The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Economic •»»i-i Review FEDERAL RESERVE BANK OF ATLANTA MARCH/APRIL 1990 De Novo Bank Performance FEDERAL RESERVE BÀ OF PHILADELPHIA! A PROPOSAL TO EASE LDC DEBT Call Options on Commodities W E. B 5 f t R Economic Review President Robert P. Forrestal Senior Vice President and Director of Research Sheila L. Tschinkel Vice President and Associate Director of Research B. Frank King Research Officers W i l l i a m C u r t H u n t e r , Basic R e s e a r c h Mary Susan Rosenbaum, Macropolicy G e n e D. Sullivan, Regional L a r r y D. W a l l , F i n a n c i a l D a v i d D. W h i t e h e a d , R e g i o n a l Public Information Officer B o b b i e H. M c C r a c k i n Publications L y n n H. F o l e y , P r o d u c t i o n |. E d w a r d R o o k s , G r a p h i c s a n d T y p e s e t t i n g M i c h a e l ). C h r i s z t , C i r c u l a t i o n The Economic Review seeks t o inform t h e p u b l i c a b o u t Federal Reserve policies and t h e economic environment and, in particular, t o narrow t h e gap b e t w e e n specialists and concerned laypersons. Views expressed in t h e Economic Review are not necessarily those of this Bank or of the Federal Reserve System. Material may b e r e p r i n t e d or abstracted if t h e Review and author are credited. Please p r o v i d e t h e Bank's Public Inform a t i o n Department with a copy of any p u b l ication containing r e p r i n t e d material. Free subscriptions and l i m i t e d additional copies are available from t h e Public Information Department, Federal Reserve Bank of Atlanta, 104 M a r i e t t a Street, N.W., Atlanta, Georgia 30303-2713 (404/521-8788). Change-of-address notices and subscription cancellations should b e sent directly t o t h e Public Information Department. Please include t h e current m a i l i n g label as well as any new information. ISSN 0 7 3 2 - 1 8 1 3 V O L U M E L X X V , N O . 2, M A R C H / A P R I L 1990, E C O N O M I C R E V I E W Covered Call Options: A Proposal to Ease LDC Debt Steven P. Feinstein and Peter A. Abken J 4 2 Determinants of De Novo Bank Performance This article proposes a method by which less developed countries can reduce their massive d e b t b u r d e n s - s e l l i n g covered calls on their chief export commodities. Which factors are most influential in determining a William C. Hunter and Aruna Srinivasan newly chartered i n d e p e n d e n t bank's probability of success? F.Y.I. Measuring Interstate Migration William ). Kahley 4 J Book Review Larry D. Wall The Separation of Commercial and Investment Banking: The Glass-Steagall Act Revisited and Reconsidered by George J. Benston COMMODITY FUTURES NVrjch 11, Covered Call Options: A Proposal to Ease LDC Debt Steven P. Feinstein and Peter A. Abken The large debt burden carried by many developing nations not only has hampered these countries' economic development but has also threatened their commercial bank lenders. Economic austerity measures aimed at reducing this debt have severely limited the funds less developed countries have available for domestic investment, further retarding their economic growth. The authors' proposal—that LDCs sell long-term, high strike price call options on their chief export commodities—would allow LDCs to generate revenue without renegotiating existing debts or giving up ownership of productive resources. any less developed countries (LDCs) struggling to pay off massive loans p o s s e s s an untapped, e x p o r t a b l e resource for which a demand exists in develo p e d nations. Revenue from such a resource would not only help to service d e b t s but also further the i n d e b t e d developing countries' economic advancement. This potential source M 2 C A T T L t i E p » £ R . (Liuti H t o M * ) p . , lb ca«5 SirtWr P„li M i r < Apr <• MavìUv-t. V i ' p A f r f Mav ? »VI 0.00 on 0 do Aft ov o 70 J«P v , î of d e b t relief is not a physical commodity but rather a type of financial asset—namely longterm, high strike price call options on their chief export commodities. The proposal in this article offers interested parties such as commodity users or speculators the opportunity to bid on the right to purchase a certain quantity of a chief export commodity for E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 a prespecified price on a given future date. In making these call options available, LDCs would in effect b e sel ling commodity price i n s u r a n c e insurance that pays off if the commodity price rises beyond s o m e high level. For example, a bidder might purchase the right (but not the obi igation) to buy Mexican oil at $27 a barrel on a given date five years hence, even though the price of oil at the time of bidding is only $ 17.25. Should the prevailing market price remain b e low the "strike" level of $27, the option owner would simply elect not to carry out the transaction. 1 Should oil rise to $30 a barrel by t h e agreed-upon expiration date, though, the bidder could purchase the oil at $3 below prevailing prices. Although the LDC must then relinquish g o o d s for l e s s than t h e going rate, revenue from the earlier option sales would have ameliorated matters during a period of low prices. The loss of potential profits would c o m e at a time of high prices when the selling country could b e s t afford it. Using call options in meeting d e b t obligations is not an entirely new idea. Numerous articles in academic journals have suggested construction of financial instruments that tie a borrower's liabilities to a commodity price, and the recent Mexican loan restructuring (described below) includes such a feature. 2 In other plans, however, options have b e e n bundled together with bonds. By selling options separately, as proposed here for the first time, an LDC can generate substantial revenue without renegotiating all its outstanding debts. Such an "unbundled" approach is possible with this proposal b e c a u s e it calls for selecting a strike price high enough to e n s u r e that t h e debtor will face the future obligation only when fulfillment of all other obligations is relatively easy. This feature would obviate the n e e d to gain the endorsement of existing creditors prior to selling the options. Keeping the options separate from d e b t instruments also affords the LDC greater flexibility in the management of its revenue flow and debt. A program of selling options when commodity prices are low and redeeming them when prices are high can help an LDC smooth revenue across periods of high and low commodity prices. A further advantage is that the market for the unbundled option is likely to b e wider than the market for options coupled with d e b t . The World Debt Problem—A Review Since 1982 the high i n d e b t e d n e s s of many developing nations has placed a severe strain on both the LDCs and the world banking community. According to the most recent accounting, LDCs owe $1.3 trillion to foreign banks, governments, and international agencies. 3 Not only have LDCs struggled to service this debt, b u t e x p o s u r e to t h e s e t r o u b l e d loans has threatened the strength of commercial banks. T h e roots of the problem can b e traced to the volatility of commodity prices and interest rates over the last two decades. The oil price increases of 1973 and 1979 led to a massive redistribution of wealth from oil importers to oil exporters, primarily those in the Middle East. T h e s e "petrodollars" accumulated in commercial banks around the world as surplus funds awaiting investment opportunities. Because of their apparent excellent potential to develop rapidly, Latin American economies were d e e m e d to b e excellent credit risks. As a result, t h e s e countries received enormous loans from commercial banks. T h e loans were predominantly short-term, with interest rates tied to the London Interbank Offer Rate (LIBOR).4 By 1980 three-fourths of the d e b t owed by Latin American countries was s e t at variable rates; 40 percent was due for repayment within one year, and 70 percent within three years. 5 From 1971 to 1980, LIBOR was on average 0.8 percent (80 basis points) less than the rate of U.S. wholesale price inflation; developing countries were effectively borrowing at very low or even negative real interest rates. 6 Later, however, rates on LDC d e b t surged as the United States tightened monetary policy in an attempt to control inflation. LIBOR averaged 9.2 percent (920 basis points) above the U.S. inflation rate from 1981 to 1982. 7 As rates rose, d e b t service obligations soared. The worldwide recession that accompanied higher interest rates further hurt LDCs. Reduced The authors are economists in the financial section of the Atlanta Fed's research department. They would like to thank Marco Espinosa, Larry Wall, and Robert Kahn for helpful comments. They gratefully acknowledge research assistance provided by Karen Hunter and Hermawan Djohari. The authors accept sole responsibility for any errors. I7 FEDERAL RESERVE BANK O F ATLANTA d e m a n d for their exports and plummeting c o m modity prices c o m p o u n d e d their misfortune. As t h e value of imports began to overwhelm that of exports, current account payment b a l a n c e s of t h e highly i n d e b t e d countries turned sharply negative. 8 By t h e e n d of 1982, 34 developing c o u n t r i e s w e r e u n a b l e to s e r v i c e their d e b t fully. 