensate through an increased supply of loans, the widening of lending margins that followed the banking problems is still likely to have reduced loan demand and thus credit growth. Australian bank lending between 1986 and 1993 is of particular interest because it was the first credit cycle following financial deregulation in that country. Emerging from a regulated era, Australian banks had limited experience in managing portfolios with a rapidly increasing proportion of risky commercial loans. Batellino and McMillan (1989) argue that deregulation encouraged banks to be active liability managers. However, deregulation and the following credit expansion required that banks also become active asset managers; they were now exposed to risks that had not been on the balance sheet during the regulated era. As numerous examples in a range of countries over the past decade illustrate, one can argue that large losses by financial institutions can exacerbate economic downturns. Given that developments in the real economy can affect the financial sector, and that developments in the financial sector can affect the real economy, monetary policy must account for changes in the pattern of financial intermediation. Understanding the linkages between the real economy and the financial sector remains a major challenge for macroeconomic policymakers. 22. In addition, relatively low inflation ensured that repayment burdens in the early phase of a loan were lessened, making borrowing more feasible for low-income earners. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 29 A P P E N D I X A Construction of Banking Data Bank Lending Measures of Bank Condition B The Capital Adequacy Return and the Impaired Assets Return are used to construct measures of bank condition for major, regional, and foreign bank categories.1 Both returns are quarterly, and both are completed on a consolidated group basis. (For confidentiality reasons, annual report data are used when presenting capital ratios and impaired assets for individual banks.) The Capital Adequacy Return is used to construct the following series: alance sheet data are used to generate lending series based on banks’ reporting to the RBA in Form D. Form D is a weekly return that covers assets and liabilities on Australian books. These data are used to construct the following series: • Total loans, defined as lending to the nonfinancial sector. • Commercial loans, defined as the sum of bills held, other promissory notes, commercial fixed loans, leasing finance, commercial overdrafts, commercial charge cards, and commercial other loans. • Housing loans, defined as the sum of secured owner-occupied housing, unsecured owner-occupied housing, and non-owner-occupied housing. • Personal loans, defined as the sum of personal fixed loans, personal overdrafts, personal charge cards, and other personal loans. The “bills held” item is included in the definition of both commercial and total lending. A client in need of finance may approach its bank with a request to draw a bill. Should the bank choose to hold the bill, it is reported as bills held in Form D. Alternatively, if the bank chooses to accept the bill but subsequently sells it to another bank, the bill is still reported as an asset (with an offsetting liability) but is not included in the definition of lending. As in the case of WBC, the treatment of bills as a lending item or a separate asset can affect growth rates. To illustrate, the chart presents growth in WBC lending, excluding bills held. • Capital base, defined as tier 1 capital plus tier 2 capital less goodwill and future income tax benefits. Tier 1 capital includes paid-up ordinary shares, nonrepayable share premium accounts, general reserves, retained earnings, noncumulative irredeemable preference shares, and minority interests in subsidiaries consistent with the foregoing components. Tier 2 capital is supplementary capital. It is classified as (1) upper tier 2 capital: general provisions for doubtful debts, asset revaluation reserves, cumulative irredeemable preference shares, mandatory convertible notes, and perpetual subordinated debt; and (2) lower tier 2 capital: term subordinated debt and limited life redeemable preference shares. • Total risk-weighted assets are a regulatory measure of assets. It is with respect to this measure that the minimum capital requirement is defined. The concept of risk-weighting assets is designed to provide capital concessions to those banks that hold relatively less risky assets. Growth in WBC Lending (Excluding Bills Held) (Percentage Change over Twelve Months) Per centage Change 30 20 Total Loans 10 Commercial Loans 0 –10 1991 1995 Source: Prudential Reports of Reserve Bank of Australia; Westpac Banking Corporation annual reports 30 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 Prior to September 1994, impaired assets were reported as “Nonperforming, Renegotiated, and Doubtful Items,” for which data began being collected in June 1990.2 The Impaired Assets Return is used to construct the following series: • Total impaired assets, defined as the sum of nonaccrual items, restructured items, other real estate–owned items, and other assets acquired through security enforcement. Nonaccrual items are those assets for which the bank does not expect further returns and hence cannot accrue income ahead of receipt. Restructured items are contracts that have been modified to provide concessions for the borrower. • Special provisions, defined as those provisions held against individually identified exposures if there is doubt surrounding collectibility. Special provisions can be held against both nonaccrual and restructured items. 1. Guidelines for the Capital Adequacy Return and Impaired Assets Return are detailed in Prudential Statements C1 and L1, respectively. 2. The Impaired Assets Return is based on RBA definitions as opposed to the Nonperforming, Renegotiated, and Doubtful Items Return, which is based primarily on banks’ subjective definitions. A P P E N D I X B Sample of Australian Banks Major Banks ending series for each major bank are constructed based on lending by the lead bank. When trading and savings banks operate as separate entities, their lending series are combined. Australia and New Zealand Banking Group Limited (ANZ) is combined with Australia and New Zealand Savings Bank between January 1990 and June 1992. Commonwealth Bank of Australia (CBA) is combined with Commonwealth Savings Bank between January 1990 and December 1992. National Australia Bank Limited (NAB) is combined with National Australia Savings Bank between January 1990 and September 1992. Westpac Banking Corporation (WBC) is combined with Westpac Savings Bank between January 1990 and September 1993. L Regional Banks Advance Bank Australia Limited (ADV) Banks of Melbourne Limited (BML) Bank of Queensland Limited (BQL) Bank of South Australia (BSA) BankWest Australia Limited (BWA) Challenge Bank Limited (CBL) Macquarie Bank Limited (MBL) Metway Bank Limited (MET) Primary Industry Bank of Australia Limited (PIB) State Bank of New South Wales (SBN) Trust Bank (TBT) Foreign Banks Bank of America Australia Limited (BAL) Barclays Bank Australia Limited (BBA) Banque Nationale de Paris (BNP) Bank of China (BOC) Bank of Singapore Australia Limited (BOS) Bank of Tokyo Australia Limited (BOT) Bankers Trust Australia Limited (BTA) Chase Manhattan Bank (CMB) Citibank Limited (CTI) Deutsche Bank Australia Limited (DBA) Hong Kong Bank of Australia Limited (HBA) IBJ Australia Bank Limited (IBJ) Lloyds Bank Limited (LBL) Mitsubishi Bank of Australia (MBA) NatWest Australia Bank Limited (NWA) Standard Chartered Bank Australia Limited (SCB) Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 31 A P P E N D I X C Breaks in Lending Series Portfolio Composition of Major Banks and Regional Banks hifts in portfolio composition shown in Chart 8 can be explained by the following breaks in lending series: S (1) In September 1990, the change in major banks’ portfolio composition reflected one bank’s reclassification of certain personal and government loans as housing and commercial loans. (2) In January 1991, major banks shifted out of commercial loans and into housing loans as a result of the acquisition of State Bank of Victoria by CBA. (3) In July 1992, the share of commercial loans in the regional bank portfolio dropped as a result of one bank’s reclassification of certain commercial loans as government loans. (4) In July 1994, the share of personal loans in the regional bank portfolio dropped as a result of one bank’s reclassification of certain personal loans as commercial loans. 32 Portfolio Composition of Major Banks Shifts in major bank portfolio composition, as presented in Chart 11, can be explained by the following breaks in lending series: (1) In November 1991, there was a shift in ANZ portfolio composition as the lending activity of subsidiaries National Mutual Royal Bank and National Mutual Royal Savings Bank was transferred to the lead bank’s balance sheet. (2) In January 1991, the share of commercial loans in the CBA portfolio dropped, with a corresponding increase in the share of housing loans. This change was due to the acquisition of State Bank of Victoria. (3) In August 1990, NAB reclassified certain housing loans as personal loans, generating the observed shift in portfolio shares. (4) In September 1990, WBC reclassified certain personal and government loans as housing and commercial loans. This reclassification resulted in an increase in the share of commercial loans in the WBC portfolio and a decrease in the share of personal loans. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 R E F E R E N C E S AUSTRALIA AND NEW ZEALAND BANKING GROUP. 1992. Annual Report. HANCOCK, DIANA, AND JAMES WILCOX. 1998. “The ‘Credit Crunch’ and the Availability of Credit to Small Business.” Journal of Banking and Finance 22 (August): 983–1014. BATTELLINO, RICK, AND NOLA MCMILLAN. 1989. “Changes in the Behaviour of Banks and Their Implications for Financial Aggregates.” Reserve Bank of Australia Research Discussion Paper 8904, July. HANCOCK, DIANA, ANDREW LAING, AND JAMES WILCOX. 1995. “Bank Capital Shocks: Dynamic Effects on Securities, Loans, and Capital.” Journal of Banking and Finance 19 (June): 661–77. BERGER, ALLEN N., RICHARD HERRING, AND GIORGIO SZEGO. 1995. “The Role of Capital in Financial Institutions.” Journal of Banking and Finance 19 (June): 393–430. BERGER, ALLEN N., ANIL KASHYAP, AND JOSEPH SCALISE. 1995. “The Transformation of the U.S. Banking Industry: What a Long, Strange Trip It’s Been.” Brookings Papers on Economic Activity, no. 2:55–218. BERGER, ALLEN N., AND GREGORY F. UDELL. 1994. “Did RiskBased Capital Allocate Bank Credit and Cause a ‘Credit Crunch’ in the United States?” Journal of Money, Credit, and Banking 26, no. 3 (part 2): 585–628. BERNANKE, BEN, AND CARA LOWN. 1991. “The Credit Crunch.” Brookings Papers on Economic Activity, no. 2:205–47. BLUNDELL-WIGNALL, ADRIAN, AND MARIANE GIZYCKI. 1992. “Credit Supply and Demand and the Australian Economy.” Reserve Bank of Australia Research Discussion Paper 9208, July. BOYD, JOHN H., PEDRO GOMIS, SUNGKYU KWAK, AND BRUCE D. SMITH. 2000. “A User’s Guide to Banking Crises.” Unpublished paper. COMMITTEE OF INQUIRY INTO THE AUSTRALIAN FINANCIAL SYSTEM. 1981. Australian Financial System: Final Report of the Committee of Inquiry. Canberra: AGPS. EDEY, MALCOLM, AND BRIAN GRAY. 1996. “The Evolving Structure of the Australian Financial System.” In The Future of the Financial System, edited by M. Edey. Sydney: Reserve Bank of Australia. FISHER, CHAY, AND CHRISTOPHER KENT. 1999. “Two Depressions, One Banking Collapse.” Reserve Bank of Australia Research Discussion Paper 1999-06, June. KAUFMAN, GEORGE. 1992. “Capital in Banking: Past, Present, and Future.” Journal of Financial Services Research 5 (April): 385–402. LOWE, PHILIP. 1995. “The Link between the Cash Rate and Market Interest Rates.” Reserve Bank of Australia Research Discussion Paper 9504, May. MACFARLANE, IAN. 1991. “The Lessons for Monetary Policy.” In The Deregulation of Financial Intermediaries, edited by I. Macfarlane. Sydney: Reserve Bank of Australia. ONGENA, STEVEN, DAVID C. SMITH, AND DAG MICHALSEN. 1999. “Distressed Relationships: Lessons from the Norwegian Banking Crisis (1988–91).” Unpublished paper. RESERVE BANK OF AUSTRALIA. 1992. “Bank Interest Rate Margins.” Reserve Bank of Australia Bulletin (May): 1–6. ———. 1997. Prudential Statements. COMMONWEALTH BANK OF AUSTRALIA. 1992. Annual Report. CONROY, F. 1997. “Managing Credit Risk—An Overview: Discussion.” In Credit Risk in Banking, edited by B. Gray and C. Cassidy. Sydney: Reserve Bank of Australia. KANG, JUN-KOO, AND RENÉ M. STULZ. 2000. “Do Banking Shocks Affect Borrowing Firm Performance? An Analysis of the Japanese Experience.” Journal of Business 73 (January): 1–24. STEVENS, GLENN. 1997. “Some Observations on Low Inflation and Household Finances.” Reserve Bank of Australia Bulletin (October): 38–47. THOMPSON, GRAEME. 1991. “Prudential Lessons.” In The Deregulation of Financial Intermediaries, edited by I. Macfarlane. Sydney: Reserve Bank of Australia. ULLMER, MICHAEL. 1997. “Managing Credit Risk—An Overview.” In Credit Risk in Banking, edited by B. Gray and C. Cassidy. Sydney: Reserve Bank of Australia. WESTPAC BANKING CORPORATION. 1991. Annual Report. GRENVILLE, STEPHEN. 1991. “The Evolution of Financial Deregulation.” In The Deregulation of Financial Intermediaries, edited by I. Macfarlane. Sydney: Reserve Bank of Australia. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 33 The Potential for Portfolio Diversification in Financial Services A L A N K . R E I C H E R T L A R R Y D . W A L L A N D Reichert is a professor of finance at Cleveland State University, Cleveland, Ohio. Wall is a research officer in the financial section of the Atlanta Fed’s research department. The authors thank Ray DeGennaro, Jerry Hanweck, Frank King, Steve Smith, and Larry White for helpful comments. T HE RECENT PASSAGE OF THE GRAMM-LEACH-BLILEY ACT REWRITES THE RULES RELATING TO THE AFFILIATION OF BANK AND NONBANK FINANCIAL SERVICES PROVIDERS. PREVIOUSLY, RULES HAD SOUGHT TO PREVENT OR AT LEAST RESTRICT COMMERCIAL BANKS’ ABILITY TO AFFILIATE WITH INVESTMENT BANKS AND INSURANCE COMPANIES. less effective over time as financial firms used advances in information processing and financial technology to avoid the rules by offering products that were functionally equivalent to those that they could not legally provide.1 The effectiveness of the rules was further diminished as sympathetic regulators reinterpreted prior law to allow their regulatees to enter other parts of the financial services industry. Nevertheless, the old rules imposed costly restrictions on a bank’s ability to provide investment banking and insurance activities. Moreover, they often had the effect of preventing investment banking and insurance firms from owning a commercial bank. The Gramm-Leach-Bliley Act removes most of the impediments to the affiliation of commercial banks with investment banks and insurance companies. Although the Gramm-Leach-Bliley Act largely eliminates the barriers to affiliation within the financial services industry, it does not necessarily follow that financial supermarkets will come to dominate the financial services industry, as some had predicted. THESE RULES HAD BECOME Any financial conglomerates that emerge must be at least as profitable as firms that focus on specific market segments, and the conglomerates must earn higher risk-adjusted rates of return if they are to dominate. Berger (forthcoming) surveys a large number of existing studies of cost and profit efficiency in the financial services industry to assess the prospects for such financial conglomerates. While pointing out that the available evidence is incomplete in a number of important ways, he concludes that the largest possible gains appear to exist from the greater risk-diversification potential of conglomerates. Conglomerates may exploit the gains from diversification by increasing the proportion of their portfolio invested in higher-return assets—for example, by holding proportionately more loans and fewer securities—without increasing the riskiness of the total portfolio. Thus, assessing the potential of diversification to reduce risk is a potentially important issue in understanding the effect of the GrammLeach-Bliley Act on financial conglomeration.2 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 35 The potential for diversification has been considered in a number of studies. Wall, Reichert, and Mohanty (1993) survey prior studies in this area. They also