9 As oil exporters, Mexico and Venezuela were special c a s e s . As oil prices leveled off and then b e g a n to drop, however, t h e s e two countries also began to have difficulty making payments on t h e substantial d e b t they had accumulated against future oil revenues. T h e p r e d i c a m e n t has taken a heavy toll on t h e LDCs. In net terms Latin America and t h e Caribb e a n exported capital in each of t h e last eight years; t h e 1989 net outflow a m o u n t e d to $24.6 billion ( s e e Barbara Durr 1989). As a result, d o m e s t i c investment and d e v e l o p m e n t have suffered. Growth over t h e last d e c a d e has slowed to a crawl, making r e p a y m e n t of t h e foreign d e b t even m o r e difficult and less likely. 1 0 Moreover, standards of living have declined. B e t w e e n 1980 and 1987, per capita consumption in t h e LDCs fell nearly 12 p e r c e n t (see Giancarlo Perasso 1989, 535). If t h e quality of life for t h e citizens of LDCs is to improve, many believe, capital must flow into t h e s e c o u n t r i e s on net, n o t in t h e o p p o s i t e direction. L e n d e r s have also suffered. In 1982 t h e sum of LDC loans on t h e b o o k s of U.S. banks was over 180 p e r c e n t of t h e capital in t h o s e banks (Jeffrey Sachs 1989a). S i n c e then, t h e market value of LDC loans has fallen sharply below b o o k value. For e x a m p l e , by March 1990, Peruvian loans c o u l d b e sold for only six c e n t s on t h e dollar, and t h e market valued Mexican d e b t at 40 perc e n t of face value. 1 1 Consequently, bank earnings have suffered, and b a n k stocks have ref l e c t e d d e t e r i o r a t i n g loan p o r t f o l i o v a l u e s (Sachs 1989a). Nonetheless, considerable progress has b e e n m a d e toward alleviating t h e crisis. Banks have r e d u c e d their LDC loan e x p o s u r e and increased capital, t h e r e b y easing fears of bank insolvency. By 1988, aggregate e x p o s u r e stood at l e s s than 80 p e r c e n t of capital (Sachs 1989a). This improvement afforded s o m e breathing room and m a d e it p o s s i b l e for banks to offer d e b t relief. R e c e n t l y , U.S. Treasury S e c r e t a r y N i c h o l a s Brady's initiative to r e d u c e LDCs' d e b t burden h e l p e d Mexico restructure its commercial b a n k 4 d e b t . Mexico received c o n c e s s i o n s on principal and interest, along with new loans (Peter Truell 1990). T h e LDC d e b t problem, however, is far from over. No single approach is likely to work across t h e board, and in m o s t e a s e s d e b t reduction will probably b e part of future financial arrangements. Yet developing countries that rely solely on d e b t reduction may find it difficult to borrow in t h e future. Nor is rescheduling old loans and securing new loans a satisfactory long-term solution. Many e x p e r t s now agree that t h e problem is o n e of solvency rather than a lack of liquidity available to t h e LDCs (Anna J. Schwartz 1989). In other words, t h e s e nations' e c o n o m i c p r o s p e c t s are sufficiently b l e a k that markets doubt LDCs can make good on their past financial obligations. Consequently, as long a s stretching payments out over longer horizons p r e s e r v e s t h e net p r e s e n t value of LDCs' liabilities, such a tactic will not end t h e crisis. Highly i n d e b t e d c o u n t r i e s m u s t t h e r e f o r e search for ways t o restore their solvency. Essentially, in order to pay off their huge d e b t s , they must raise more revenue. B e t t e r e c o n o m i c planning and m e a s u r e s to s t e m d o m e s t i c capital flight are t h e sorts of actions that will have lasting positive effects (A. Schwartz 1989 and Rudiger Dornbusch 1987). S o m e policymakers and analysts advocate d e b t - e q u i t y swaps a s m e a n s by which LDCs can pay off loans without incurring new liabilities; however, t h e s e transactions r e p r e s e n t p e r m a n e n t s a l e s of an LDC's productive resources (such as forests, mines, and factories) a n d s o c a n n o t p r o v i d e r e v e n u e on a continuing basis. C o n s e q u e n t l y , such swaps serve only as temporary palliatives. Furthermore, t h e manner in which d e b t - e q u i t y swaps are e x e c u t e d usually results in money supply expansion in t h e LDC, thus fueling inflationary pressures. 1 2 A New Proposal Selling high strike price call options on export c o m m o d i t i e s is o n e way LDCs could g e n e r a t e revenue without selling their productive resourc e s or renegotiating existing d e b t . Thus c o m modity price volatility, which contributed to t h e d e b t crises in t h e first place, could provide a E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 partial key to its solution. It is precisely the volatility of LDC export prices that makes high strike price call options valuable today, just as uncertainty makes various forms of insurance more desirable. Because a call option affords the buyer the right, but not the obligation, to buy at a given price on a given date, investors find it profitable to exercise the option only in the event that the commodity's market price rises above the strike price. Thus, an LDC that sells high strike options incurs an obligation that n e e d s to b e fulfilled only if commodity prices increase substantially— a circumstance in which fulfillment of all obligations would b e easier for the LDC. Selecting a high strike price for the call options enhances the instrument's marketability. High strike price call options n e e d not b e designated as "senior" obligations (those that must b e met in advance of existing debt) in order to attract buyers. Although call option transactions entail s o m e risk of default, t h e option's value remains high, even when default risk and junior designation are taken into consideration, b e c a u s e the chance of default is small when the commodity reaches the high strike price. (See the example of option pricing in the box on page 10.) As junior obligations, option s a l e s would not require approval by existing creditors—an obstacle that frequently blocks additional loans to highly i n d e b t e d countries. Since LDCs would b e issuing call options on their own export commodities, the countries would bear little financial risk in meeting their obligations. When calls are covered, the issuer simply sells the available commodity to the option holder for the strike price. When call options are not covered, that is, when the underlying commodity is not in the possession of the issuing party, the issuer takes the chance of having to purchase quantities of the commodity at high prices to fulfill the contract. LDCs could either sell covered calls to new investors or swap them for a portion of outstanding debt. To ensure a fair price to the LDC for the calls, the sale and distribution could b e conducted via a closed-bid auction in a manner similar to the Morgan Guaranty Mexican debtswap of February 1988. LDCs could s e t a minimum acceptable price in advance, or they could retain the right to reject low bids. The covered call is a simple use of options. Additional flexibility could b e achieved through use of combinations of options, which would, for example, keep an LDC from sacrificing all of the additional revenue forgone if prices rose above the strike price. A more elaborate use of options could tailor the payoff contingencies to b e t t e r suit the LDC and potential investors. One such strategy is the call option spread, which would provide a cap on the payout in the event of a very high commodity price. A spread is conceptually equivalent to the LDC's selling a call at a given strike price and simultaneously buying a call at a higher strike price. 1 3 This combination entitles the investor to an increasing payout as the commodity price rises in the range between the two strike prices. If the commodity price rises further, though, the cap c o m e s into play. LDC governments might prefer this arrangement for pol itical reasons. Setting a cap on the contingent liability would also b e advantageous in the c a s e of agricultural commodities, for which quantity risks caused by uncertain harvests are significant. Supporting Theory As is nowclear, the LDC is not the only party to bear commodity price risk; the LDC's creditors share that risk. While periods of high prices are characterized by large capital inflows and rapid growth, low prices strain LDC economies and often result in d e b t crises as the revenue n e e d e d to service outstanding d e b t dries up. At the same time d e b t held by the LDC's creditors fluctuates in value according to the LDC's creditworthiness, which in turn d e p e n d s on the commodity price. If the risk of price fluctuation s u b j e c t s both the LDC and the creditor to the specter of default during low-price times, and if default is costly to the two parties, then s o m e sort of revenue smoothing is advantageous to both parties. One way to smooth revenue across high- and low-price periods is for the LDC to sell "claims" on the high-price period. High strike price options are exactly such a claim. They allow the LDC to transfer money from potential high-price periods into an immediate low-price period. According to option pricing theory, the more volatile the commodity price is, the more such I7 FEDERAL RESERVE BANK O F ATLANTA Table 1. Estimated Value of Vulnerable Oil Options For Various Terms and Strike Prices (Johnson-Stulz Pricing Method: Spot Oil Price = $17.25; Oil Price Volatility = 28 percent per year; Initial Pool Value = $10.77 per option; Pool Volatility = 42 percent per year; Correlation between Oil Price and Pool Value = .5; Interest Rate = 8 percent per year) . 0, Strike pnce per barrel Years until expiration: 2 3 4 5 6 7 $25 $1.27 $2.01 $2.59 $3.04 $3.39 $3.65 26 1.11 1.83 2.42 2.88 3.23 3.51 27 .96 1.67 2.25 2.72 3.09 3.38 28 .84 1.51 2.10 2.57 2.96 3.25 29 .73 1.44 1.95 2.43 2.84 3.13 30 .63 1.25 1.82 2.30 2.70 3.01 31 .55 1.14 1.69 2.17 2.57 2.89 32 .48 1.04 1.58 2.06 2.46 2.78 an option is worth. Consequently, t h e u s e of LDC c o v e r e d calls is m o s t f e a s i b l e in exactly t h o s e c a s e s where it is most necessary. An Example of an LDC Covered Call While t h e features and theoretical underpinnings of LDC covered calls are straightforward, a realistic e x a m p l e can clarify t h e features outlined. Mexico o w e s $ 1 0 0 . 3 billion t o foreign interests. 1 4 Payment of principal and interest e x c e e d e d $15 billion in 1988, an a m o u n t equal to 4 6 p e r c e n t of export r e v e n u e s . 1 5 Yet Mexico consistently produces b e t w e e n 880 million and 1.1 trillion barrels of oil each year. S i n c e 1982 over 4 7 0 million barrels per year have b e e n e x p o r t e d . 1 6 Proven reserves amount to 70 billion barrels. Thus, t h e supply and production of Mexican oil is reliable. Under conditions prevailing in t h e s e c o n d quarter of 1989, when t h e market price for Mexican oil s t o o d at $17.25 per barrel, an option to sell oil five years later at a price of $27 would have b e e n worth approximately $2.72 ( s e e T a b l e l). 1 7 This figure is c o m p u t e d using Herb Johnson and R e n e S t u l z ' s (1987) pricing model, which is similar to t h e Black-Scholes model but 6 also accounts for default-risk. 1 8 T h e box on page 10 d e s c r i b e s t h e methodology and explains t h e assumptions and parameter values used. With a price of $2.72 for an option on o n e barrel of Mexican oil, a sale of call options on o n e year's quantity of oil exports could net Mexico $1.28 billion in current r e v e n u e (Table 2). A s a l e of o p t i o n s on five y e a r s ' e x p o r t s would bring $6.4 billion. 1 9 This r e v e n u e would b e suffic i e n t to retire 6.37 p e r c e n t of Mexico's foreign bank d e b t at face value, and m o r e than 15.9 perc e n t at current market discounts. By comparison, t h e recent Mexican financing package negotiated under t h e Brady plan framework extinguished $7 billion in commercial b a n k d e b t . In addition, a fixed r e d u c e d interest rate was s e c u r e d on over $22 billion in claims. However, this d e b t relief was partially offset by new loans totaling nearly $ 1.4 billion from banks and a b o u t $5.75 billion from official s o u r c e s to provide for principal repayments and guarantee interest ( s e e Jorge C. C a s t a ñ e d a 1990 and Truell 1990). T h e Mexican government has e x p r e s s e d an int e r e s t in further reducing their commercial bank d e b t over time, and t h e r e v e n u e raised by t h e proposal p r e s e n t e d h e r e could allow such reductions without a drawdown in reserves. T h e s a l e of o p t i o n s is not a s a l e of oil; it merely s e t s t h e highest price that Mexico can E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 Table 2. Estimated Revenue to Mexico For Issues of Options with Various Terms and Strike Prices (billions of U.S. dollars) (Based on Johnson-Stulz vulnerable option valuation; assumes options cover 470 million barrels, approximately one year's exports; all other assumptions are as given in Table 1) Years until expiration: Strike price per barrel 2 3 4 5 6 7 $25 $.60 $.94 $1.20 $1.40 $1.60 $1.70 1.50 1.70 26 .52 .86 1.10 1.40 27 .45 .79 1.10 1.30 1.50 1.60 28 .39 .71 .99 1.20 1.40 1.50 29 .34 .68 .92 1.10 1.30 1.50 .30 .59 .86 1.10 1.30 1.40 31 .26 .54 .79 1.00 1.20 1.40 32 .23 .49 .74 .97 1.20 1.30 30 charge t h e option purchaser for oil during t h e expiration year. Should t h e price of oil remain low, Mexico would b e free t o sell oil in any manner it c h o o s e s . Only if t h e price r e b o u n d s and e x c e e d s t h e $27 strike price would Mexico b e obliged to sell oil to t h e option holders for $27 per barrel. At that price, though, t h e flow of revenue would b e sizable, and Mexico's financial situation would b e greatly improved. In either c a s e , t h e s a l e of covered calls would provide Mexico with a d d e d r e v e n u e now, r e v e n u e that could b e u s e d to e a s e t h e current financial crisis. T a b l e l p r e s e n t s t h e Johnson-Stulz prices for vulnerable Mexican oil options—that is, t h o s e s u b j e c t to default risk—using various c o m binations of strike prices and maturities. T a b l e 3 p r e s e n t s Black-Scholes prices, which a s s u m e no d e f a u l t risk b u t c o r r e s p o n d t o t h e s a m e parameter values. (These p a r a m e t e r s are given in t h e box and in t h e tables.) T h e difference b e tween the Black-Scholes and johnson-Stulz prices is t h e credit spread, that is, t h e reduction in t h e value of t h e option that is d u e to t h e risk of default. As a p e r c e n t a g e of t h e option price, t h e credit s p r e a d is lower for higher strike prices. T h e attractiveness of t h e option is that its c o n tingent liability is paid only when funds are plentiful and thus effectively d o e s not c o m p e t e with existing d e b t for available funds. Mexico is not t h e only e x a m p l e of a country that could b e n e f i t from selling call o p t i o n s on an export commodity. Brazil, for example, could sell options on s o y b e a n s and coffee. Chile and Peru might sell options on c o p p e r , Bolivia, tin options. Any commodity-exporting country could m a k e u s e of a similar strategy. T a b l e 4 lists several LDC c a n d i d a t e s and t h e i r principal export commodities. The Proposal from the LDCs' Perspective A s a l e of high strike price covered calls would p r o v i d e an LDC with m u c h - n e e d e d r e v e n u e when revenue is otherwise scarce. Unlike a d e b t equity swap—which involves p e r m a n e n t s a l e of productive resources, often at d e p r e s s e d prices— t h e LDC covered call entails no loss of control or ownership over productive resources. S a l e of LDC c o v e r e d calls, on t h e other hand, relinq u i s h e s only s o m e potential profits for a fixed a m o u n t of time—that is, t h e difference b e t w e e n t h e market price of t h e commodity and t h e option's strike price (if t h e difference is positive) for a p r e s p e c i f i e d quantity of output. T h e obligation would b e "costly" to t h e LDC in high-price periods b e c a u s e t h e LDC would I7 FEDERAL RESERVE B A N K O F ATLANTA Table 3. Estimated Value of Default-Free Oil Options For Various Terms and Strike Prices (Black-Scholes Pricing Method: Spot Oil Price = $17.25; Oil Price Volatility = 28 percent per year; Interest Rate = 8 percent per year) Years until expiration: Strike price per barrel 2 3 4 5 6 $25 $1.41 $2.45 $3.47 $4.43 $5.33 $6.18 26 1.23 2.23 3.22 4.17 5.08 5.93 27 1.06 2.02 2.99 3.93 4.83 5.69 28 .92 1.83 2.78 3.71 4.60 5.45 29 .80 1.66 2.58 3.49 4.38 5.23 30 .69 1.51 2.39 3.29 4.18 5.02 31 .60 1.37 2.22 3.11 3.98 4.82 32 .52 1.24 2.07 2.93 3.79 4.63 fulfill its obligation by selling at a lower-thanm a r k e t p r i c e ; however, t h a t p r i c e would b e much higher than t h e market price had b e e n at the t i m e t h e option was written. Thus, t h e transaction would still b e profitable to t h e LDC and would further e n h a n c e its welfare, d e s p i t e its obligatory nature. Furthermore, if t h e LDC had u n u s e d p r o d u c t i o n capacity, e x e r c i s e of t h e o p t i o n s would provide additional customers and greater total revenue. Should t h e commodity price not rise a b o v e t h e strike price, t h e option would expire unexercised and t h e LDC would face no further enc u m b r a n c e . In order to s u p p l e m e n t i n c o m e during t h e continued low-price state, t h e LDC might then wish to issue additional covered calls against another future period's production. The Proposal from the Creditor's Perspective S a l e s of c o v e r e d calls by LDCs would b e n e f i t creditors for several reasons. T h e new instrum e n t e m b o d i e s s o m e attractive i n v e s t m e n t features. LDCs' selling covered calls r e p r e s e n t s a f e a s i b l e alternative to d e m a n d i n g new loans from creditors. T h e r e v e n u e covered calls provide LDCs would e n h a n c e t h e val ue of t h e other 8 7 LDC d e b t held by t h e creditors. Although cove r e d calls require no servicing prior to expiration, unlike d e b t forgiveness, they have inherent value; thus t h e accounting t r e a t m e n t is m o r e favorable to banks than is outright d e b t forgiveness. A n o t h e r a d v a n t a g e o u s f e a t u r e of an LDC covered call is its ability to let t h e investor/ creditor share m o r e fully in t h e fortunes of t h e LDC during t i m e s of increasing c o m m o d i t y prices. A portfol io of loans to an LDC carries cons i d e r a b l e e x p o s u r e to commodity price risk, but offers limited reward should prices rise. LDC covered calls would grant creditors a c c e s s t o g r e a t e r u p s i d e p o t e n t i a l . Just a s c o m m o d i t y prices are theoretically u n b o u n d e d , t h e p o t e n tial gain from owning an LDC covered call would b e unlimited. Although t h e option would expire worthless should commodity prices not recover, t h e s a m e could b e said of LDC d e b t . Nonperformance on loans is q u i t e p o s s i b l e during t h e low-price p e r i o d s in which an option would expire out of t h e money. Thus, LDC d e b t and LDC covered calls have certain downside features in c o m mon. Certainly t h e r e a r e p r i c e s c e n a r i o s in which LDC d e b t would perform b e t t e r than LDC c o v e r e d