provide an analysis of the strengths and weaknesses of alternative methodologies, which is adapted here in the box on page 8. In addition, they present new evidence using Internal Revenue Service (IRS) data on industry-level return on assets. They find an emerging consensus that pairwise combinations of individual firms provided insignificant gains from diversification at best and an increase in risk at worse. However, following Litan (1985) and using IRS data over the 1974–89 period, they find gains from forming portfolios of bank and nonbank activities when using IRS rather than indiAlthough the Gramm-Leachvidual bank data. Certain portfolios of Bliley Act largely eliminates bank and nonbank the barriers to affiliation activities showed within the financial services substantially higher returns for lower industry, it does not necesrisk. However, Wall, sarily follow that financial Reichert, and Mohanty supermarkets will come to (1993) also find that their results were dominate the financial time-dependent. For services industry. example, they find that life insurance was the dominant part of the lowest-risk portfolios in the 1974–80 subperiod whereas it was a relatively small part of those portfolios in the 1981–89 subperiod. This article updates the contribution of Wall, Reichert, and Mohanty (1993) in several ways. It begins with a brief review of the literature since 1993. Next, it summarizes the legal changes resulting from the recent passage of the Gramm-LeachBliley Act. The article then extends the earlier empirical analysis, first by updating the analysis of the potential gains from diversification using IRS data on return on assets through 1997 and then by extending the diversification analysis to consider profitability as measured by return on equity using IRS data from 1990 to 1997. The article concludes with some speculation about the potential impact of recent technological changes on the benefits of forming financial conglomerates. Literature Review everal recent studies have addressed the issue of the benefits and risk associated with bank holding company diversification. Whalen (1999a) examines the overseas insurance activities of S 36 U.S. bank holding companies for the period from 1987 to 1997. Since U.S. banks are required to report separately on foreign bank activities, he is able to use the return on assets (ROA) associated with these foreign activities as his measure of performance. Whalen finds that mean returns in insurance activities exceeded the returns to banking as well as the returns on other nonbanking activities by a significant margin. On a stand-alone basis, insurance activities appear to be somewhat riskier than other nonbanking services, but when combined with traditional banking activities the combination can noticeably improve a bank’s risk/return opportunities. Whalen concludes that insurance activities that constitute less than 10 percent of a diversified financial firm’s total assets should not present a major problem for regulators. In a similar study (1999b), Whalen examines the foreign securities activities of U.S. banks. Using industry-level data he finds that the average security returns are similar to the returns of traditional banking activities, while measures of risk are somewhat higher. According to firm-level data, average security returns exceeded banking returns by a substantial margin, while security-related risk was higher. However, Whalen concludes that, taking into account the low correlation of foreign investment bank returns with those of their U.S. bank affiliates, overseas securities activities have reduced risk for U.S. banks. Kwan (1998) studies the relative risk and return associated with both securities underwriting and trading by comparing the performance of bank holding company securities affiliates called Section 20 subsidiaries with their commercial bank affiliates. Kwan finds that those securities subsidiaries tend to be riskier but not always more profitable. On the other hand, the low correlation of securities and bank returns provides some degree of diversification benefits. Kwan concludes that underwriting activities generate likely diversification benefits for both security trading and traditional commercial banking activities. Two papers expand the focus to examine combinations of banks with financial firms that are currently unaffiliated with banks. Laderman (2000) generates synthetic banks to simulate the entry of both large banks and all banks into eight other types of firms. She finds that all of the activities she examined would reduce bank risk but that in some cases the benefits would occur only at trivial levels of investment in nonbank activities. The largest weights were on firms involved in various aspects of insurance and securities. Allen and Jagtiani (1999) also generate synthetic banks to simulate the impact of both insurance and securities activities. They find that these nonbank activities reduce the Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 firm’s total risk but serve to increase systematic market risk. The authors conclude that the benefits of diversification alone are not sufficient to support the expansion of bank power in the securities and insurance areas. The Gramm-Leach-Bliley Act he changes in the Gramm-Leach-Bliley Act may be characterized as facilitating the continued evolution of the financial system rather than permitting revolutionary change. While the barriers separating commercial banking from investment banking and insurance may have at times seemed almost impenetrable, changes in both the markets and the regulators’ interpretations of existing law had allowed considerable competition across different subsectors of the financial services industry over the last several decades. Pre-Gramm-Leach-Bliley Act. The GlassSteagall Act adopted in 1933 prohibited commercial banks from being affiliated with firms “engaged principally” in investment banking activities such as the issuing, floatation, underwriting, public sale, or distribution of securities.3 At the same time, the act did not completely prohibit commercial banks from conducting all types of investment banking services. For example, banks could underwrite municipal general obligation bonds and offer certain investment services through their trust departments. Banking organizations could also offer a variety of investment banking products in foreign markets, allowing U.S. banking organizations to develop some experience in investment banking.4 Although commercial banks obtained explicit authorization to provide a limited range of securities services, the Glass-Steagall Act effectively kept them out of the mainstream of domestic investment banking. Investment banks had somewhat more success in offering close substitutes for traditional T banking products. The rapid growth of the commercial paper market has reduced large firms’ use of short-term commercial loans from banks. Investment banks also used money market mutual funds to provide a substitute for demand deposits at commercial banks. Banks gradually expanded their investment banking activities as the regulatory agencies reinterpreted ambiguous parts of the statutory law.5 Probably the most significant reinterpretation relates to the Glass-Steagall Act’s prohibition on commercial banks’ affiliation with firms that “principally engaged” in bank-ineligible activities. Bank lawyers argued that this provision allowed banks to be affiliated with a securities firm as long as the securities firm Although commercial was not “principally engaged” in ineligible banks obtained explicit activities. 6 In 1987 authorization to provide a the Federal Reserve limited range of securities Board agreed with this interpretation. 7 services, the GlassAt first, the securities Steagall Act effectively subsidiaries approved kept them out of the by the Federal Reserve (Section 20 submainstream of domestic sidiaries) were subinvestment banking. ject to strict limits on what they could underwrite and the extent to which they could engage in bank-ineligible activities. These activities were also subject to “firewalls” that limited the potential for the securities affiliate to put the commercial bank at risk but may have limited possible synergies between the two types of affiliates. The various restrictions were relaxed in subsequent years as commercial banks and their regulators gained experience with Section 1. For example, investment banks offered money market mutual funds as a substitute for bank deposits, and banks offered syndicated loans as a substitute for underwriting bonds. 2. Whether the potential for diversification gains is in fact realized will depend in part on how the combined firms are managed after the takeover. Hypothetical combinations, such as those formed in this article, are inadequate for addressing the question of how the new activities will be managed. The issue of how diversification works in practice is addressed by Wall (1987), Kwan (1998), and Whalen (1999a, 1999b). 3. See Section 9.02 of Fein (1998) for a discussion of the Glass-Steagall Act. 4. See Section 12.01 of Fein (1998) for a discussion of U.S. banking organizations’ ability to engage in securities activities outside the United States. 5. See Chapter 9 of Fein (1998) for a discussion of banks’ authority to engage in securities activities in the United States. See also Section 1.04 for a review of the key administrative rulings authorizing banks to engage in additional securities activities, and Section 1.05 for the relevant court cases. 6. A securities activity was considered “eligible” if the banks could legally engage in the activity without being restricted by the Glass-Steagall Act. 7. See the Federal Reserve Board’s decision on Citicorp/J.P. Morgan & Co., Inc./Bankers Trust New York Corporation in the Federal Reserve Bulletin (1987, 473ff). Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 37 20 subsidiaries.8 The relaxation of the restrictions, combined with changes in investment banks’ mix of business, allowed the Federal Reserve to approve the acquisition of Citicorp by Travelers Group, a financial firm with a large investment banking operation.9 Bank progress in entering the insurance field has been slower. The laws governing banking organizations restricted their ability to provide insurance services. However, regulatory reinterpretations of the statutes by the Office of the Comptroller of the Currency (OCC) in the years leading up to the Gramm-Leach-Bliley Act were opening the door to bank sale of insurance products. A 1916 federal law permits national banks to act as an insurance agent in any community of fewer than 5,000 people. The OCC interThe Gramm-Leach-Bliley preted this stipulation as requiring the actual Act ends the need for insurance activities to commercial banks to find be located in commuloopholes to enter investnities with a population no greater than ment banking and insur5,000 residents but ance and for investment permitting marketing banks and insurance firms of the insurance nationally to commuto find loopholes to enter nities of all sizes.10 commercial banking. Federal bank regulators also used their authority to control bank organizations’ entry into other types of financial services. The Federal Reserve limited the activities permissible to bank holding companies under authority provided by the Bank Holding Company Act of 1956 and its amendments. All three federal bank regulators—the Federal Reserve, OCC, and the Federal Deposit Insurance Corporation—placed limits on the activities of subsidiaries of their banks. In the years prior to the Gramm-Leach-Bliley Act, the OCC had undertaken an extensive review of its rules and had determined that subsidiaries of national banks could provide a wide variety of financial services. This interpretation was incorporated into the agency’s Part 5 rules.11 Although national banks might have provided a wide range of heretofore impermissible activities via subsidiaries given sufficient time, relatively few applications for new activities were approved under Part 5 prior to the Gramm-Leach-Bliley Act. Post-Gramm-Leach-Bliley Act. The GrammLeach-Bliley Act ends the need for commercial banks to find loopholes to enter investment banking and insurance and for investment banks and insurance firms to find loopholes to enter commercial banking. Amending the Bank Holding Company 38 (BHC) Act, the Gramm-Leach-Bliley Act allows commercial banks to be affiliated with a wide range of financial services. As a part of its effort to reduce the barriers between financial services companies, the act repeals the Glass-Steagall prohibitions on the affiliation of commercial and investment banking. At the same time, however, the act added a new barrier separating banking and commerce. The Gramm-Leach-Bliley Act created a new type of bank holding company, called the financial holding company, under which nonbank activities are organized as subsidiaries of the holding company. If a bank holding company elects to become a financial holding company, all of its subsidiary banks and thrifts must meet several criteria: they must be well capitalized and well managed and must have at least a satisfactory Community Reinvestment Act rating.12 Financial holding companies are authorized to engage in a variety of activities, including (1) lending, trust, and other banking activities, (2) insurance activities, and (3) securities underwriting and dealing. The Federal Reserve, with the approval of the Secretary of the Treasury, may also expand this list of activities to include other “financial” or “incidental” activities. The criteria for approving additional activities include clear consistency with the purposes of the Gramm-Leach-Bliley Act; actual or likely changes in the financial services marketplace, including innovation in financial and information technology; and assessment of an activity as “necessary or appropriate” to enable financial holding companies to compete or to use technology effectively in providing financial services. While the Gramm-Leach-Bliley Act seeks to maintain the separation of banking and commerce, the law recognizes that investment banks, merchant banks, and insurance companies may acquire controlling interests in companies in the ordinary course of business. If nonbank providers were prohibited from owning merchant banks, their cost of owning a bank would significantly increase. Thus, the act permits financial holding companies to own a controlling interest in any company. However, to limit the mixing of banking and commerce, a financial holding company must acquire the interest in the ordinary course of business and the financial holding company must act as a passive investor.13 The Gramm-Leach-Bliley Act also allows national banks to have a subsidiary that engages in any activity authorized directly for the bank or any financial activity except insurance underwriting, insurance investments, real estate investment or development, and merchant banking.14 The Secretary of the Treasury may expand the list of permitted activities subject to approval by the Federal Reserve Board. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 The act limits the total assets of all financial subsidiaries of a bank to 45 percent of the bank’s assets or $50 billion, whichever is smaller. The Gramm-Leach-Bliley Act closes a loophole in a previous law that allowed holding companies that owned no bank charters and only one thrift charter to be affiliated with commercial activities.15 The act grandfathers any holding company that owned a thrift charter on May 4, 1999, with the restriction that the owner of the thrift charter may not transfer the charter to another corporation. Empirical Analysis he Gramm-Leach-Bliley Act has resolved the public policy problem of what restrictions should be placed on banking organizations’ ability to provide nonbank services. However, financial holding companies must still evaluate which services to provide, an evaluation that may depend in part on the diversification gains. Further, supervisors will need to evaluate the safety and soundness implications of the new combinations arising under the financial holding company umbrella. While several studies provide recent evidence using firm-level data, heretofore none have reexamined industry-level data from the most comprehensive publicly available source, IRS corporate income tax returns. An update of Wall, Reichert, and Mohanty’s (1993) analysis of IRS data may be useful given the many differences between data from the 1990s and the 1970s and 1980s, including (1) a different macroeconomic environment with significantly lower inflation and somewhat higher growth in the T 1990s, (2) continuing innovation in the financial sector, most notably a continuation of disintermediation, and (3) changes in financial technology and in regulatory limits, allowing financial services firms to provide a wider range of services.16 Data. The data are obtained from the IRS Corporate Income Tax Returns for the major sector “Finance, Insurance, and Real Estate” (Major Group 60) for the years 1974–97. This is the same source of the earlier data Wall, Reichert, and Mohanty used (1993). This article follows Wall, Reichert, and Mohanty in focusing on using ROA as the primary measure of profitability. To facilitate comparison, the research reproduces their results for the 1974–80 and 1981–89 periods and provides new results for the 1990–97 period. The ROA analysis is supplemented with an analysis of return on equity (ROE) for the 1990–97 period. In some years, the corporate returns publication contains two related columns, “Net Income” and “Deficit.” Net income refers to total net income for corporations reporting positive earnings for the year. Deficit refers to total losses incurred by corporations reporting losses for the year. In this study, aggregate industry profits for the year were calculated by subtracting the deficit figure from the net income figure to provide a comprehensive picture of industry performance for that year. In some other years a net profit for the total industry is provided by the IRS. A breakdown of the industry categories used by the IRS is provided in the appendix. Descriptive Statistics. Table 1 presents the results of the ROA analysis by major sector. The first 8. See Section 9.05 of Fein (1998) for a review of the original firewalls and the subsequent modifications. 9. Travelers Group applied to the Board of Governors of the Federal Reserve System to acquire Citicorp and, thereby, become a bank holding company (which would be renamed Citigroup). The Travelers Group owned Salomon, Smith Barney Inc., one of the largest securities firms in the United States. The Board approved the application on September 23, 1998 (Federal Reserve Bulletin 84 [1988], 985-1016). In approving the Travelers application, the Board found that Salomon, Smith Barney’s bank-ineligible investment banking activities fell within the 25 percent of revenue test imposed at that time on Section 20 subsidiaries (page 1006 of the Bulletin) and, thus, that the investment banking activities of Travelers were not an obstacle to the transaction. (However, the Board did rely on certain commitments of the new Citigroup to impose certain limits on its securities and investment banking activities.) 10. For an informative discussion of the legislative issues leading up to the Gramm-Leach-Bliley Act, see the interview with Representative James Leach in the March 2000 issue of The Region (Vol. 14, No. 1) published by the Federal Reserve Bank of Minneapolis. 11. See the testimony of Eugene A. Ludwig, Comptroller of the Currency, before the Subcommittee on Finance and Hazardous Materials of the Committee on Commerce of the U.S. House of Representatives, July 17, 1997. 12. Section 103 of the Gramm-Leach-Bliley Act lays out the requirements for a bank holding company to become a financial holding company and the range of permissible activities for a financial holding company. 13. Section 103 of the Gramm-Leach-Bliley Act. 14. Section 121 of the Gramm-Leach-Bliley Act establishes the limits on national bank provision of financial services through affiliates. 15. Section 401 of the Gramm-Leach-Bliley Act imposes restrictions on holding companies that own only one thrift charter. 16. Disintermediation is the replacement of financing via loans funded by intermediaries with loans funded by financial markets. Examples of disintermediation include the replacement of commercial loans by banks with commercial paper sold to financial markets and the replacement of mortgage loans held by thrifts with mortgage-backed securities held by a variety of investors. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 39 column indicates the seven broad industry categories used by the IRS to classify the finance sector. The second column indicates, for each industry, the three data periods used in the analysis: (1) the prederegulation period, 1974–80; the initial deregulation period, 1981–89; and the most recent period, 1990–97, which has witnessed continued deregulation and rapid industry consolidation. Congress passed the Depository Institution Deregulation and Monetary Control Act (DIDMCA) in March 1980. DIDMCA is widely regarded as the major piece of legislation that formally ushered in the era of financial deregulation. DIDMCA was shortly followed by the Garn-St. Germain Act of 1982, which accelerated the deregulation process. The third and fourth columns give the mean industry ROA for each period and the associated industry rank (with larger ROAs ranked higher). The next column indicates the degree of earnings volatility associated with each industry, as measured by the coefficient of variation of ROA over the given time period for industries with positive mean returns. The coefficient of variation of ROA is the ratio of the standard deviation of returns over the period to the mean return over the period.17 In the next column, volatility is ranked from low to high (thus, industries with low volatility are ranked high). The remaining columns indicate the simple correlation between the ROA of bank holding companies and each industry group. The value of the correlations is ranked from highly positive to highly negative. Thus, industries with high negative correlations are ranked high while industries with high positive correlations are ranked low. ROA is a broad measure of the efficiency with which resources are employed within an industry and is calculated by dividing annual net income by total year-end assets.18 As indicated in Table 1, while the banking sector ranked only fifth out of seven in terms of average ROA, it experienced a threefold increase in average earnings during the 1990–97 period compared with the two earlier periods. The only other sector to experience such a dramatic improvement was other credit agencies. While banking historically has recorded a relatively low degree of earnings volatility, the sector was ranked number one as reflected by the lowest coefficient of variation during the decade of the 1990s. From a simple pairwise correlations perspective, the holding and other investment companies sector had the second-lowest TA B L E 1 Industry ROA Volatility and Correlation Analysis, 1974–97 Coefficient of Variation of ROA ROA Industry a b Period Mean Value Rank Correlation with Bank Holding Companies Value Rank Value Rank Bankinga 1974–80 1981–89 1990–97 .29 .24 .84 7 5 5 25.2 65.3 39.6 2 3 1 .87 .96 .99 6 7 7 Credit Agencies (other than banks) 1974–80 1981–89 1990–97 .35 –.40 1.68 6 7 3 89.2 163.7 172.1 7 6 6 .27 .68 .32 2 6 2 Insurance (broad category ) 1974–80 1981–89 1990–97 1.48 .61 1.10 4 3 4 36.6 78.6 46.9 3 5 4 .85 .66 .88 5 5 6 Insurance Agents and Brokers 1974–80 1981–89 1990–97 7.15 3.74 6.03 1 2 1 21.8 54.5 41.3 1 2 2 .76 .63 .77 4 4 4 Real Estate (broad category ) 1974–80 1981–89 1990–97 1.61 .20 –.09 3 6 7 63.3 237.5 NMb 6 6 .94 –.43 .46 7 3 3 Security, Commodity Brokers and Services 1974–80 1981–89 1990–97 1.29 .59 .83 5 4 5 52.1 69.3 47.8 5 4 5 –.23 –.70 .82 1 2 5 Holding and Other Investment Companies 1974–80 1981–89 1990–97 4.66 6.40 2.99 2 1 2 37.6 19.2 45.8 4 1 3 .60 –.90 .34 3 1 1 Includes mutual savings banks, bank holding companies, and other (or independent) commercial banks. Not meaningful. Source: Wall, Reichert, and Mohanty (1993) and authors’ calculation using data from IRS corporate income tax returns 40 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 TA B L E 2 Banking Industry ROA Volatility and Correlation Analysis, 1990–97 Coefficient of Variation of ROA ROA Industry a Mean Value Bank Holding Companies Mutual Savings Banks Independent Banks (not mutual savings banks or bank holding companies) Savings and Loans Personal Credit Institutions Business Credit Institutions Other Credit Agencies Life Insurance Mutual Insurance Other Insurance Insurance Agents Real Estate Operators and Lessors of Buildings Lessors of Mining and Oil Properties Lessors of Railroad Properties Condominium Management and Coops Subdividers and Developers Other Real Estate Security Brokers, Dealers Commodity Brokers, Dealers Regulated Investment Companies Real Estate Investment Trusts Small Business Investment Trusts Holding and Other Investment Companies a b Correlation with Bank Holding Companies Rank Value Rank Value Rank 0.85 0.96 13 11 42.6 48.0 2 7 1.00 0.09 23 4 0.59 0.18 0.99 0.60 0.82 0.93 1.13 1.44 6.03 17 18 10 16 14 12 9 7 2 91.0 156.2 59.1 76.1 43.7 48.4 87.4 45.9 41.3 17 18 12 14 4 8 15 5 1 0.05 –0.24 0.61 0.66 0.54 0.91 0.40 0.80 0.77 3 2 14 16 12 21 9 20 19 –0.05 8.48 4.55 20 1 3 3 10 0.17 0.64 0.52 5 15 11 –1.16 –0.68 0.11 0.70 3.01 3.20 2.48 –0.41 23 22 19 15 5 4 6 21 NM NM 1585.8 49.4 55.3 46.7 62.4 NM –0.27 0.30 0.52 0.76 0.95 0.25 0.70 0.29 1 8 10 18 22 6 17 7 1.35 8 89.9 0.61 13 NMb 43.4 52.8 19 9 11 6 13 16 Detailed statistics for each nonbank industry group are available in Wall, Reichert, and Mohanty (1993). Not meaningful. Source: Authors’ calculation using data from IRS corporate income tax returns degree of positive correlation and at the same time was ranked high in terms of average ROA and low earnings volatility during the 1990–97 period. Table 2 presents similar information for the 1990–97 period for twenty-three industry sub-categories. The most profitable sectors proved to be lessors of mining and oil properties, followed by insurance agencies and lessors of railroad properties. With an average ROA of 0.85, bank holding companies ranked 13. The three least-profitable sectors were condominium and coop management, real estate subdividers and developers, and small business investment trusts. In terms of earnings volatility, insurance agencies ranked the lowest, followed closely by bank holding companies and lessors of mining and oil properties. At the other extreme, the three most volatile sectors proved to be “other” real estate, small business investment trusts, and operators and lessors of buildings. Based on pairwise correlations with bank holding company earnings, the three sectors with the smallest correlations are condominium and coop management (–.27), savings and loans (–.24), and independent banks (.05). However, these three sectors ranked in the bottom third in terms of both earnings and volatility. While ROA is useful for some purposes, ROE provides information that may be of special interest to investors. For example, banking is a low-margin, high-leverage industry. Thus, a bank with an ROA of 17. The coefficient of variation of ROA and of ROE may be thought of as a sort of inverse Sharpe ratio. The Sharpe ratio is a measure of the return to bearing risk and is defined as the ratio of excess returns (mean returns less the risk-free rate) to the standard deviation of returns. The primary difference between the two ratios is that the Sharpe ratio incorporates the return to a riskless asset whereas the coefficient of variation does not. 18. Albeit ROA incorporates only on-balance sheet assets; off-balance-sheet exposures are not incorporated in the ratio. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 41 B O X Evaluation of Prior Studies he studies of portfolio diversification discussed in Wall, Reichert, and Mohanty (1993) and updated here have approached analyzing the effects on portfolios of diversification into nonbank activities in a variety of ways.1 The methodologies used each have significant strengths and weaknesses that it is important to understand in evaluating the current state of the literature. T Measurement of Portfolio Performance Most studies of the effect of portfolio diversification focus on one or two risk measures: the coefficient of variation of some return measure or the risk of failure calculated using accounting or market data. The coefficient of variation is, as noted in the text, merely the variability of returns (standard deviation of returns) divided by the expected return. The risk of failure incorporates a firm’s equity capital, its expected returns, and standard deviation of returns to provide a measure of the likelihood that a firm will experience losses that exceed its capital. Risk of failure is a more direct measure of the primary regulatory concern: Would increased participation in nonbank activities make banks more or less likely to fail? However, implicit in the risk-of-failure measure is the assumption that the combined organization’s capital structure will be the sum of its individual premerger capital structures, an assumption that may not be appropriate if regulators require higher postmerger capital levels. Further, calculation of the risk of failure requires data on premerger capital structures that may not be available from some data sources. A further consideration in evaluating portfolio performance is the perspective of bank owners and managers. Most studies focus on risk issues because that is the regulator’s concern. Bank owners and managers, however, actually undertake mergers on the basis of the effect of diversification on both the return and risk of the combined organization. Banks may engage in a risk-reducing merger if the reduction in their expected return is not too large, but they may also be willing to undertake higher risk if the increase in expected return is sufficiently large.2 Thus, a full analysis of the effect of diversification on returns must consider both the banks’ and the regulators’ perspectives.3 Formation of Portfolios The various studies of bank mergers take three different approaches to forming the portfolios for analysis. Some studies limit their analysis to three or four statistics: industry-average mean returns, industry-average standard deviations of returns, industry-average coefficients of variation of returns on assets, and the correlation of industry returns with banking returns. Looking at industry statistics alone does not allow an 42 easy determination of the change in risk that results from combining different industries into a single firm. For example, an industry might have a higher standard deviation of returns than banking, but the returns may be negatively correlated with banks’ returns. Thus, it is not always clear whether the higher standard deviation of a particular firm from this industry combined with a banking firm will increase the risk to the postmerger organization or whether its negative correlation with banking will generate less risk. An alternative to using overall industry statistics is to combine industries in pairs—banks and one nonbank industry at a time. This approach provides for simultaneously considering the effects of expected return, the standard deviation of returns, and the correlation between returns (as well as the capital positions of the two firms, when appropriate). Perhaps most importantly, this approach has the advantage of corresponding with actual bank behavior. Because firms typically engage in one merger at a time, the concern to banks and their regulators at any given point in time is the desirability of a particular pairwise combination. The third alternative in examining portfolios is to analyze efficient portfolios of banks and several nonbank industries. As discussed in the article, the term “efficient portfolio” refers to one whose combinations produce the most return for any given level of return variability (or, equivalently, the least return variability for any given return). These portfolios may contain firms operating in only two industries (or in some cases a single industry). However, as Litan (1985) found, some efficient portfolios are likely to contain multiple industries. Examining portfolios of unique service products is advantageous because it is the approach that banks should take from a portfolio risk and return perspective. Thus, basing public policy solely on the risk effects of pairwise mergers may impose significant social costs if it results in policies that prevent the formation of efficient portfolios of bank and nonbank firms. Timing of Aggregation to Industry Level The various studies take two approaches to the aggregation of firm data into industry statistics. Some studies combine individual firms into a single industry before conducting any analysis, and others calculate the mean and variability of returns for individual firms (and across pairs of firms) and then aggregate the