calls; n e v e r t h e l e s s , s o m e c r e d i t o r s / investors might prefer t h e different risk-return profile of t h e option. E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 Another potentially attractive feature of LDC covered calls could b e their greater liquidity relative to LDC bank d e b t and equity stakes. As a standardized security, LDC covered calls should b e easier to sell than bank d e b t and equity holdings. Indeed, an active market for LDC covered calls might develop in response to their availability. Banks would then have an avenue to reduce their LDC exposure, should they wish to, by swapping bank d e b t for calls and then selling those calls in the market. The proposal should also appeal to creditors b e c a u s e of the effect it would have on the entire portfolio of LDC investments. The options would only pay off when LDC funds are plentiful, and the initial sale or swap of the options would reduce the LDC's d e b t burden. Hence, the LDC could better service all of its obligations regardless of the behavior of commodity prices. The LDC would potentially have more funds for development and thus face enhanced future financial prospects. Accordingly, the value of all the LDC's outstanding d e b t could appreciate. Clearly, purchase of LDC covered calls is preferable to d e b t forgiveness from the creditors' point of view. Like d e b t forgiveness, selling LDC covered calls would relieve the debtor of s o m e d e b t service obligations and make service of remaining d e b t more manageable. This change would enhance the value of remaining debt, just as d e b t forgiveness d o e s . By receiving LDC covered calls, however, creditors would maintain a claim on LDC funds that might b e c o m e available at a later date. Furthermore, whereas banks must write off d e b t that is forgiven, receipt of calls would preserve s o m e capital since LDC covered calls are a valuable asset. The Market for LDC Covered Calls Users of the LDC's export commodity face price risk exactly opposite of that borne by the LDC: users suffer when prices are high and prosper when prices are low. Purchase of an LDC c o v e r e d call would b e a form of insurance against excessive price hikes. It would insulate the user against severe fluctuations in the commodity price, thereby facilitating investment planning and marketing decisions. Should the price remain low, the option would expire Table 4. Selected Heavily Indebted Countries and Their Chief Export Commodities Export Revenue from Commodity as a Percent of Total Exports, 1982-88 Country Commodity Argentina Wheat Corn 10.8 7.8 Bolivia Natural Gas Tin 49.3 25.1 Brazil Soybeans Coffee Chile Copper 45.4 Colombia Coffee Fuel Oil 39.8 13.4 Ecuador Crude Oil Bananas 54.7 9.4 Côte d'Ivoire Cocoa Beans Coffee 29.7 18.3 Mexico Petroleum 57.0 Morocco Phosphates 18.5 Peru Copper Zinc 16.8 9.5 Philippines Coconut Products 10.8 Uruguay Wool 19.7 Venezuela Petroleum 86.0 9.5 8.4 Source: International M o n e t a r y Fund. worthless, much in the s a m e way an accident insurance policy returns nothing when no accident occurs. During low-price periods, however, the option owner would continue to enjoy the low price of the essential commodity. Processors and distributors of oil products, along with the U.S. Department of Energy, which purchases oil to supply the nation's strategic reserves, might b e potential buyers of LDC covered calls on oil. Agents who wish to speculate on the price of oil might also b e interested, and they would add liquidity to the market. Investment managers could u s e LDC covered calls to hedge over the long term against the adverse effects commodity price shocks can I7 FEDERAL RESERVE BANK OF ATLANTA exert on investment portfolios. Currently, few instruments are available that allow a g e n t s to s p e c u l a t e or h e d g e prices over t h e long term. Potential Problems This plan is s u b j e c t to criticism in several a r e a s , and certain d e t a i l s would have t o b e worked out. For example, creditors and investors might fear that an LDC would r e n e g e on its obligation should t h e price of oil rise substantially a b o v e t h e option strike price. This b e havior is unlikely, however, s i n c e it would b e akin to default on a loan. An LDC's initial s a l e of LDC covered calls in order t o retire d e b t would exhibit a willingness to honor international financial a g r e e m e n t s . For countries striving t o d o s o during hard times, reneging during e a s i e r circ u m s t a n c e s s e e m s unlikely. T h e LDC would have to b e a r inflation risk. Worldwide inflation over t h e life of t h e option might lower t h e real value of t h e strike price at which t h e commodity would have to b e sold. If this occurred, however, t h e s a m e inflation that would m a k e fulfillment of t h e option obligation costl ier to t h e LDC would also rel ieve t h e LDC of s o m e of its d e b t burden. Inflation i n c r e a s e s t h e real c o s t of t h e option obligation but d e c r e a s e s the real cost of t h e d e b t obligations. As long a s the term of t h e options were similar to or shorter than t h e interval at which banks reprice loans in r e s p o n s e to inflation a n d changing i n t e r e s t rates, t h e two countervailing inflation effects c o u l d o f f s e t o n e a n o t h e r . In this way, LDC covered calls could possibly offer creditors and d e b t o r s a h e d g e against inflation risk. Logistical details n e e d to b e a d d r e s s e d on a c a s e - b y - c a s e basis. For instance, to avoid simultaneous e x e r c i s e of all options, which would put a t r e m e n d o u s strain on t h e LDC's ability to deliver t h e commodity, t h e option could b e written s o that a forward delivery contract would b e sold to t h e holder, with t h e delivery d a t e d e t e r m i n e d by t h e order in which t h e e x e r c i s e r e q u e s t is received. Conclusions High strike price covered call o p t i o n s are a market-oriented solution that should b e explored further and given serious consideration as a r e s p o n s e to t h e current LDC d e b t crisis. B e c a u s e their u s e offers LDCs a new source of revenue without adding to an already difficult d e b t burden, sale of call o p t i o n s would b e n e f i t both LDCs and t h e holders of existing LDC d e b t . By smoothing income across p e r i o d s of high and low commodity prices and providing insurance against periods of excessively high commodity prices, this financial instrument holds advant a g e s for LDCs, c o n s u m e r s of their export c o m m o d i t i e s , and investors. Thus, it p r e s e n t s a promising new approach to easing d e b t obligations and furthering e c o n o m i c a d v a n c e m e n t in highly i n d e b t e d countries. An Explanation of t h e Johnson-Stulz Pricing Model The Johnson-Stulz method is useful for valuing "vulnerable" options, that is, options that face some risk of default. The method relies on various assumptions similar to those required by the Black-Scholes methodology, which is commonly used to price exchange-traded and other defaultfree options. Both models assume that the price of the underlying asset (oil in the Mexican example) moves randomly and smoothly through time and that the percent change over any instant of time is independent of the change at any other point in 10 time. 1 Therefore, the intrinsic value of the optionthat is, the difference between the strike price and the underlying asset price—also moves smoothly. Black and Scholes showed that a portfolio consisting of the underlying asset and a short position in bonds could exactly replicate the option. Because the replicating portfolio and actual option offer the same payoff at expiration, theirvalues must be the same any time before expiration. Otherwise, arbitrage profits could be realized. The BlackScholes methodology essentially prices the op- E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 tion by adding up the observable prices of the assets in the replicating portfolio. The Johnson-Stulz method differs from the Black-Scholes method in that it assumes the option writer has limited assets available to meet the potential liability of the option. Mexico, for example, has a limited pool of funds with which to cover its option obligations upon exercise. If the pool value falls below the option obligation, the option owners do not receive the full option payoff upon expiration. 2 Instead, they receive the available funds in the pool. In other words, at expiration the vulnerable call pays the minimum of the option's intrinsic value or its share of the pool (the total pool divided by the number of options issued). In pricing the option used in the example, the pool was valued in the following manner: for a given period, Mexico's current account surplus, which was sometimes negative, was added to the previous period's foreign currency reserves, and required debt service was subtracted. 3 This number, however, was often negative since Mexico partially financed large trade deficits and debt service by obtaining new loans. To apply the JohnsonStulz method, it was necessary to adjust the pool, normalizing it to equal zero at the level at which default would b e likely. The unadjusted pool reached its historical minimum in the third quarter of 1981; the next year Mexico suspended interest payments. Since that time the total quantity of foreign debt has risen. Moreover, both the ratio of debt service to GNP and the ratio of external liabilities to GNP have similarly been much higher in several of the years since the time of the payments moratorium.4 Thus, the rapid rate of capital outflow, rather than the excessive level of debt amassed, apparently was the main factor prompting Mexico's interest suspension in 1982. To normalize the pool, therefore, it was assumed that Mexico would make payments on the options as long as the pool value remained above its historical minimum. If that minimum were ever again achieved, it would lead to default. To normalize the pool in this way, the historical minimum ( - $ 5 . 4 billion) was subtracted from each entry in the time series of pool values. 5 Finally, the pool per option was computed by dividing the total pool value by the number of options issued. Table I presents Johnson-Stulz prices for Mexican oil options using various combinations of strike prices and maturities. The computation uses data from the second quarter of 1989. The oil price was $17.25 per barrel at that time. The total value of the pool was $5.1 billion, or $10.77 per option if options are issued on one full year's worth of exports, 470 million barrels. Parameter inputs were estimated using quarterly data from 1982 to 1989. The standard deviation of percent changes in the pool value is 43 percent per year. The volatility of the oil price is 28 percent per year. The prices in Table 1 may overstate the true value of these options for the following reason: the oil price may not follow a random walk as the Johnson-Stulz, like the Black-Scholes, model assumes. Oil prices may revert to some long-run level, which may vary over time. If this were true, the oil prices would not be as variable overtime as the models predict, and so the computed prices would b e too high. However, there is no firm empirical evidence that oil prices are meanreverting. Nevertheless, the Johnson-Stulz prices, though possibly high, are closer to the true value of the default-risky options than the Black-Scholes prices. Further modeling research should achieve more reliable values for these options. The Johnson-Stulz methodology takes into account the relationship between the value of the pool and the underlying commodity. When oil prices are high, the funds available to Mexico to service outstanding debt and pay off options are more plentiful. The correlation coefficient between these two variables is 0.5, estimated over the sample period running from 1982 to 1989. The higher this correlation, the lower the option default risk will be. A high positive correlation would be expected for countries like Mexico whose economies depend largely on single export commodities. Thus, high strike price covered call options are attractive instruments for these countries, since the relationship between commodity price and the country's financial resources mitigates default. Notes 'More precisely, the asset price is assumed t o follow a diffusion process (a continuous-time geometric Brownian motion). This assumption is standard in o p t i o n pricing models. For t h e sake of greater realism, it would be desirable to allow discrete jumps in the process, which, for example, might be due to oil price I7 FEDERAL RESERVE BANK O F ATLANTA shocks. However, such a model poses serious technical problems in valuing vulnerable options. These details will be addressed in future research. 2 Since Mexico would own the oil on which the option was written, however, a default on the options would entail selling the oil to a third party on the spot market rather than selling to the option owner at the strike price. 3 Data were provided by the International Monetary Fund and Banco d e Mexico. 4 The Economist Intelligence Unit (19891, 35. 5 Since this computation involved subtracting a negative number, it in effect a d d e d to the pool. This recognizes the historical record that the unadjusted pool could become negative without resulting in a default. Notes 'The type of option proposed here is the "European" option, which cannot be exercised until maturity. 2 E. Schwartz (1982) cited two instances where commoditylinked bonds—which grant the lender the option to take a given quantity of a commodity instead of the principal at maturity—had been issued. O'Hara (forthcoming) described how commodity-linked bonds could be used to shift c o m m o d i t y price risk from LDCs to risk-neutral banks. Many other examples exist. The "value recovery clause" in the recent Mexican d e b t package allows banks to receive additional payments starting in 1996 should the price of oil b e above $14 per barrel in constant 1989 dollars by that time. The additional payments will be subject to a cap proportional to the amount of old loans each bank tenders in exchange for the new fixed-interest bonds. ^The World Bank (1989a), 2. 4 LIBOR is the rate of interest on large loans between creditworthy international banks. It is commonly used as the base rate for floating-rate international loans, much in the manner that the prime rate is often used as the base for floating rates on loans in the United States. ^The World Bank (1988), xi. ^See Schuker (1988), 134. 7 lbid. d e v e l o p i n g countries amassed current account surpluses totalling $11.9 billion (U.S.) over the five years from 1971 through 1975. Over the period 1981-86, LDCs amassed deficits totaling $242.7 b i l l i o n (International Monetary Fund, 1987 Yearbook, 136). 9 See Schuker (1988), 134. l0 Latin America's per capita output fell 7 percent in the 1980s, whereas it had grown 40 percent in the 1970s. See Farnsworth (1990). 1 'See "LDC Debt News" (1990), 12. l2 See Sachs (1989b), 92, and DiLeo and Remolona (1989). 13 In practice, it would not be necessary to issue two options; one contingent claim contract can be written with the same features. '"See "Country Risk-Watch," (1989/90), 94-95. ,5 The World Bank (1989b), 254. l6 See Banco d e Mexico. l7 World Oil Price Table, Weekly Petroleum Status Report. l8 For a description of the Black-Scholes model, see Black and Scholes (1973). 19 This is the sum of the val ues of five series of options covering export production for the period between three and seven years into the future. Three-year options would cover the oil to be produced three years from now, fouryear options would cover production four years from now, and so forth. References Banco d e Mexico. "Indicadores del Sector Externo." Indicadores Economicos, updated monthly. Black, Fischer, and Myron Scholes. "The Pricing of Options and Corporate Liabilities." journal of Political Economy 81 (1973): 637-59. Castañeda, )orgeC. "Mexico's Dismal Debt Deal." New York Times, February 25, 1990. "Country Risk-Watch " The International Economy 3 (December 1989/)anuary 1990): 94-95. DiLeo, Paul, and Eli M. Remolona. "On Voluntary Conversions of LDC Debt." Federal Reserve Bank of New York, unpublished manuscript, July 24, 1989. Dornbusch, Rudiger. "International Debt and Economic Instability." Federal Reserve Bank of Kansas City Economic Review (January 1987): 15-32. 12 Durr, Barbara. "Debt, Inflation Continue to Hamper Latin America." Financial Times, December 22, 1989. The Economist Intelligence Unit. Mexico: Country Profile, 1989-90. Annual Survey of Political and Economic Background. London: The Economist Intelligence United Limited, 1989. Farnsworth, Clyde. "U.S. Falls Short on Its Debt Plan for Third World." New York Times, January 9, 1990. International Monetary Fund. International Financial Statistics. Washington, D.C.: IMF, various issues. Johnson, Herb, and René Stulz. "The Pricing of Options with Default Risk." Journal of Finance 42 (June 1987): 267-80. "LDC Debt News." American Banker, March 20, 1990, 12. O'Hara, Maureen. "Financial Contracts and International Lending." Journal of Banking and Finance (forthcoming). E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 Perasso, Giancarlo. "The Pricing of LDC Debt in the Secondary Market: An Empirical Analysis." Kyklos 42, fasc. 4 (1989): 533-55. Sachs, leffrey. "New Approaches to the Latin American Debt Crisis." Essays in International Finance, no. 174. International Finance Section, Department of Economics, Princeton University. Princeton, New Jersey, July 1989a. . "Making the Brady Plan Work." Foreign Affairs (September 1989b): 87-104. Schuker, Stephen A. American "Reparations" to Germany, 1919-33: Implications for the Third-World Debt Crisis. Princeton Studies in International Finance, no. 61. International Finance Section, Department of Economics, Princeton University. Princeton, New Jersey, July 1988. Schwartz, Eduardo. "The Pricing of C o m m o d i t y Linked Bonds." Journal of Finance 37, no. 2 (1982): 525-39. Truell, Peter. "Mexico, Creditor Banks Complete Pact Covering $48 Billion of Debt." Wall Street Journal, January 11, 1990. The World Bank. World Debt Tables, 1987-88: External Debt of Devloping Countries. Vol. I, Analysis and Summary Tables. Washington, D.C., 1988. . World Debt Tables, 1989-90: External Debt of Developing Countries. Vol. I, Analysis and Summary Tables. Washington, D.C., 1989a. . World Debt Tables, 1989-90: External Debt of Developing Countries. Vol. 2, Country Tables. Washington, DC., 1989b. Schwartz, Anna J. "International Debts: What's Fact and What's Fiction." Economic Inquiry 27 (January 1989): 1-19. I7 FEDERAL RESERVE BANK O F ATLANTA Determinants of De Novo Bank Performance William C. Hunter and Aruna Srinivasan This study expands the focus of previous research on the performance of newly chartered banks. The authors' methodology not only identifies the determinants of financial success hut also ranks these factors in terms of their relative importance. The research sample consists of 169 independent de novo banks chartered in 1980 and still operating as independent organizations in 1988. By gauging new banks' performance as compared with that of similar, established banks in the same state of charter, this study avoids possible biases due to differences in regional economies. dentifying the characteristics that make a financial institution successful has b e c o m e a significant consideration, e s p e cially given the failure of record n u m b e r s of banks and savings and loan institutions during the past d e c a d e . While much attention has b e e n directed toward pinpointing the causes of financial distress in e s t a b l i s h e d institutions, analysts are focusing increasingly on determining the factors that lead to success in newly chartered financial institutions. By understanding the e l e m e n t s that contribute to soundness in d e novo banks, regulators and management, as I The authors are, respectively, a research officer and an economist in the financial section of the research department at the Federal Reserve Bank of Atlanta. They thank Rob McDonough and Willy Wolfe for excellent research assistance. 