figures across all firms in the industry (or in the pair of industries). The major disadvantage of the first approach is that individual firms enter into mergers with specific firms, not with broad industries. On the Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 other hand, industry aggregate figures may be a better proxy for the expected future distribution of returns to the extent that two conditions hold—that is, if withinindustry differences primarily arise from regional economic conditions and if firms within industries are combining across regions. Another advantage of using industry aggregates is that spurious results in the formation of portfolios may be less likely. An efficient portfolio is formed by looking at an individual entity’s “assets” to determine the combinations that produce an efficient set of portfolios. These assets may be defined as entire industries or as individual firms within industries. Obviously, the number of separate assets for inclusion in an efficient portfolio will increase dramatically if individual firm returns are used rather than industry returns. In general, an increase in the number of assets is likely to increase the chances of identifying lower-risk portfolios. Thus, Boyd, Graham, and Hewitt (1993) argue that random chance favors the possibility that a risk-reducing portfolio will be found using individual firm data even if there is not a real opportunity for diversification to reduce risk. Use of Market or Accounting Data Banks’ and nonbank firms’ returns may be measured using accounting or financial market data. The drawback to using accounting data is that they are not perfectly correlated with economic returns. Firms often try to smooth accounting data through time, producing reported returns that are deliberately low in the good years and high in the bad years. If firms across different industries have unequal ability to smooth their accounting earnings, then accounting-based risk measures may not provide accurate interindustry comparisons of risk. Using accounting data has some appeal, however. First, market data is typically available only for the largest firms in an industry, so it clearly is more limited than accounting data. In addition, regulators rely heavily on accounting figures in their evaluation of a bank’s financial condition. Studies that rely on accounting data use two sources of information: accounting data from the individual firm’s public financial statements prepared according to generally accepted accounting principles (GAAP) and accounting data published by the IRS for all firms in an industry prepared according to IRS accounting rules. Each data set has its advantages. GAAP rules are intended to fairly present a firm’s performance over time whereas IRS rules also reflect a number of public policy decisions. For example, to encourage banks to hold state and local government obligations, IRS rules allowed banks to understate their income by excluding the interest from holding these obligations. Another advantage of using GAAP data is that they are available at the individual firm level, and IRS data are available only for an entire industry. On the other hand, IRS data reflect a broad cross-section of firms in an industry while public financial statements are only available for the largest firms. Overall Evaluation of Prior Studies As the above discussion suggests, there appears to be no single “correct” methodology. Each has advantages and disadvantages. Ideally, the different approaches would produce consistent results confirming that individual findings were not the result of a unique methodology. 1. This box is adapted from Wall, Reichart, and Mohanty (1993). 2. See Boyd and Graham (1986) for a discussion of the issue of managerial incentives. 3. Boyd, Graham, and Hewitt (1993) point out that an important issue in evaluating the effect of a merger is the purchase price paid for the target by the acquiring organization. Virtually all studies of historical data implicitly assume that no premium will be paid to the target. Researchers use this assumption not because it is realistic but because they have no good basis for determining the likely magnitude of the takeover premium. only 1.25 percent and a capital-to-asset ratio of 8 percent, would earn a 15.6 percent ROE. On the other hand, a brokerage firm with an ROA of 5 percent and a capital-asset ratio of 50 percent, would earn only a 10 percent ROE. Table 3 replicates Table 2 using ROE over the 1990–97 period. The use of ROE rather than ROA improved the profitability rankings of depositories (bank holding companies, mutual savings banks, independent banks, and savings and loans), as would be expected from their generally low capital-to-asset ratios. The lower capital ratios also boosted the standard deviation of returns. The net result of the increase in the profitability ratio and the standard deviation for bank holding companies is small; they had the second lowest coefficient of variation using ROA and the lowest coefficient of variation using ROE. The use of ROE rather than ROA caused some changes in the rankings by correlation with bank holding companies, but the differences are generally small. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 43 Efficient Portfolios Using Return on Assets. Table 4 forms efficient portfolios of selected sectors that provide both traditional banking-related services, such as mutual savings banks, savings and loans, and personal and business credit companies and those that provide nontraditional financial services, such as security and commodity brokers and dealers, life insurance underwriters, and security and insurance agents and brokers.19 The portfolios are formed subject to the assumption that investment in any given industry cannot be less than zero and that there is no risk-free asset. Efficient portfolio combinations of activities that have traditionally been closely related to banking are presented in the top part of Table 4. The portfolio calculation program calculates the efficient frontier and provides detailed information on portfolios at points selected by the user. 20 In order to provide an overview of the results, portfolios are obtained at various levels of return. Given that the program rounds returns and variances, the point selected to obtain a portfolio for each level of return is the one with the lowest standard deviation of return. Thus, portfolio B for the 1990–97 period was selected to have an ROA of 0.7 percent. The lowest standard deviation of return for a portfolio with an ROA of 0.7 percent is 0.2 percent. The resulting coefficient of variation is 28.6 percent. A portfolio that yields these returns would invest the following percentages in each industry: 31.0 percent in mutual savings banks, 28.1 percent in bank holding companies, 33.9 percent in savings and loan associations, 4.3 percent in personal credit companies, and 2.7 percent in business credit companies. Evaluating the results requires a rough standard for judging the extent to which bank holding companies should be allowed to diversify. One reasonable standard suggests that bank holding companies should be allowed to diversify as long as the nonbank activity does not increase the coefficient of variation above that of bank holding companies by themselves. For 1974–80, the mean ROA is 0.27 and the coefficient of variation is 30.6; for 1981–89 the mean ROA is 0.26 and the coefficient of variation is 34.4. TA B L E 3 Banking Industry ROE Volatility and Correlation Analysis, 1990–97 Coefficient of Variation of ROE ROE Industry Mean Value Bank Holding Companies Mutual Savings Banks Independent Banks (not mutual savings banks or bank holding companies) Savings and Loans Personal Credit Institutions Business Credit Institutions Other Credit Agencies Life Insurance Mutual Insurance Other Insurance Insurance Agents Real Estate Operators and Lessors of Buildings Lessors of Mining and Oil Properties Lessors of Railroad Properties Condominium Management and Coops Subdividers and Developers Other Real Estate Security Brokers, Dealers Commodity Brokers, Dealers Regulated Investment Companies Real Estate Investment Trusts Small Business Investment Trusts Holding and Other Investment Companies a Rank Value Rank Value Rank 8.2 13.4 8 5 38.4 44.3 1 4 1.00 0.06 23 4 8.1 7.5 5.6 8.1 10.5 7.3 4.2 4.5 16.4 9 11 13 10 7 12 16 14 3 99.7 184.5 47.1 52.0 42.5 47.8 85.2 43.6 61.8 17 18 6 10 2 7 16 3 13 0.06 –0.39 0.85 0.64 0.63 0.83 0.40 0.65 0.48 5 1 21 17 16 20 11 18 14 –0.4 19.1 10.6 19 1 6 NMa 62.8 53.7 NM 14 12 0.06 0.25 0.46 3 7 13 –2.3 –4.5 –0.9 15.9 18.1 3.3 4.4 –1.2 22 23 20 4 2 17 15 21 NM NM NM 51.0 52.6 46.7 47.8 NM NM NM NM 9 11 5 8 NM –0.31 0.30 0.34 0.79 0.91 0.25 0.52 0.16 2 9 10 19 22 8 15 6 3.3 18 81.0 15 0.41 12 Not meaningful. Source: Authors’ calculation using data from IRS corporate income tax returns 44 Correlation with Bank Holding Companies Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 Traditional Banking Activities. Looking at traditional banking activities in Table 4 and applying the above standard to the 1974–80 period, portfolio D, a financial services holding company comprising mutual savings banks (37.5 percent), personal credit (15.5 percent) and business credit (47 percent) institutions, and less than 1 percent savings and loans and bank holding companies would have generated an ROA approximately three times as large as reported by bank holding companies alone, with roughly the same level of risk (coefficient of variation of 25.0 percent vs. 30.6 percent). During the middle period (1981–89), financial institutions with approximately the same degree of risk as bank holding companies, portfolio A, would have allocated more than 93 percent of their assets to mutual savings banks, with the remaining assets invested in personal credit institutions. In contrast to the earlier periods, bank holding companies are included in all of the portfolios over the 1990–97 period in proportions that range from approximately 10 percent to 33 percent. All of the portfolios have a coefficient of variation less than bank holding companies by themselves. Mutual savings banks remain a substantial part of all portfolios and dominate the highest-return portfolio. Savings and loan companies also enter the lower-return efficient portfolios and provide more than one-half of the assets in the lowestrisk portfolio. Personal credit and business credit institutions also enter all but the highest-return portfolio, but these activities generally receive less weight than in prior periods. The variance of returns and the coefficient of variation of returns in the last period are greater than those during the 1974–80 period but less than those of the 1981–89 period. Nontraditional Banking Activities. In terms of nontraditional activities, Table 4 presents a much wider range of possible ROAs, reflecting the greater earnings opportunities available outside traditional banking services. For the two earlier periods, bank holding companies play a significant role but only for portfolios A and B, which yield relatively low ROAs (1.0–1.5 percent). Table 4 indicates that, even at low levels of return, during the 1970s a dramatic increase in ROA could be achieved by diversifying into a variety of nontraditional activities while at the same time reducing risk by approximately 50–60 percent. For example, in portfolio B a financial services company that invests only 14 percent of its assets in bank holding companies and the majority of its assets (70 percent) in life insurance would generate a coefficient of variation of only 6.7 percent and an expected ROA of 1.5 percent, compared with a pure bank holding company that reported a coefficient of variation of 30.6 percent and an average ROA of 0.27 percent during the same period. Of the remaining financial sectors, either life insurance underwriting or insurance agents and brokers dominate the portfolio for higher expected returns. For the 1980s, an optimal portfolio mix suggests that bank holding company involvement disappears entirely beyond an expected ROA of 2.5 percent and a coefficient of variaAs financial holding comtion of 7.2 percent. panies become more diverThese figures compare sified, it will become more quite favorably to an ROA of only 0.26 perdifficult to measure the cent and a coefficient benefits of diversification of variation of 34.4 by simply forming efficient percent for bank holding companies alone portfolios. Diversification during this period. will take place as much During the 1990s the within as between industry picture changes quite dramatically, with bank classifications. holding companies contributing meaningfully in virtually all portfolios with an ROA of 4.5 percent or less (portfolios A–G). The remainder of the efficient portfolios including nontraditional activities consists largely of insurance agents and regulated investment companies. The only exception is the lowest-return portfolio, in which security brokers are approximately one-fifth of the portfolio. Commodity brokers, life insurance underwriters, and subdividers and developers do not enter the efficient portfolio (no greater than 0.1 percent) except for a small share (1.4 percent) devoted to subdividers and developers in portfolio I. The standard deviation of ROA and the coefficient of variation of ROA are higher during the last period than either of the preceding two periods. However, the coefficient of variation of ROA is not 19. The mathematics of linear algebra limits the number of industries to no more than the number of years in a sample. If the optimization program is given more industries than years then one or more industries will become linear combinations of the other industries and the solution to each of the efficient portfolios will not be unique. The table illustrates the optimal combination of each industry for selected ROAs. 20. The program used to form the portfolios is The Investment Portfolio, version 1.0, designed by Edwin J. Elton, Martin J. Gruber, and Christopher R. Blake. The program sometimes had problems forming efficient portfolios when the portfolio share of an industry was allowed to vary between 0 percent and 100 percent. However, in these cases it was able to estimate the efficient portfolio when each asset was constrained to consist of no more than 99.99 percent of the portfolio. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 45 TA B L E 4 Efficient Portfolios of Both Traditional and Nontraditional Financial Services, ROA (Percent)a Traditional Activities 1974–80 A B C D Efficient Risk and Return Combination Mean ROA Standard Deviation ROA Coefficient of Variation 0.5 0.1 26.0 0.7 0.2 24.3 0.9 0.2 24.4 1.0 .2 25.0 Associated Portfolio Allocations Mutual Savings Banks Bank Holding Companies Savings and Loan Associations Personal Credit Institutions Business Credit 80.2 0.0 0.0 5.5 14.3 63.1 0.0 0.0 9.5 27.4 46.0 0.0 0.0 13.5 40.5 37.5 0.5 0.5 15.5 47.0 A B C D Efficient Risk and Return Combination Mean ROA Standard Deviation ROA Coefficient of Variation 0.3 0.1 40.0 0.4 0.3 62.5 0.6 0.5 90.0 0.8 0.8 103.8 Associated Portfolio Allocations Mutual Savings Banks Bank Holding Companies Savings and Loan Personal Credit Business Credit 93.1 0.0 0.0 6.9 0.0 75.9 0.0 0.0 24.1 0.0 41.4 0.0 0.0 58.6 0.0 6.9 0.0 0.0 93.1 0.0 1981–89 1990–97 A B C D Efficient Risk and Return Combination Mean ROA Standard Deviation ROA Coefficient of Variation 0.5 0.1 20.0 0.7 0.2 28.6 0.9 0.3 33.3 1.0 0.4 40.0 Associated Portfolio Allocations Mutual Savings Banks Bank Holding Companies Savings and Loan Personal Credit Business Credit 25.2 10.0 52.4 0.0 12.4 31.0 28.1 33.9 4.3 2.7 39.9 33.4 6.6 20.1 0.0 86.0 14.0 0.0 0.0 0.0 substantially greater and is generally slightly less for each of the portfolios of nontraditional activities than for bank holding companies by themselves. The benefits for diversification into nontraditional financial activities during the two earlier periods were two-dimensional in that the bank holding company could achieve a significant increase in expected earnings while achieving a substantial reduction in risk. On the other hand, the benefits for diversification during the past decade had little to do with risk reduction but appear to have been related almost entirely to a substantial increase in expected earnings. Efficient Portfolios Using Return on Equity. The ratio of equity capital to assets varies across different types of financial services providers.21 Thus, the efficient portfolio allocations using ROE as the measure of profitability may be different from those using ROA. Further, the allocations based on ROE may provide more insight to managers of financial firms, given that theory suggests that firms should 46 focus on maximizing shareholder wealth rather than return on assets. Thus, Table 5 provides efficient portfolio allocations using ROE for the 1990–97 period. The allocations for traditional activities using ROE in Table 5 appear roughly the same as the allocations using ROA in Table 4. The allocations are not exactly comparable because a 10 percent allocation in Table 4 indicates that 10 percent of assets should be invested in the industry whereas the same percentage allocation in Table 5 indicates that 10 percent of equity should be invested in the industry. Bank holding companies constitute a large fraction of the optimal portfolios with intermediate levels of ROE but are not included in either the very low or very high return portfolios. In contrast, mutual savings banks and savings and loans are in all of the efficient portfolios, with mutual savings banks dominating the highreturn portfolios. All of the portfolios of traditional activities had a lower coefficient of variation of ROE than bank holding companies have by themselves, a Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 TA B L E 4 ( C o n t i n u e d ) Nontraditional Activities 1974–80 A Efficient Risk and Return Combination Mean ROA 1.0 Standard Deviation ROA 0.1 Coefficient of Variation 10.0 Associated Portfolio Allocations Bank Holding Companies Security Brokers Commodity Brokers, Dealers Life Insurance Underwriters Insurance Agents, Brokers Regulated Investment Companies Subdividers and Developers 1981–89 Efficient Risk and Return Combination Mean ROA Standard Deviation ROA Coefficient of Variation Associated Portfolio Allocations Bank Holding Companies Security Brokers Commodity Brokers, Dealers Life Insurance Underwriters Insurance Agents, Brokers Regulated Investment Companies Subdividers and Developers 1990–97 a 46.2 6.8 2.9 43.2 0.0 0.0 0.0 B C D E F G H I 1.5 0.1 6.7 2.5 0.2 10.0 3.0 0.4 12.0 3.5 0.5 13.7 4.0 0.6 14.8 4.5 0.7 15.8 5.0 0.8 16.6 6.0 1.1 17.8 13.6 9.7 3.4 70.4 0.0 2.9 0.0 0.0 7.2 6.5 68.0 9.9 8.4 0.0 0.0 6.1 8.4 58.1 17.1 10.2 0.0 0.0 5.0 10.4 48.1 24.3 12.1 0.0 0.0 3.9 12.3 38.3 31.5 13.9 0.0 0.0 2.8 14.2 28.4 38.7 15.7 0.0 0.0 1.7 16.2 18.6 45.9 17.6 0.0 0.0 0.0 18.1 0.0 60.7 21.2 0.0 A B C D E F G H I 1.0 0.1 5.0 1.5 0.1 6.0 2.5 0.2 7.2 3.0 0.2 7.0 3.5 0.3 8.9 4.0 0.4 10.0 4.5 0.5 11.1 5.0 0.6 12.2 6.0 0.9 14.7 76.6 0.0 1.0 6.4 1.3 9.3 5.3 54.3 0.0 1.7 13.6 3.5 15.1 11.8 9.5 0.0 3.1 28.1 7.8 26.6 24.9 0.0 0.0 6.7 26.6 11.2 31.0 24.4 0.0 0.0 12.4 18.6 15.6 34.5 18.8 0.0 0.0 18.7 10.6 19.9 38.2 13.2 0.0 0.0 23.9 2.7 24.3 41.6 7.6 0.0 0.0 26.0 0.0 28.5 45.5 0.0 0.0 0.0 6.2 0.0 32.6 61.2 0.0 C D E F H I A B G Efficient Risk and Return Combination Mean ROA 1.0 Standard Deviation ROA 0.4 Coefficient of Variation 40.0 1.5 0.5 33.3 2.5 0.9 36.0 3.0 1.1 36.7 3.5 1.1 31.4 4.0 1.5 37.5 4.5 1.7 37.8 5.0 1.9 38.0 6.0 2.3 38.3 Associated Portfolio Allocations Bank Holding Companies Security Brokers Commodity Brokers, Dealers Life Insurance Underwriters Insurance Agents, Brokers Regulated Investment Companies Subdividers and Developers 76.1 0.0 0.0 0.0 1.8 22.1 0.0 55.4 0.0 0.0 0.0 19.7 24.9 0.0 45.0 0.0 0.0 0.0 28.8 26.3 0.0 34.4 0.0 0.0 0.0 38.0 27.7 0.0 24.3 0.0 0.0 0.0 46.6 29.0 0.0 13.8 0.0 0.0 0.0 55.8 30.5 0.0 2.9 0.0 0.0 0.0 65.2 31.9 0.0 0.0 0.0 0.0 0.0 98.6 0.0 1.4 69.4 22.6 0.0 0.0 0.0 7.9 0.0 The results are rounded to one decimal place. As a result, the portfolio allocations sometimes sum to slightly more or less than 1.0. Further, the rounding was applied independently to the mean, standard deviation, and coefficient of variation for ROA for the results obtained from Wall, Reichert, and Mohanty (1993). Thus, the coefficient of variation for 1974–80 and 1981–89 is not necessarily equal to the reported standard deviation divided by the mean. Source: Wall, Reichert, and Mohanty (1993) and authors’ calculation using data from IRS corporate income tax returns result that mirrors the finding for the coefficient of variation of ROA for traditional activities. The results forming portfolios using ROE for the nontraditional activities, also seen in Table 5, can be usefully compared with the portfolios formed using ROA. There are several similarities: bank holding companies enter into all of the portfolios except the highest-risk portfolio, insurance agents enter the higher-return portfolios, and regulated investment companies are in the lowest-risk portfolios. Another similarity is that life insurance does not enter any of the portfolios for the 1990–97 period. There are also several differences: security brokers enter the highreturn portfolios using ROE rather than the lowreturn portfolios using ROA, and commodity brokers and dealers dominate the highest-return portfolios 21. Some finance theories hold that the ratio of equity to assets should vary across industries depending upon factors such as the volatility of the industry’s earnings and the costs incurred by the firm if it becomes financially distressed. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 47 TA B L E 5 Efficient Portfolios of Both Traditional and Nontraditional Financial Services, ROE 1990–97 (Percent)a Traditional Activities A B C D E F Efficient Risk and Return Combination Mean ROE Standard Deviation ROE Coefficient of Variation 8.0 1.7 21.3 9.0 1.8 20.0 10.0 1.8 18.0 11.0 2.4 21.8 12.0 3.3 27.5 13.0 4.6 35.4 Associated Portfolio Allocations Mutual Savings Banks Bank Holding Companies Savings and Loan Companies Personal Credit Companies Business Credit Companies 28.9 0.0 8.2 62.8 0.0 29.3 34.0 12.0 24.7 0.0 35.6 49.5 14.9 0.0 0.0 54.9 28.3 16.3 0.0 0.5 74.3 0.0 16.8 0.0 8.9 91.9 0.0 8.1 0.0 0.0 Nontraditional Activities A a B C D E F G H I Efficient Risk and Return Combination Mean ROE Standard Deviation ROE Coefficient of Variation 3.0 1.2 40.0 5.0 1.5 30.0 7.0 2.3 32.9 9.0 3.2 35.5 11.0 4.2 38.2 13.0 5.3 40.7 15.0 6.4 42.7 17.0 7.6 44.7 18.0 8.5 47.2 Associated Portfolio Allocations Bank Holding Companies Security Brokers Commodity Brokers/Dealers Life Insurance Underwriters Insurance Agents, Brokers Regulated Investment Companies Subdividers and Developers 7.6 0.0 0.0 0.0 0.0 83.9 8.5 33.8 0.0 0.0 0.0 0.0 66.2 0.0 65.7 0.0 0.0 0.0 3.4 31.0 0.0 88.6 0.0 0.0 0.0 10.2 1.2 0.0 67.7 6.4 10.6 0.0 15.3 0.0 0.0 45.9 10.3 24.2 0.0 19.6 0.0 0.0 24.6 14.1 37.5 0.0 23.8 0.0 0.0 2.6 18.0 51.3 0.0 28.2 0.0 0.0 0.0 0.0 89.8 0.0 10.1 0.0 0.0 The results are rounded to one decimal place. As a result, the portfolio allocations sometimes sum to slightly more or less than 1.0. Source: Authors’ calculation using data from IRS corporate income tax returns using ROE whereas the industry never enters the efficient portfolio using ROA. Another difference is that several of the portfolios formed using ROE have lower coefficients of variation than bank holding companies by themselves. Nevertheless, portfolios with ROEs as high as 11 percent may be formed from nontraditional activities, which have a lower coefficient of variation of ROE than bank holding companies by themselves. Implications of the Empirical Results. Consistent with Wall, Reichert, and Mohanty’s (1993) findings, the above results suggest that the optimal portfolio is time-varying.22 One new finding is that bank holding companies are a more important element of the efficient portfolio for both traditional and nontraditional activities in the 1990s than they were in the 1970s and the 1980s. Although the empirical analysis is not designed to explain why banks were more important in the 1990s, two plausible hypotheses are that banks have benefited more from the stable macroeconomic environment in the 1990s and that banks may have benefited from relaxed restrictions on their ability to expand geographically and into new product areas. 48 One limitation of the results is that they may imply portfolio combinations that are not feasible, at least for large financial holding companies. For example, the analysis of efficient portfolios suggests that the highestreturn portfolio using ROE invests 90 percent of its equity in commodity brokers and dealers while the highest-return portfolio using ROA over the 1990–97 period invests almost 99 percent of its assets in insurance agents and brokers. While commodity brokers and dealers and insurance agents were especially profitable parts of the financial services industry, they were, in terms of assets, a relatively small portion. Conclusion he Gramm-Leach-Bliley Act sweeps away most of the barriers limiting the affiliation of banks with nonbank financial services providers. The focus now shifts to financial services executives who must decide which combinations provide the best opportunities to increase shareholder wealth. Existing empirical evidence suggests that an important consideration in this decision is the potential gain from portfolio diversification into new activities. The available empirical evidence also T Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 suggests that the potential for such gain clearly exists. However, the results provided by this article emphasize that the efficient combination may change over time, perhaps for reasons such as the macroeconomic environment or technology. One positive result from the analysis for bankers is that in recent years banks have become a larger part of the efficient financial services portfolio than they were in the 1970s or the 1980s. As financial holding companies become more diversified, it will become more difficult to measure the benefits of diversification by simply forming efficient portfolios. Aggregate measures of industry profitability, such as those provided by the IRS, will mask any synergies already present in the performance data. That is, diversification will take place as much within as between industry classifications. On the other hand, the IRS and other data sources may define a new category to capture the performance of diversified financial holding companies. Another trend that is currently unfolding is the trend toward using the Internet for the production and delivery of financial services. In a review of the Gramm-Leach-Bliley Act, Barth, Brumbaugh, and Wilcox (2000) discuss the future of banking and how passage of the act may allow banks to recapture some of their lost market share. They raise the issue as to whether the “portfolio model” as evidenced by Citigroup, which combines in a single institution a wide range of banking, insurance, and investment services, will be competitive in a world of “financial portals” like Yahoo Finance. In a recent study Morgan Stanley Dean Witter (1999) discusses how financial services will possibly evolve on the Internet over the next several years. They project that financial services on the Internet will grow at an annual compound rate of growth of 34 percent from 1999 to 2003. Annual revenue from a wide range of financial services such as banking, brokerage services, auto insurance, and term life insurance, as well as credit card fees, is expected to grow from approximately $100 billion in 1999 to $435 billion by 2003. In terms of broad trends, Morgan Stanley Dean Witter forecasts increased competitive pressures as technological and regulatory impediments fall and as product unbundling and price transparency put a squeeze on profit margins. Their evaluation of various financial models predicts that “vertical portals” will become the ultimate distributor of financial services. Morgan Stanley Dean Witter (1999) defines a financial services vertical portal as a web site devoted to financial services where cyber customers can buy or get information on a wide variety of financial products. Its competitive strength relates to its breadth of product offerings, user friendliness, twenty-four-hour, seven-days-a-week availability, and the ability to customize. At a minimum, a financial vertical portal should allow the user to get current financial information; obtain brokerage services; review account balances; see and pay bills; plan for retirement; purchase life, auto, and home insurance; and obtain a mortgage or a credit card. In practice, financial firms appear to be trying a variety of strategies. Some firms with large investments in brick and mortar are creating or participating in portals (for example, see Power 2000). In contrast, one provider of financial services over the Internet, E*Trade, has acquired a commercial bank, formerly called Telebank, to provide a wider variety of services. Thus, the Internet may reduce the potential synergistic gains from financial companies owning the providers of a variety of financial services. If so, the impact of the Internet will be to reinforce Berger’s (2000) finding that the largest potential benefits of conglomeration lie in reduced financial risk due to portfolio diversification. The findings presented in this article suggest that such portfolio diversification may allow financial firms to earn higher rates of return at little or no increase in risk. However, the results also suggest that the efficient combination of services from a portfolio diversification perspective varies through time. 22. That the optimal portfolios vary through time does not suggest that the government should restrict possible combinations. Firms will restructure (selling some activities and buying others) if the gains from restructuring the firm’s activities are sufficiently large. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 49 A P P E N D I X IRS Industry Categories he IRS corporate returns publication breaks the general finance, insurance, and real estate sector down into twenty-three minor industry groups as follows (a brief description for some of the less obvious industries is included). T Banking • Mutual savings banks • Bank holding companies, including both one bank and multibank holding companies • Independent banks, excluding mutual savings banks and bank holding companies Credit Agencies • Savings and loan associations • Personal credit institutions, which are establishments primarily engaged in providing loans to individuals and establishments engaged in financing retail sales made on the installment plan and automobile financing • Business credit institutions, which are establishments engaged in making loans to business and agricultural enterprises, such as short-term business credit institutions (commercial finance companies), accounts receivable and commercial paper factoring, direct financing of working capital, captive automobile finance companies (for example, GMAC), mercantile financing, and so forth. Real Estate • Real estate operators and lessors of buildings, including firms that operate and lease but do not develop real property, such as operators of commercial and office buildings, retail establishments and shopping centers, and so forth • Lessors of mining, oil, and similar properties • Lessors of railroad property, including firms such as airport leasing offices, landholding offices, and others • Condominium management and cooperative housing associations • Subdividers and developers, including firms engaged in subdividing real property into lots and in developing them for resale on their own account Security, Commodity Brokers and Services • Security brokers, dealers, and flotation companies, including establishments engaged in the purchase, sale, and brokerage of securities and those, generally known as investment bankers, that originate, underwrite, and distribute securities issues • Commodity contract brokers and dealers; security and commodity exchanges; and allied services, firms that buy and sell commodity contracts on either the spot or future basis for their own account or the account of others and that provide investment advice regarding securities to companies and individuals on a contractual or fee basis, and so forth Insurance • Life insurance companies • Mutual insurance companies (except life or marine and certain fire or flood insurance companies) Insurance Agents and Brokers • Agents and brokers dealing in insurance • Organizations offering services to insurance companies and policyholders, such as insurance claim adjusters Holding and Other Investment Companies1 • Regulated investment trusts, including a wide range of firms such as open and closed-end investment funds, money market mutual funds, unit investment trusts, and so forth • Real estate investment trusts (REITs), including firms engaged in closed-end real estate investments or related mortgage assets that meet the requirements of the amended Real Estate Investment Act of 1960, such as mortgage investment trusts, mortgage and realty trusts, and real estate investment trusts • Small business investment trusts (SBITs) 1. Excludes bank holding companies. 