14 well as depositors and other investors, may b e a b l e to identify problems in time to react judiciously. Bank chartering agencies, too, s e e k a better understanding of the economics of startup banking operations in today's less regulated environment. This knowledge would aid in prescribing appropriate criteria for the charter application review and approval process. This article investigates the financial performance of independent d e novo banks chartered in 1980 and still surviving in 1988. The study compares earnings of t h e s e institutions, which were drawn from a national sample, in their first, third, fifth, and eighth years of operation with those of similarly situated, established banks during the s a m e calendar years. A d e novo bank is considered financially successful if its earnings, measured by the return on assets (ROA), are at least 80 percent of the m e d ian ROA for established banks with less than $100 milE C O N O M I C R E V I E W , M A R C H / A P R I L 1990 lion in a s s e t s in t h e s a m p l e bank's s t a t e of charter. This study develops a statistical model of the surviving banks' financial performance in 1981, 1983, 1985, and 1988 to identify and analyze several factors that are crucial in determining the probability of financial success. 1 This analysis reveals that only a small s u b s e t of the factors typically cited as key influences on d e novo banks' survival plays a systematic role during the start-up phase of operations. Specifically, the study identifies the primary determinants of newly chartered banks' financial s u c c e s s as: (1) the overall quality of the bank's credit policies, exemplified by its procedures for credit evaluation and approval as well as loan monitoring and collection; (2) the bank's ability to control operating costs such as wages and salaries; and (3) the bank's level of capitalization at startup. Factors such as market structure and economic conditions in the bank's local trade area, while important in a generic sense, do not exert much impact on the probability of financial succ e s s when considered independently. A Previous Study of De Novo Bank Performance Though a number of studies have examined the performance of d e novo commercial banks, research by Nasser Arshadi and Edward C. Lawrence (1987) is most relevant to the analysis presented in this article. 2 Arshadi and Lawrence studied the financial performance of 438 banks chartered between 1977 and 1979 in their third and fifth years of operation, using canonical correlation analysis, a statistical technique useful for determining the maximum correlation b e tween two sets of variables. T h e two researchers examined variables both internal and external to banks' decision making in relation to several gauges of performance: the return on assets, the ratio of interest and f e e s on loans to total loans, and the bank's share of the total loans in the primary trade area. Endogenous factors included indicators of b a n k c o s t such as t h e ratio of salaries and wages to total assets, loan portfolio composition (the proportion of the portfolio invested in consumer, real estate, commercial, and industrial loans), the return on the loan portfolio, and deposit composition. Among the exogenous factors studied were indicators of market structure, such as the number of competing banks, as well as economic and demographic conditions in the primary trade area—for example, growth in effective buying income and population per banking office. Based on their empirical analysis, Arshadi and Lawrence concluded that the performance of newly chartered banks d e p e n d s most critically on endogenous factors such as cost structure, a s s e t size, and t h e c o m p o s i t i o n of the loan portfolio. A notable finding of their study was the relatively minor influence on performance exerted by market structure and economic conditions in the bank's trade area. While Arshadi and Lawrence's results are interesting, the limitations of their methodology, canonical correlation analysis, precluded explicit calculation of the marginal value or impact of bank-specific and market characteristics on a d e novo bank's chances of financial success. Their analysis therefore could not rank the factors in terms of their relative importance. In addition, b e c a u s e Arshadi and Lawrence analyzed s a m p l e banks' a b s o l u t e performance rather than performance relative to similarly situated banks, regional factors such as t h e state of the oil industry in the Southwest or the farm economy in the Midwest may have biased their conclusions. Research Methodology and Variable Definitions In contrast to Arshadi and Lawrence's study, the explicit goals of the analysis in this article are to characterize the marginal impact of bankspecific attributes on the probability that a d e novo bank will b e financially successful and to determine whether exogenous market and regulatory factors play a significant role in determining success. By paying particular attention to the financial performance of each sample bank relative to the performance of similar banks in the sample bank's state of charter, this study avoids possible biases due to differences in regional economies. The Performance Measure and Statistical Model. For the purposes of this study, a bank I7 F E D E R A L RESERVE B A N K O F ATLANTA was d e e m e d financially successful if its return on assets (ROA) was equal to or greater than 80 percent of the median ROA for banks in the state of charter with total assets less than $100 million. Similarly, a bank was identified as unsuccessful if its ROA was less than 80 percent of the median ROA for banks in the comparison group. A binary (indicator) variable was assigned for each sample bank and was s e t equal to a value of one if the bank was classified as financially successful and zero if the bank was considered unsuccessful. The binary variable, which d e n o t e s the occurrence or nonoccurrence of the event "financial success," was developed to allow the use of probit analysis, a technique that estimates the impact of observable characteristics on the probability of an event's occurring. For a detailed description of the probit technique s e e the box on page 23. 3 In this study, the probit technique permitted the determination of bank-specific and market or regulatory factors' marginal impact on the probability that a bank's financial performance will b e classified as successful. 4 For each year of analysis, the following probit model describing financial success was estimated: Sj = a + fix + ei, where S(- is the performance indicator variable for bank /', /3 is a vector of parameters to b e estimated, X is a vector of bank-specific attrib u t e s and exogenous market and regulatory factors, and e,- is an error term. The model was estimated for 1981, 1983, 1985, and 1988, when the sample banks were one, three, five, and eight years old, respectively. T h e s e years are consistent with bank chartering agencies' practice of examining d e novo banks within three to five years after granting their charters. study is that the probability of financial success will b e inversely related to the variables SALARIES (the ratio of wage and salary e x p e n s e s to total assets) and LTIMEDEP (the ratio of large time deposits to total deposits). Holding other factors constant, increases in personnel exp e n s e s imply lower bank profitability and a reduced ROA. Likewise, the larger the proportion of deposits on which a bank pays interest rates that are highly competitive, the lower its profit margin should be. On the other hand, ceteris paribus gains in the variable NOWACC (the ratio of NOW accounts and other savings accounts subject to interest rate ceilings to total deposits) would b e expected to increase the probability of financial success since this serves as a proxy for the bank's level of core deposits (that is, t h e bank's lowest-cost, most s t a b l e deposits). Lower funding costs should improve the bank's profitability, raising the ROA. The logic of the relationship between a bank's cushion of capital and its financial performance is also essentially straightforward. Typically, the better capitalized a d e novo commercial bank is, the better its chances for survival. Equity capital acts as a buffer against the unforeseen losses incurred by all banks; it is critical to newly chartered banks' probability of survival. This study assumes that if all other factors remain constant, increases in a bank's ratio of total capital to total assets (CAPITAL) will increase its probability of financial success. 7 Variable Definitions. The specific bank attributes and exogenous factors included as components of t h e X v e c t o r in the probit model were s e l e c t e d to represent the following eight areas: bank operating cost structure, leverage, loan portfolio composition, credit policy, liquidity, local market structure, state branching law, and the economic climate of the local trade area. 5 The specific variables included in each category are defined in Table l. 6 T h e relationship between the probability of financial s u c c e s s and the composition of a bank's loan portfolio is harder to characterize. Ordinarily, the more diversified its loan portfolio, the more likely the bank is to avoid significant l o s s e s resulting from concentrating its lending in an area or industry which unforeseen e v e n t s may adversely affect. N o n e t h e l e s s , a bank may concentrate its lending in few sectors and remain quite profitable if it has a d e q u a t e credit evaluation, monitoring, and collection policies. Many banks' credit policy standards also specify the types of loans to b e made and their proportions of the total portfolio. In t h e s e c a s e s the relationship between loan portfolio composition and the probability of success dep e n d s critically on the quality of t h e bank's credit policies. With respect to the variables measuring bank cost structure, the a priori expectation in this A bank's credit policy is generally considered a key determinant of its long-run survival prob- 16 E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 Table 1. Variable Definitions Performance Measure: = 1, if ROA > 8 0 percent of median ROA of target group 3 0, otherwise S Cost Structure: SALARIES LTIMEDEP NOWACC = ratio of wage and salary expenses to total assets = ratio of large time deposits to total deposits = ratio of NOW accounts and other savings accounts subject to interest rate ceilings to total deposits Loan Portfolio Composition: REALEST CONS C&l = ratio of real estate loans to total loans = ratio of consumer loans to total loans = ratio of commerical and industrial loans to total loans Credit Policy: LOANLOSS = ratio of charge-offs net of recoveries to total loans Liquidity: FEDFUNDS = ratio of net federal funds purchased to total deposits LOANDEP = ratio of total loans to total deposits Leverage: CAPITAL = ratio of total equity capital to total assets Market Structure: MSA = 1, if the bank is chartered in a metropolitan statistical area (MSA) 0, otherwise = concentration ratio of three largest firms in the local market 3FIRMCR Regulatory Structure: STATEBR = 1, 0, = 1, 0, LTDBR if bank is chartered in statewide branching state otherwise if bank is chartered in limited branching state otherwise Economic Conditions: CHGPI a The target group is composed = change in personal income growth in the local market of all banks in the state of charter with total assets less than $100 ability. As n o t e d a b o v e , t h r e e primary comp o n e n t s of a bank's credit policy are its credit evaluation p r o c e d u r e s and standards, monitoring policy, and collection procedures. T h e varia b l e LOANLOSS r e p r e s e n t s t h e effectiveness of t h e credit policy. Clearly, this policy will control t h e quality of t h e loan portfolio, which, in turn, will d e t e r m i n e t h e bank's net l o s s e s and, hence, its e c o n o m i c viability. Thus, in t h e a b s e n c e of other changes, increases in t h e bank's ratio of net charge-offs to total loans should d e c r e a s e profitability a n d t h e p r o b a b i l i t y of financial success. Making e x a n t e predictions a b o u t t h e impact of t h e variables measuring s a m p l e b a n k liquidFEDERAL RESERVE B A N K O F ATLANTA million. ity ( F E D F U N D S a n d LOANDEP) on a b a n k ' s probability of s u c c e s s is problematic without d e t a i l e d knowledge of t h e bank's credit policy and its loan d e m a n d . T h e s a m p l e banks are net s e l l e r s of federal funds in all years. This transaction is essentially a short-term loan by o n e b a n k to another, which is borrowing to m e e t its reserve r e q u i r e m e n t s or loan d e m a n d . A d e c r e a s e in t h e FEDFUNDS variable (the variable b e c o m e s m o r e negative) could b e taken a s a signal that t h e b a n k has additional temporary e x c e s s liquidity and is investing in this market short term ; a d e c r e a s e could b e interpreted to mean, however, that t h e bank is refusing to m a k e additional loans or hold m o r e securities. Although I 7 t h e latter explanations might suggest that t h e b a n k is being t o o conservative or turning down legitimate loan r e q u e s t s , such a conclusion is not warranted without an understanding of t h e bank's loan policies and loan d e m a n d . In t h e s a m e m a n n e r c e t e r i s p a r i b u s i n c r e a s e s in a bank's l o a n - t o - d e p o s i t ratio, LOANDEP, could e n h a n c e or detract from t h e bank's profitability under various loan-quality scenarios. Like portfolio composition, s o m e environmental variables could influence financial performance in o p p o s i t e ways. For example, it is difficult to specify t h e ex a n t e effect of a d e novo bank's l o c a t i o n — r e p e s e n t e d by t h e MSA (metropolitan statistical area) variable—on t h e probability of success. Such factors as t h e bank's credit policy and t h e nature of c o m p e t i t i o n in t h e bank's primary area will greatly affect t h e MSA variable's impact. I n d e p e n d e n t of its location, and holding other factors equal, a newly chartered b a n k will b e l e s s likely to p r o s p e r during t h e start-up p h a s e t h e m o r e c o n c e n t r a t e d t h e primary trade area or local market is (the higher t h e variable 3F1RMCR). 8 To t h e e x t e n t that t h e m a r k e t is d o m i n a t e d by b a n k s with strong c u s t o m e r relationships—and, h e n c e , significant market shares—the new b a n k will b e forced to c o m p e t e on t h e b a s i s of rates on loans and d e p o s i t s a s well as other services whose provision entails additional operating costs without any guarant e e of increased revenues. Alternatively, b a n k s in c o n c e n t r a t e d markets may earn high profits without having to offer competitive prices. T h e s e profits should provide a protective umbrella for new entrants. S t a t e branching laws can play an important role in determining t h e long-run financial succ e s s of newly chartered banks. Although branching is a proven m e t h o d by which banks can grow and diversify their retail d e p o s i t and loan portfolios, b r a n c h i n g a l s o allows c o m p e t i t o r s t o e n t e r a bank's trade area. Thus, relative to t h e c a s e of unit banking laws, t h e effect of statewide and limited branching laws on t h e probability of financial s u c c e s s could b e either positive or negative. General e c o n o m i c conditions in t h e s a m p l e banks' primary trade areas are m e a s u r e d by t h e variable CHGPI, t h e c h a n g e in personal income growth in t h e local market. Clearly, ceteris parib u s increases in this variable should improve 18 d e novo b a n k s ' p r o b a b i l i t y of financial success.9 D a t a a n d S a m p l e Selection. T h e full s a m p l e consists of 169 i n d e p e n d e n t b a n k s chartered in 1980 in 32 states. New i n d e p e n d e n t banks warrant distinct scrutiny b e c a u s e r e c e n t e v i d e n c e (John R o s e and Donald Savage 1984) indicates that their performance differs from that of new b a n k s f o r m e d a s s u b s i d i a r i e s of m u l t i b a n k holding c o m p a n i e s . As did earlier studies, this research excludes industrial banks, trust c o m panies, " p h a n t o m " banks organized to facilitate acquisition of an ongoing bank, and b a n k s organized by foreign banking organizations. T h e year 1980 was s e l e c t e d a s t h e starting point for two reasons. First, t h e Office of t h e Comptroller of t h e Currency (OCC), t h e principal regulator of national banks, liberalized its policy for granting national b a n k charters in that year. T h e OCC also a d o p t e d a new weighting s c h e m e which p l a c e d greater e m p h a s i s on t h e agency's " a p p r a i s a l of t h e organizing g r o u p a n d its operating plan for establishing and operating a b a n k " and less e m p h a s i s on " t h e community's e c o n o m i c and c o m p e t i t i v e characteristics." T h e revised policy reflected t h e OCC's e x p e r i e n c e that " a strong organizing group . . . |was] generally a b l e t o establish and o p e r a t e a successful b a n k even in t h e most economically d i s t r e s s e d or m o s t highly c o m p e t i t i v e markets." S e c o n d , using 1980 a s t h e b a s e year p e r m i t t e d t h e examination of d e novo banks during a period in which b a n k regulation was b e i n g liberalized. This focus is especially significant since little is known a b o u t t h e effect of interest-rate, product, and geographic market deregulation on new b a n k performance. Of t h e s a m p l e b a n k s o p e n e d in 1980, 67 perc e n t were located in metropolitan areas (MSAs), and 123 of t h e 169 b a n k s had s t a t e charters. By t h e e n d of 1988, only 96 of t h e original 169 banks were still operating independently. Seventyt h r e e b a n k s had either failed or m e r g e d with other financial institutions, including multibank holding c o m p a n i e s . T h e financial and regulatory data on each s a m p l e bank were taken from t h e Reports of Condition and Reports of Income filed annually with t h e Federal Reserve System. T h e data on