50 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 REFERENCES ALLEN, LINDA, AND JULAPA JAGTIANI. 1999. “The Impact of New Bank Powers (Securities and Insurance Activities) on Bank Holding Companies’ Risks.” Federal Reserve Bank of Chicago Emerging Issues Series Paper S&R-99-1E, September. BARTH, JAMES R., DAN BRUMBAUGH JR., AND JAMES A.WILCOX. 2000. “The Repeal of Glass-Steagall and the Advent of Broad Banking.” Office of the Comptroller of the Currency Economic and Policy Analysis Working Paper 2000-5, April. BERGER, ALLEN N. Forthcoming. “The Integration of the Financial Services Industry: Where Are the Efficiencies?” North American Actuarial Journal. BOYD, JOHN H., AND STANLEY L. GRAHAM. 1986. “Risk, Regulation, and Bank Holding Company Expansion into Nonbanking.” Federal Reserve Bank of Minneapolis Quarterly Review 10 (Spring): 2–17. BOYD, JOHN H., STANLEY L. GRAHAM, AND R. SHAWN HEWITT. 1993. “Bank Holding Company Mergers with Nonbank Financial Firms: Effects on the Risk of Failure.” Journal of Banking and Finance 17 (February): 43–63. FEIN, MELANIE L. 1997. Securities Activities of Banks. 2d ed. New York: Aspen Publishers, Inc. KWAN, SIMON. 1998. “The Securities Activities by Commercial Banking Firms’ Section 20 Subsidiaries: Risk, Return, and Diversification Benefits.” Federal Reserve Bank of San Francisco Working Paper 98-10, May. LITAN, ROBERT E. 1985. “Evaluating and Controlling the Risks of Financial Product Deregulation.” Yale Journal on Regulation 3 (Fall): 1–52. MORGAN STANLEY DEAN WITTER EQUITY RESEARCH. 1999. The Internet and Financial Services Report: Financial$ervices.com. New York: Morgan Stanley Dean Witter. POWER, CAROL. 2000. “Citi Advances in Online Payments, Aggregation.” American Banker 165, July 19, 1. WALL, LARRY D. 1987. “Has Bank Holding Companies’ Diversification Affected Their Risk of Failure?” Journal of Economics and Business 39 (November): 313–26. WALL, LARRY D., ALAN K. REICHERT, AND SUNIL MOHANTY. 1993. “Deregulation and the Opportunities for Commercial Bank Diversification.” Federal Reserve Bank of Atlanta Economic Review 78 (September/October): 1–25. WHALEN, GARY. 1999a. “The Risks and Returns Associated with the Insurance Activities of Foreign Subsidiaries of U.S. Banking Organizations.” Office of the Comptroller of the Currency. Unpublished paper. ———. 1999b. “Trends in Organizational Form and Their Relationship to Performance: The Case of Foreign Securities Subsidiaries of U.S. Banking Organizations.” Journal of Financial Services Research (September/ October): 181–218. LADERMAN, ELIZABETH S. 2000. “The Potential Diversification and Failure Reduction Benefits of Bank Expansion into Nonbank Activities.” Federal Reserve Bank of San Francisco Working Paper 2000-01, January. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 51 The Shifty Laffer Curve Z S O LT B E C S I The author is an economist in the regional section of the Atlanta Fed’s research department. It has been said that the virtue of the Laffer curve is that you can explain it to a congressman in half an hour and he can talk about it for six months. —Hal Varian, Intermediate Microeconomics W ITHOUT TAXES THERE ARE NO GOVERNMENT SERVICES. PEOPLE UNDERSTAND THIS REALITY BUT ALSO PREFER TO GET THE MOST FROM THEIR GOVERNMENTS AT THE LEAST COST. IN THE UNITED STATES, ANY NUMBER OF POLITICIANS AT THE LOCAL, STATE, AND FEDERAL LEVELS OWE THEIR SUCCESS TO EMPHASIZING TAX CUTTING. According to logic, this voter response means that people are opting for fewer government services across-the-board or are voting for changes in the mix of services rendered. It is at this point that things become complicated because what happens to expenditures influences how much revenue a government needs to collect. In other words, tax policy cannot be made in isolation from expenditure policy because the mix of expenditures affects economic activity and thus the revenue yield from tax policy. To understand the impacts of tax policy, one needs to know what determines tax revenues. A good place to start is with what is popularly known as the Laffer curve, which shows how tax rates and tax revenues are related.1 Essentially, the Laffer curve posits that as tax rates rise continuously from zero, tax revenues rise up to some maximum after which tax revenues fall. This curve became famous early in the 1980s when supply-side theorists argued that lower tax rates would mean higher revenues because existing rates were too high to maximize tax revenues—that is, tax rates were so high that fewer taxed goods were being produced and the overall effect was lower tax revenues. While conceptually simple, the Laffer curve came under increasing scrutiny after tax cuts based on supply-side arguments apparently failed to “deliver the goods.” Tax rates fell but tax revenues did not rise accordingly, and the United States resorted to deficit spending. In part, the expected outcome did not occur because there are important theoretical limitations that produce the deceptive simplicity of the Laffer curve. This article examines the macroeconomic and conceptual issues that may have made a difference.2 Understanding these considerations may shed more light on why the 1980s supply-side experiment did not produce the desired results. It should also help frame future budget discussions. Because most analyses of the Laffer curve occur in a static framework that has proved inadequate, this analysis presents a simple dynamic model that resembles the discussion in Baxter and King (1995). This framework is useful for analyzing the long-run effects of tax policies.3 In addition, the model can easily be extended to analyze the disposition of government revenues and the consequent effects on national Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 53 income. It turns out that how the government spends its tax revenues—on consumption, investment, or transfers—is important for understanding the Laffer curve. In fact, a different Laffer curve is associated with the different ways revenues are spent, and it is important to know which curve one is operating on when designing tax policies. Otherwise, one might be riding the wrong curve, so to speak, and thus miscalculating revenue effects. Background erhaps one of the first things one learns in studying the economics of taxation is that taxes alter equilibrium prices and quantities of taxed goods. A tax on any good x introduces a tax wedge between the price demanders pay and the price suppliers receive. Thus, the equilibrium quantity of good x will fall unless demand or supply is perfectly inelasTax policy cannot be made tic. When the tax rate is adjusted upward, in isolation from expenditure tax revenues will rise policy because the mix of as long as the perexpenditures affects ecocentage rise in the tax rate exceeds the pernomic activity and thus the centage fall in quanrevenue yield from tax policy. tity. However, as one lets the tax rate rise at a given percentage rate, the quantity of x falls, implying that the percentage change of quantity will rise. At some point the percentage fall in quantity dominates the percentage rise in tax rates so that further tax rate increases cause tax revenues to fall. At the point at which tax revenues begin to fall, tax revenues are at a maximum.4 This revenue-maximizing point is a sort of Holy Grail for policymakers interested in defending the impact of various budgetary reforms. One can easily see these points in a simple demand-and-supply graph (see Chart 1). The intersection of supply and demand gives the before-tax equilibrium quantity, Q*, and price, P*. Introducing a tax drives a wedge between the price demanders pay and the price suppliers receive. Thus, a tax causes equilibrium quantity to fall to Q** and the before-tax price to rise to P**. The after-tax price is the before-tax unit price after taxes have been subtracted, or P** – T. At Q** the amount of tax revenues collected is given by the rectangle Q** × T. As can be easily verified by comparing rectangles for different tax rates, tax revenues first rise as tax rates are raised from small levels because the tax P 54 rate effect on revenues tends to dominate. But after a while tax revenues start to shrink because the quantity effect dominates the tax rate effect. The rate at which the revenue-maximizing point occurs determines whether tax rates for a given product should be raised or lowered from current levels. The answer depends in part on the relative demand-and-supply elasticities, or how sensitive quantity demanded or supplied is to price changes. Generally, the more inelastic and the steeper the curves are, the higher the revenue-maximizing tax rate is. This relationship holds because the percentage reduction in quantities tends to be small and less likely to dominate a given tax rate change than if curves were more elastic. This pattern can be easily verified by drawing steeper demand or supply curves in Chart 1 and comparing rectangles for a given tax rate. As a rule, demand or supply curves tend to be more inelastic the more broadly the tax is defined or the fewer substitution possibilities there are (either on the supply or demand side). For example, the revenue-maximizing tax rate on chocolate bars will tend to be lower than the revenue-maximizing tax rate on food, both of which in turn are likely to be lower than the revenue-maximizing rate on cigarettes. Similarly, the revenue-maximizing state sales tax rate should be lower than for federal sales taxes given that people can avoid state taxes by moving. The theoretical Laffer experiment deals only with the effects on revenues from changing tax rates. However, in the real world tax rates are usually not changed in isolation. What the government does with the revenues it receives will also determine where revenues are maximized. So far it has been assumed that the government did nothing with its revenues so that expenditures had no effects. This scenario is essentially like assuming that the government wastes its revenues, no better than throwing them into the ocean. If instead tax revenues were returned lumpsum to taxpayers, or in a way that would not affect taxpayers’ behavior, the negative wealth effects of the tax would be offset. This approach would increase tax revenues relative to throwing the money away. However, because the taxed activity has become more expensive relative to untaxed activities, a substitution effect remains whereby the quantity of the taxed activity falls relative to all other activities. But what if the government actively spends its revenues, as it invariably does? If the government uses revenues to buy more of the taxed good, it will increase the demand for the good. This move will tend to offset the decline in quantity caused by the tax increase, and both tax revenues and the revenuemaximizing tax rate will tend to rise. Finally, if the revenues are used to add to the public capital stock, Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 C H A R T 1 Derivation of Laffer Curve D C P3** Supply B P2** Price A P2**–T2 Implies C Revenues P4** • Demand B D • P3**–T3 • P4**–T4 A Q4** Q3** Q2** Q* 0 Quantity T2 T3 T4 1 Tax Rate the supply of good x may increase and, again, the quantity decline will be offset and the revenuemaximizing tax will tend to be higher.5 Graphically, when revenues are used by the government to increase demand, an outward shift occurs at the same time the tax is imposed. As seen in Chart 2, the shift in demand counteracts the quantity reduction when taxes are raised in isolation. Thus, equilibrium quantity falls by a lesser amount than before. Also, as can be seen by comparing revenue rectangles, tax revenues rise by a larger amount than if no taxes are raised. This observation suggests that the revenue-maximizing tax rate under a balanced-budget policy is higher than if expenditures do not keep pace with tax revenues.6 Alternatively, when revenues are used to increase the supply of the good, the supply curve shifts to the right instead of the demand curve. However, the qualitative result is the same. Fullerton (1982) summarizes the Laffer curve literature. For the most part, this literature was comfortable with the assumption that tax revenues adjust smoothly to tax rate changes.7 Strong assumptions about the shape of individual preferences and firm-production functions were employed by theorists and empiricists alike. This literature also tended to use mostly static frameworks. Thus, the focus of the research was to empirically investigate the shape of the Laffer curve and determine where current tax rates were on this curve. The majority of the papers found that for U.S. income taxes, tax rates were on the upward-sloping portion of the Laffer curve. Thus, it was assumed, a reduction of income tax rates would lower tax revenues. With Malcomson (1986), studies began probing the strong assumptions leading to a simple Laffer curve using static general equilibrium models.8 Guesnerie and Jerison (1991) show for general demand functions and technologies that Laffer curves can have many shapes. Their argument is consistent with the idea that when the Laffer curve exhibits several peaks, moving to one peak may not maximize revenues 1. The idea behind the Laffer curve has been around for a long time, as long as 200 years by some accounts. See Fullerton (1982) and Blinder (1981) for historical references. 2. There are also empirical limitations, but the focus of the article is on the macro and conceptual issues. 3. The model is also simple enough to allow an explicit solution. It is related to simple models found in Becsi (1993) and Koenig and Huffman (1998). While supply-side arguments for lowering tax rates rely heavily on the growth effects of fiscal policies, the model can easily be extended along the lines of Ireland (1994). 4. The existence of a revenue-maximizing point can be proved using elementary calculus. All that is needed is the assumption that tax revenues are a continuous and differentiable function of tax rates. Also, tax revenues must be zero when tax rates are zero or when tax rates are at some very high rate. With these assumptions, Rolle’s Theorem states that there exists a tax rate such that tax revenues are maximized. 5. Of course, raising public capital may also affect demand inasmuch as it affects the utility derived from good x. Symmetrically, public consumption may affect the supply side. Thus, public consumption and investment will be treated symmetrically in utility and production in this article. 6. In the case of very high tax rates, where higher rates in isolation mean lower revenues, a balanced-budget approach might cause an offset to the reduction in tax revenues. 7. In other words, the mathematical assumptions of Rolle’s Theorem (see Blinder 1981) were respected. 8. See also Malcomson (1988), who shows that the tax function could be discontinuous at some tax rates. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 55 CHART 2 Tax Revenues with and without Government Spending This formulation says that as the µi parameters rise, public services substitute more closely for a unit of private consumption. Similarly, total output is defined as the sum of private output produced for profit, y, and output produced as a direct by-product of government activities: Tax Rate E {{ T3 • • A T2 • Demand Q2** Q3** Q* Quantity globally unless it is the highest one. Finally, Gahvari (1989) shows that how the budget is balanced when tax rates are changed will affect the shape of the Laffer curve. In particular, a lump-sum transfer leads to a normally shaped Laffer curve while government consumption may eliminate the downward-sloping portion. Essentially, the positive effects on production of an increase of government spending may dominate the contractionary quantity effects of rising tax rates. If the expansionary effects are strong enough, an increase in tax rates will always be associated with an increase in total revenues. This article elaborates on this last view. Description of the Model his section develops a simple dynamic macroeconomic model consisting of household, production, and government sectors. To study the long-run effects of taxes, attention is turned to the steady-state equilibrium of the model where all variables are constant through time. Despite its simplicity, the model is a useful starting point for analyzing the steady-state effects of various fiscal policies. In particular, it allows one to explore the Laffer curve in a long-run context and also illustrates how the Laffer curve depends on the disposition of tax revenues. To start the analysis of how public expenditures affect household and firm decisions, it is useful to look at broad measures of consumption and output. First, composite consumption is defined as private consumption, c, plus the services derived from public consumption, cg, and public capital, kg.9 In short, composite consumption, x, is given by T x ≡ c + µ c c g + µ k k g. 56 y + Ac c g + Ak k g. Supply B This formulation says that a unit of government expenditures will increase total output by Ai. In other words, Ac is the marginal product of public consumption, and Ak is the marginal product of public capital. While this specification is very simple, it has the drawback that private and public output are substitutes. It is assumed that households would like to maximize composite consumption and leisure obtained in each period of their lives.10 However, they are constrained by their budgets. In other words, purchases of consumption goods and savings can never exceed after-tax earnings from working and past savings. The solution of this problem leads to wellknown optimality conditions for constrained utility maximization: the marginal rate of substitution (MRS), which equals the ratio of the marginal utilities of two goods, is equated to the price ratio of the two goods. In other words, the MRS is the rate at which the individual is willing to sacrifice one good in return for another to keep lifetime utility constant. The price ratio is the rate at which the two goods can be substituted and still satisfy the budget constraint. The difference between the MRS and the price ratio is that the former is determined by individuals’ tastes and the later is determined by the marketplace. Optimality means simply that tastes and market realities are in harmony.11 Optimality forces households to adjust consumption and labor until the marginal rate of substitution of composite consumption and leisure is equal to the after-tax wage rate: MRSh = (1 − t y )w, (1) where h is the fraction of time a person spends working. Alternatively, 1 – h is the fraction of time devoted to leisure. To understand this equation, consider what happens when an individual works more. Suppose the increase in work time is ∆h. In this case utility will fall with the reduction in leisure time unless consumption rises sufficiently. Consumption must rise by MRSh × ∆h to keep utility constant. Thus, MRSh gives the desired increase in consumption for a unit increase of labor (or unit loss of leisure). Alternatively, the budget constraint Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 indicates that if labor rises by ∆h, after-tax labor earnings will rise by (1 – ty)w∆h units. Thus, consumption can rise only as much as labor income. To see that individuals will adjust their consumption and labor until the MRS equals the price ratio, suppose that the MRS is smaller than the price ratio. In this case, a given reduction of leisure will be rewarded with more consumption (from additional wages) than individuals require to keep utility constant. Thus, they will work more because overall utility rises when work effort and consumption are increased. As labor and consumption are increased, the MRS rises because leisure is scarcer and further sacrifice requires more consumption in order to keep utility constant. Finally, the MRS will rise until condition (1) is satisfied. Households adjust consumption and savings across time until the MRS of consumption in adjacent periods equals the after-tax interest rate: budget constraint allows. Thus, current consumption will be raised relative to future consumption. In the production sector firms use labor and private capital to produce their output. Competitive firms vary their labor and capital mix until profits are maximized. Profit maximization by the firm implies that the firm adjusts inputs until its marginal products equal its factor costs. These conditions can be succinctly represented with a small amount of notation. The marginal product of labor is denoted MPh and is the additional output from varying labor by one unit. Similarly, MPk is the marginal product of physical capital. Also, the unit cost of labor is the wage rate, w, and the cost of capital is the rental rate, r. With this notation, firms maximize profits when MPh = w (3) MPk = r. (4) and MRSx = r(1 − t y ). (2) The logic behind this condition is similar to that of condition (1). When current consumption is reduced by ∆x, the next period’s consumption must rise by –MRSx × ∆x to keep utility constant. In steady state, the MRSx reflects an individual’s impatience to consume early. An impatient household requires a higher return for a sacrifice of current consumption. From the budget constraint, decreasing current consumption by ∆x allows the household to increase savings by ∆k = –∆x. An increase in savings will cause next period’s earnings to rise by r(1 – ty)∆k, which is the increase in capital earnings from additional savings. Thus, the price ratio in equation (2) measures how much additional future consumption one can have if current consumption is reduced by one unit. If condition (2) does not hold with equality, then households will adjust their savings. For instance, if the MRS exceeds the price ratio, then the individual requires more future consumption to keep utility constant for the unit sacrifice of current consumption than the Intuitively, when the firm is in a situation in which the marginal product of an input exceeds the unit cost of the input, profits can be raised by hiring more of the input in question. As more of the input is employed, the marginal product tends to fall because of diminishing returns. Hiring of the input will proceed until the marginal products again equal marginal costs. Finally, the public sector pursues a balancedbudget strategy and purchases consumption and investment goods and makes lump-sum transfers, l g, from the proceeds of its income tax collections. The government’s budget constraint is described by c g + k g + l g = t y ( wh + rk), (5) where the right-hand side of the equation depicts the source of tax revenues from labor and capital income and the left-hand side shows uses of funds. 9. The public good aspects of public consumption such as spending on health care, housing, education and defense will affect individual utility. Some of these expenditures will be closer substitutes for private spending than others. The services from public capital such as highways and streets, educational structures, and public utilities could also enter private utility. 10. Literally, it is assumed that lifetimes are infinite, an assumption that can be viewed as a useful abstraction of long lives. In addition, technically oriented readers will find it useful to know that the model assumes that lifetime preferences are intertemporally separable and that preferences over consumption and leisure are logarithmic. Furthermore, production is Cobb-Douglas (see the appendix), and capital depreciates fully in each period. As is well known, these popular assumptions yield an explicit solution and can be a useful starting point for dynamic analyses. However, it must be noted that the strong assumptions on the form of the utility and production functions may limit the shape of the associated Laffer curves. 11. Again note that long-run optimality conditions are derived by assuming that a steady state exists. A household is in steady state when asset holdings do not change across time; thus, consumption, labor, and savings are time invariant and time subscripts can be dropped. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 57 All markets are assumed to equilibrate in all periods. Thus, aggregate demand equals aggregate supply, or c + k + c g + k g = y + Ac c g + Ak k g . (6) Here total output supplied by firms is given on the right-hand side. The left-hand side shows private and government demand. This equation is just another way of writing the gross domestic product (GDP) identity with the government sector broken out. Description of Steady-State Equilibrium he six equations introduced above are enough to describe a simple economy in steady state and deduce the effects of income taxes and the effects of public spending.12 Equations (2) and (4) together determine the marginal The Laffer curve became product of private capfamous early in the 1980s ital and also the private capital-labor ratio. when supply-side theorists Thus, raising the inargued that lower tax rates come tax rate reduces would mean higher revenues the after-tax marginal product of private because existing rates were capital below its equitoo high to maximize tax librium level. To rerevenues. store the steady-state marginal product of capital, the firms cut back on capital, thus causing the capital-labor ratio to rise and the productivity of capital to rise. While the income tax has a large effect on the productivity of capital, government consumption and investment do not have any effect. The effect of these variables on the model economy is through the GDP identity, which is considered next. Once the productivity of capital is determined, equation (6) determines the share of output that goes to consumption. Thus, anything that enhances the productivity of capital will raise the consumptionoutput ratio. Furthermore, an increase in the fraction of output devoted to public consumption or public investment will lower the fraction of output that goes to consumption. However, care must be taken to distinguish between demand and supply effects of government spending. If the marginal product from public input is zero so that there are no supply effects, then crowding out of consumption is one-forone. To include supply effects one must also keep track of the productivity of government spending. If the marginal product of public services is greater T 58 than zero, the share of consumption remaining for output will fall by less than one-for-one. Because it is likely that the marginal product for public capital exceeds the marginal product of public consumption, public consumption will have a greater crowding-out effect than public capital. Given the consumption-output ratio, equations (1) and (3) pinpoint the steady-state level of labor. The focus is on three ways that labor in this economy is altered. First, anything that causes the consumptionoutput ratio to rise raises MRSh in equation (1). Because MRSh exceeds the price ratio, individuals adjust consumption and leisure to reduce MRSh and bring equation (1) back to equality. As discussed previously, households work and consume less and increase the time devoted to leisure. Second, given the ratio of consumption to output, a rise in the income tax rate lowers the after-tax marginal product of labor in equation (3). To restore the equilibrium marginal product, work effort must fall because of diminishing returns. At the same time, this falling work effort lowers MRSh in equation (1) until households are happy with a lower after-tax marginal product of labor. Finally, given the consumptionoutput share, increasing the output share of public consumption or capital tends to raise MRSh. This effect induces households to substitute away from consumption toward leisure and to reduce aggregate labor. However, the substitution effect on labor is offset more when there is a greater decline in the consumption share. So far, the equilibrium capital-labor ratio (or productivity of private capital), the equilibrium level of labor, and the consumption-output ratio have been determined. Because private output is produced with private capital and labor, it is easy to find, given that equilibrium labor and capital and the form of aggregate production are known. Qualitatively, output changes will reflect input changes, and the effects of the various policy changes on output will be traced out below. It is also possible to calculate the effect on consumption and capital of a policy change because it is known how the consumption-output ratio and the capital-output ratio (or productivity of private capital) respond as well as how output responds. Finally, it should be noted that although the productivity of capital is not observed, the real (inflationadjusted) interest rate, which in equilibrium reflects the marginal product of capital, is observed. Theoretical Effects of Balanced-Budget Income Tax Changes A Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 s discussed above, a simple income tax will cause private inputs to fall. Increasing the income tax causes the ratio of private capital to private output to fall and the private capital-labor ratio to rise because it lowers the after-tax marginal product of private capital. Because the marginal product of labor is also lowered and the capitallabor ratio has already been determined, labor must fall. Thus, output and the private capital stock fall in the long run. Because the productivity of private capital falls, the consumption-output ratio rises. The total effect on consumption seems uncertain because the share of consumption rises at the same time private output falls. Normally, these two factors combine to raise consumption (at the expense of savings and output). Finally, notice that tax revenues rise or fall depending on whether output falls proportionately less than the tax rate rises. To keep its budget balanced, the government has to do something with the revenue change. Thus, the effects of different expenditure strategies must be weighed against the effects of the tax rate changes. A lump-sum transfer or tax has only wealth effects and does not affect the long-run equilibrium at the margin. On the other hand, increasing public consumption or capital will affect the steady state of the economy much as it was shown to do in the simple demandand-supply analysis at the outset of this paper. Suppose public consumption adjusts with income tax rates to balance the budget. While the capitaloutput ratio is unaffected because the after-tax marginal product of capital is unchanged, the consumptionoutput ratio falls since fewer resources are left over. The share of consumption falls less than one-for-one if the marginal product of the government expenditures is positive.13 It can be shown that the increase in the share of public services and the fall of the consumption-output ratio together cause the marginal rate of substitution of leisure and consumption to fall below the market wage. Bringing the marginal rate of substitution back into equilibrium requires increasing consumption, but doing so is only possible by working more. However, since more labor implies that the productivity of capital rises, private capital rises to keep the capital-labor ratio constant. The rise in private inputs increases income tax revenues and raises private (and total) output. Thus, an income tax with budget-balancing public consumption causes a smaller reduction in GDP than if expenditures did not change. How do the effects of public consumption differ from the effects of public investment? The differ- ence depends on the relative marginal products of consumption and investment and on their relative substitutability with private consumption. It seems reasonable that the marginal product of public capital is greater than the marginal product of public consumption. Assume that Ac < Ak and for simplicity that µc = µk, and let the share of public consumption and the share of public investment increase equally. In this case, the consumption-output ratio is crowded out to a greater extent by a rise in the share of public consumption than by public investment. This relationship exists because increasing public capital raises total production more, leaving more resources for consumption. However, since the consumptionoutput ratio falls more with public consumption, the marginal utility of consumption rises more. Thus, to reequilibrate the optimal marginal rate of substitution, households increase their work effort more with A different Laffer curve is public consumption associated with the differthan with public inent ways revenues are vestment. Thus, private capital, labor, spent, and it is important output, and consumpto know which curve one tion rise more when is operating on when public consumption is increased than when designing tax policies. the share of public capital rises by an equal amount. In essence, since increasing the share of public capital causes total output to increase more, private inputs (and output) are required to rise less than with an equal increase in the share of public consumption. Since factor incomes rise less with an increase in public investment than with an increase in public consumption, tax revenues rise less, too. Just as reasonable is the supposition that public consumption is a closer substitute for private consumption than for public investment. Assume that µc > µk and for simplicity that Ac = Ak. Thus, increasing the share of public consumption or investment reduces the share of consumption equally. However, the marginal utility of consumption rises by a greater amount with public capital because for a given increase in public capital composite consumption will 12. The discussion focuses on an illustrative case that allows a closed-form solution (see the appendix). The solution is simplified by assuming that all forms of government expenditure can be written as linear functions, εiy, of “private” output, y. In this case, it is possible to write all endogenous variables as linear functions of y and then solve for y itself. 13. Since increasing the share of public consumption (or investment) also tends to lower the marginal utility of consumption, the negative effect on consumption is reinforced. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 59 CHART 3 Laffer Curves under Alternative Spending Arrangements fall more than with a decrease in public consumption. Thus, increasing the share of public capital will increase labor, private capital, and output more than an equal increase in the share of public consumption will. In summary, the expansionary effect of an increase in public consumption or investment is positively related to the substitutability with private consumption and with the size of the marginal product. While it is easy to imagine that µc > µk or that Ac < Ak, it is more difficult to see what the overall effect might be. This issue is analyzed in the next section. • Revenues • Government Consumption Government Investment • Evaluating Laffer Curve Experiments hich of these competing influences on labor, private capital, and output tends to dominate? It turns out that the net effect of an increase in the share of public expenditures εi can be described very simply. It can be shown that in this simple model the effect of εi on labor, private capital, and private output is proportional to (1 – Ai – µi) for i = c, k.14 In other words, the effect of any government expenditure adjusts the pure demand effect by subtracting a supply effect Ai and a demand substitution effect µi. A few studies have tried to quantify (1 – Ai – µi). Overall, the evidence seems to suggest that A c + µc ≤ Ak + µk.15 Thus, it seems likely that public consumption will have a stronger positive effect on labor, private capital, and private output while public capital will have a stronger positive effect on total output.16 In particular, increasing public consumption at the expense of public capital will raise private inputs and tax revenues but lower total output. This finding has strong implications for the Laffer curve since the response of total revenues to a change in the income tax depends on changes in income from private inputs. Increasing the income tax rate tends to raise the average tax rate and to reduce private inputs. As tax rates continue to rise, the percentage fall in private-factor income eventually dominates a given percentage rise in the income tax rate. At this point, total revenues will begin to fall if tax rates rise any further. Since lump-sum transfers have no long-run macroeconomic effects, balancing the budget with lump-sum transfers will not affect the Laffer curve. In contrast to lump-sum transfers, increasing the share of public capital will cause private-factor incomes to rise, offsetting the tax-induced contractionary effect. Thus, with budget-balancing increases of public capital, tax revenues will be higher than if lump-sum transfers were used. As indicated in Chart 3, the Laffer curve with public capital expenditures will be above the Laffer curve for lump-sum transfers. It also can be shown that the revenue-maximizing Lump-Sum Taxes W 60 0 1 Tax Rate income tax rate will be greater when public capital is used than when it is not. The downward-sloping part of the Laffer curve occurs at higher tax rates on the higher curve than on the lower curve. In other words, it is less likely that tax revenues increase when income tax rates and public capital are reduced simultaneously than when lump-sum transfers have been reduced. Lastly, increasing the share of public consumption is likely to cause income from private inputs to rise more than if public capital were increased. Thus, tax revenues will be higher if government consumption is used to balance the budget. Equivalently, the Laffer curve for public consumption lies above the Laffer curve of public investment (and it can be shown that the revenue-maximizing tax rate will be higher, too). This possibility is also depicted in Chart 3 along with the other two possibilities. Which of the three Laffer curves in Chart 3 is the correct one for the 1980s under the Reagan administration? Answering this question requires a quick look at the data, which reveals three important features of the times. The two well-publicized features are the federal marginal tax cuts and the deficit-financed spending (on transfers and government consumption).17 Another important feature of the data for the period is that public capital investment dropped relative to public consumption, continuing a trend started in the mid-1960s (see Chart 4).18 Thus, to some extent higher government consumption was paid for by lower government investment. When government consumption is increased at the expense of government investment, the total effect on tax revenues equals the effect on GDP that is proportional to Ac + µc – Ak – µk. Because GDP falls, less revenue is collected than before at prevailing tax rates. In essence such a policy shifts all existing Laffer curves down. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 CHART 4 Nondefense Government Investment as a Share of Total Government Expenditures, 1950–99 Share 0.16 0.14 0.12 1960 1970 1980 1990 Note: This series takes a ratio of nondefense government investment and total government expenditures. Both series include local, state, and federal expenditures. Now suppose that the preexisting income tax rate was higher than it ought to be to maximize revenues. In other words, suppose that the prevailing tax rate was on the downward-sloping portion of the Laffer curve as indicated by point A in Chart 5. Under these circumstances, lowering the tax rate would tend to increase revenues. However, if government consumption rises at the expense of government investment, the Laffer curve shifts down. Thus, rather than rising on the original curve, tax revenues fall from point A in Chart 5 to point B. At this point, lowering of tax rates would still increase revenues. However, the additional revenues from lowering tax rates would be insufficient to offset the decline in revenues brought about by the expenditure switch. Thus, it seems that supply-siders may have overlooked an important determinant of the position of the Laffer curve.19 Under the Clinton administration there have been two developments with implications that can be explained using the current analysis: tax rates have risen, and government investment has risen relative to government consumption.20 If one 14. This relationship can be shown by totally differentiating the closed-form solution in the appendix. 15. Aschauer (1989a, 1989b, 1990) cites evidence that the marginal product of public consumption, Ac, is close to zero and that the marginal rate of substitution between private and public consumption, µc, is in the range (0.2, 0.4). However, Kuehlwein (1992) finds no evidence for the substitutability of public and private consumption. Thus, µc is more likely in the range (0, 0.4). To date there exists no empirical evidence on the size and sign of µk. Aschauer (1990) finds that the marginal product of public capital, Ak, may be close to four. Tatom (1991a, 1991b), however, argues that these estimates may be overstated by 40 percent, if not more. 16. Notice that when Ak + µk < (>) 1 an increase in public investment will crowd private capital in (out). Aschauer (1989a) argues that public capital may have two effects. First, if public capital raises the marginal productivity of private capital, it will crowd private capital in. Second, if public capital rises, it will raise output creating a positive wealth effect for households, which will raise consumption and lower savings. Thus, private capital is crowded out. Aschauer finds that the first effect comes to dominate over time. While this article does not consider this effect, assuming a small enough Ak + µk is a rough approximation. For the second effect Aschauer seems to assume that Ak + µk > 1. 17. The calculation abstracts from the deficit-financed increase in government spending because ultimately it must be paid for with future tax increases, future spending reductions, or higher growth of incomes. Ireland (1994) shows that deficitfinanced increases in government spending will eventually pay for themselves through higher growth. However, it may take a long time. 18. Note that the chart compares nondefense government investment to total expenditures. Both investment and expenditure numbers include outlays at the local, state, and federal levels. Also see, for instance, Baxter and King (1995). 19. One implication of this analysis is that empirical studies of the Laffer curve must carefully control for the effects of all types of government expenditures. 20. This statement refers to Chart 4. Government investment and consumption numbers include expenditures at the local, state, and federal levels. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 61 CHART 5 Laffer Curve Depicting the Switch from Government Investment to Consumption as Tax Rates Are Reduced Revenues • •A Consumption •B Investment 0 1 Tax Rate believes that the downward-sloping portion of the Laffer curve is relevant, then such a policy would be a move from point B to point A in Chart 5. However, many economists would argue that the United States is on the upward portion of the Laffer curve. In this case, the positive effect on tax revenues from an increase in tax rates would be reinforced by the shift in government expenditures. In either case, the analysis suggests higher tax revenues, an outcome the data bear out. 62 Conclusion ature and tax policy abhor a vacuum. If tax policy is designed without reference to expenditure policy, it is possible that the effects on tax revenues may be miscalculated. To make this case, a simple neoclassical growth model was developed and the long-run effects of government expenditures and income taxes were analyzed. It was shown that a reduction of tax rates would increase income from labor and private capital and would increase output. Reducing public capital at the same time will tend to lower private inputs and production and thus lower income tax revenues, in turn reducing the tax revenues derived from a cut in income tax rates. The larger the productivity of public capital is or the more precipitous its decline, the likelier it is that tax revenues will fall. By this argument, cutting income taxes at the same time that public investment falls and government consumption rises, as occurred in the 1980s, increases the likelihood that the government loses tax revenues. In this case, a revenueincreasing strategy would have been to lower income tax rates but increase public investment at the expense of government consumption. As a general rule, raising public investment relative to public consumption will tend to add to tax revenues. More importantly, realizing that the Laffer curve is shifty (in the sense that it moves with external shocks) should lead to better tax-policy design. N Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 A P P E N D I X A Closer Look at the Model Households maximize the utility function where total output in steady state is Σ s ≥1(1 / ρ)s −1[ln( cs + µ c cg + µ k ksg−1 ) + α ln(1 − hs )] subject to a budget constraint that is summarized by y + y g = (1 + Ac ε c + Ak ε k ) y. Furthermore, dividing both sides of the government’s revenue constraint by y implies cs + ks = (1 − t y )( ws hs + rs ks–1 ) + l sg , ε c + ε k + ε l = ty. where lsg is the lump-sum transfer (or tax), ks–1 is physical capital accumulated up to period s, and ks is the additional holdings of capital. This equation implies the following first-order conditions that correspond with equations (1) and (2) in the text: MRSh ≡ α / (1 − h ) = (1 − t y ) w, 1 / ( c + µ c c g + µ k kg ) (A1) and MRSx ≡ ρ = r(1 − t y ), (A2) respectively, with subscripts dropped to indicate that variables are in steady state. Firms produce according to a Cobb-Douglas production function, y = kθh1–θ. Under these circumstances the first-order conditions corresponding to equations (3) and (4) in the text are MPh ≡ (1 − θ )( y / h ) = w, (A8) Finally, the market clearing conditions now look like c + k + ε c y + ε k y = (1 + Ac ε c + Ak ε k ) y. (A9) As long as the marginal products of the public inputs are less than unity, the demand effects of public expenditures dominate the supply effects. Using the last five equations, a closed-form solution to the model is easily found. The solution proceeds much like the exposition in the text. From (A6), steadystate capital is a linear function of equilibrium output. Thus, the average productivity of capital is given by k (1 − t y )θ = . y ρ (A10) Substituting (A10) into (A9) yields c / y = 1 − (1 − Ac )ε c − (1 − Ak ) ε k − ( k / y ), (A11) (A3) and which in turn, after substitution into (A5), yields MPk ≡ θ( y / k) = r, (A4) respectively. Combining household and firm-optimality conditions and imposing a steady state yields α (A7) (c / y + µ c ε c + µ k ε k ) y = (1 − t y )(1 − θ )( y / h ) (A5) 1– h h= (1 − t y )(1 − θ ) (1 − t y )(1 − θ ) + α [(c / y ) + µ c ε c + µ k ε k ] . (A12) Then, output can be found by rewriting the production relationship as θ y = ( k / y ) 1− θ h (A13) and inserting (A10) and (A12). Finally, consumption and capital are found by multiplying (A10) and (A11) with (A13). and ρ = (1 − t y ) θ ( y / k), (A6) Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000 63 REFERENCES ASCHAUER, DAVID A. 1989a. “Does Public Capital Crowd Out Private Capital?” Journal of Monetary Economics 24 (September): 171–88. IRELAND, PETER N. 1994. “Supply-Side Economics and Endogenous Growth.” Journal of Monetary Economics 33 (June): 559–71. ———. 1989b. “Is Public Expenditure Productive?” Journal of Monetary Economics 23 (March): 177–200. KOENIG, EVAN F., AND GREGORY W. HUFFMAN. 1998. “The Dynamic Impact of Fundamental Tax Reform—Part 1: The Basic Model.” Federal Reserve Bank of Dallas Economic Review (First Quarter): 24–37. ———. 1990. “Is Government Spending Stimulative?” Contemporary Policy Issues 8 (October): 30–46. BAXTER, MARIANNE, AND ROBERT G. KING. 1995. “Fiscal Policy in General Equilibrium.” American Economic Review 83 (June): 315–34. BECSI, ZSOLT. 1993. “The Long (and Short) of Taxation and Expenditure Policies.” Federal Reserve Bank of Dallas Economic Review (Third Quarter): 51–64. BLINDER, ALAN S. 1981. “Some Thoughts on the Laffer Curve.” In The Supply-Side Effects of Economic Policy, edited by Laurence H. Meyer. St. Louis: Center for the Study of American Business at Washington University and Federal Reserve Bank of St. Louis. FULLERTON, DON. 1982. “On the Possibility of an Inverse Relationship between Tax Rates and Government Revenues.” Journal of Public Economics 19 (October): 3–22. KUEHLWEIN, MICHAEL A. 1992. “Disaggregate Evidence on the Substitutability between Public and Private Spending.” Hoover Institution Working Paper E92-16, June. MALCOMSON, JAMES M. 1986. “Some Analytics of the Laffer Curve.” Journal of Public Economics 29 (April): 263–79. ———. 1988. “Some Analytics of the Laffer Curve: Reply,” Journal of Public Economics 35 (February): 131–32. TATOM, JOHN A. 1991a. “Public Capital and Private Sector Performance.” Federal Reserve Bank of St. Louis Review 73 (May–June): 3–15. ———. 1991b. “Should Government Spending on Capital Goods Be Raised?” Federal Reserve Bank of St. Louis Review 73 (March–April): 3–15. VARIAN, HAL R. 1990. Intermediate Microeconomics. 2d ed. New York: W.W. Norton Company. GAHVARI, FIROUZ. 1989. “The Nature of Government Expenditures and the Shape of the Laffer Curve.” Journal of Public Economics 40 (November): 251–60. GUESNERIE, ROGER, AND MICHAEL JERISON. 1991. “Taxation As a Social Choice Problem: The Scope of the Laffer Curve Argument.” Journal of Public Economics 44 (February): 37–63. 64 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 2000