personal income were gathered from various issues of t h e U.S. C o m m e r c e Department's Survey of Current Business.. E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 Table 2. Means of Variables for Different Years 1981 Variable 3 (Age 1 ) 1983 (Age 3) 1985 (Age 5) 1988 (Age 8) SALARIES 0.02 0.02 0.02 0.02 LTIMEDEP 0.31 0.24 0.21 0.17 NOWACC 0.06 0.05 NA NA 0.35 0.47 REALEST 0.22 0.29 CONS 0.30 0.28 0.25 0.21 C&l 0.41 0.38 0.34 0.28 LOANLOSS 0.00 0.01 0.02 0.02 -0.07 -0.08 -0.10 -0.04 LOANDEP 0.66 0.66 0.67 0.65 CAPITAL 0.15 0.10 0.07 0.07 MEAN ROA (Percent) 1.06 0.12 -0.44 0.37 MEDIAN ROA (Percent) 1.28 0.48 0.39 0.70 MEAN ROE (Percent) 6.37 0.42 -11.81 1.43 MEDIAN ROE (Percent) 7.53 4.38 4.45 9.28 165 141 123 96 FEDFUNDS Number of Surviving Banks a See Table 1 for variable NA = Not applicable definitions. in the indicated years. The Empirical Findings The mean values of each of the continuous variables included in the analysis are given in Table 2 for 1981, the first full year of operation for the sample banks, and for 1983, 1985, and 1988, along with the median and mean ROA of the surviving banks. T h e mean ROA of the s a m p l e banks was a very favorable 1.06 percent for 1981, the first full year of operation. However, the ratio returned to more normal levels (less than 1 percent) as the surviving sample banks b e c a m e more established. Table 3 profiles the performance of the sample banks by age, showing the number and perc e n t a g e of t h e surviving s a m p l e b a n k s with ROAs greater than zero, greater than or equal to 80 percent of the median ROA of established banks with total assets not exceeding $100 million in the relevant state of charter, and greater than or equal to 150 percent of the median ROA of the established banks. Since the majority of the sample banks were chartered in late 1980, the zero age category typically represents operating performance over only one or two months. About 86 percent of the sample banks surviving for at least one year posted positive net income in the first year of operation ; about 26 percent of the sample banks at that age demonstrated earnings superior to t h o s e of their established counterparts—that is, their ROA indices were 150 percent or more of the median of their more established counterparts. 1 0 I7 FEDERAL RESERVE BANK O F ATLANTA Table 3. De Novo Bank Earnings Performance by Age, 1980-88 Break-Even Analysis Age Number of Banks Earnings Comparability with Established Banks Percent 3 Number of Banks Percent b Earnings Superiority of De Novo Banks Number of Banks 0 76 45.0 20 11.8 3 1.8 1 142 86.1 96 58.2 42 25.5 3 99 70.2 54 38.3 22 15.6 5 80 65.0 54 43.9 20 16.3 8 73 76.0 49 51.0 18 18.8 a Percent of de novo banks at given age with ROA > 0. b Percent of de novo banks at given age with ROA > 80 percent of the median ROA for established charter with up to $100 million in assets. 0 Percent of de novo banks at given age with ROA >_ 150 percent of the median ROA of established The results of the probit model for the four time periods studied are presented in Table 4, along with goodness-of-fit measures. 1 1 The estimated probit model performs reasonably well for each year. T h e e s t i m a t e d probability of achieving the observed value (0 or 1 ) e x c e e d s 0.5 in 70 percent or more of the total c a s e s in all years. The likelihood ratio, indicating how well the s e l e c t e d model specification explains financial performance, is also significant in all c a s e s at the 0.01 level. Among the individual variables, CAPITAL and SALARIES have the expected signs with statistically significant coefficients in all four models. T h e LOANLOSS variable is significant with the correct sign in all years except 1988. The proportion of total loans in consumer (CONS) and business loans (C&I) is significant in the first year of operation, while the share of real estate lending in total loans (REALEST) is positive and significant in the later years of operation. The only exogenous variable of any significance was STATEBR (statewide branching) in all years except the fifth. The probit analysis results indicate that differences in operating costs, credit policy, and leverage account for most of the performance variations among the sample banks relative to the established target group during the early years of operation. Other variables—namely, those measuring the composition of the bank's loan portfolio (REALEST, CONS, and C&I), li20 Percent 0 banks in state of banks. quidity (FEDFUNDS and LOANDEP), market structure (MSA and 3FIRMCR), and local economic conditions (CHGPI)—do not exhibit consistent, significant effects on the performance of new banks studied during the first eight years of operation. T h e positive and significant coefficient on the state branching laws variable (STATEBR) in 1981, 1983, and 1988 suggests that being chartered in a state that allows statewide branching increases new banks' probability of success. Statewide branching also allows established banks to s e t up d e novo branchés, thereby increasing the competition i n d e p e n d e n t d e novos face. Under these circumstances, the evidence that statewide branching enhances the financial performance of new independent banks corroborates the arguments of Constance Dunham ( 1989) and others that consumers prefer small independent banks over large, multioffice institutions. 12 The insignificant coefficients on the NOWACC variable (the ratio of NOW accounts and other savings deposits subject to interest rate ceilings to total deposits) and LTIMEDEP (the ratios of large time deposits to total deposits), except for 1981, are especially interesting. The Depository Institutions Deregulation and Monetary Control Act of 1980, which phased out ceilings on deposit rates over a six-year period, was expected to increase competition in financial services. In this context, the low profile of LTIMEDEP is surE C O N O M I C R E V I E W , M A R C H / A P R I L 1990 Table 4. Results of Probit Model for Different Years of Operation 1981 (Age 1) Variable INTERCEPT 1983 (Age 3) -0.09 -1.61 SALARIES -93.70* -58.64* LTIMEDEP -1.52* NOWACC REALEST 1985 (Age 5) 1988 (Age 8) -5.06* -5.79* -55.50* -49.65* -1.09 1.33 -0.39 2.62 NA NA 0.77 -0.15 4.36* 4.57* -2.21 CONS 1.80* 0.17 3.64 3.13 C&l 2.46* -0.80 2.08 4.30* LOANLOSS -39.35* FEDFUNDS -105.09* -159.33* 2.97 0.16 -0.24 -2.21 LOANDEP 0.63 1.57 2.30 0.56 CAPITAL 4.29* 9.50* 34.00* 24.15* 0.07 0.61 -0.31 -0.14 -0.37 1.64 -0.17 -0.17 MSA 3FIRMCR STATEBR 0.59* 0.66* 0.15 0.34 1.50* LTD BR -0.01 -0.09 -0.23 CHGPI 1.58 -4.79 1.29 2.29* 0.67 Goodness-of-Fit Measures Number of Right Predictions 118 102 108 74 Percent of Right Predictions 72 72 88 77 Likelihood Ratio Test 36.5 55.9 94.5 34.4 Degrees of Freedom 15 15 14 14 * Statistically significant at the 1 percent * Statistically significant at the 5 percent * Statistically significant at the 10 percent NA = Not applicable in the indicated level. level. level. years. prising, suggesting that d e novo banks did not use the relaxation of interest rate ceilings to spur growth by bidding for additional deposits. (In fact, the ratio of large time deposits to total deposits fell by nearly one-half at the sample banks from 1980 to 1988). The insignificance of NOWACC is partially attributable to the fact that, on average, NOW accounts and other savings deposits made up less than 10 percent of sample banks' total deposits in 1981 and 1983. T h e coefficients on t h e liquidity variables (FEDFUNDS and LOANDEP) were insignificant for the most part. T h e sample banks were net sellers of federal funds in all years, and their FEDERAL RESERVE B A N K O F ATLANTA loan-to-deposit ratio held constant. Contrary to Arshadi and Lawrence's results, this study indicates that market structure variables had no significant impact on t h e new banks' probability of success. As noted earlier, a major advantage of using probit analysis is that the marginal impacts of the independent variables on a new bank's probability of s u c c e s s can b e a s s e s s e d . T a b l e 5 shows the marginal effects computed at the means of the regressors. Operating costs and credit policy dominate new bank performance in the first year of operation. For example, a ceteris paribus increase of 1 percent in the ratio I7 Table 5. Marginal Effects of Change in Independent Variables On Probability of Success of De Novo Banks 1985 (Age 5) 1988 (Age 8) Variable 1981 (Age 1 ) 1983 (Age 3) INTERCEPT -0.03 -0.46 -0.19 -2.30 -19.75 -0.88 SALARIES -36.50 -16.89 -2.07 LTIMEDEP -0.59 -0.31 0.05 NOWACC -0.15 0.75 NA NA REALEST 0.30 -0.04 0.16 1.82 CONS 0.70 0.05 0.14 1.25 0.96 -0.23 0.08 1.71 -15.33 -30.26 -5.93 1.18 FEDFUNDS 0.06 -0.07 -0.08 0.13 LOANDEP 0.25 0.45 0.09 0.22 C&l LOANLOSS CAPITAL 1.67 2.73 1.27 9.61 MSA 0.03 0.18 -0.01 -0.06 -0.14 0.47 -0.01 -0.07 0.23 0.19 0.01 0.59 3FIRMCR STATEBR LTDBR -0.00 0.03 -0.01 0.27 CHGPI 0.61 -1.38 0.05 0.91 NA = Not applicable in the indicated years. of wages and salaries to total assets (SALARIES) reduces the probability of financial success of new banks by 36.5 percent. An increase of similar magnitude in the ratio of loan losses net of charge-offs to total loans (LOANLOSS) lowers the probability of success by 15 percent. Although t h e relative importance of operating costs and credit policy (as measured by their marginal impacts) declines in the four years examined, t h e s e two factors clearly dominate in explaining overall d e novo bank success. The marginal impact of CAPITAL is strongest in the eighth year of operation. The remaining statistically significant i n d e p e n d e n t variables showed small marginal effects. significant in most years examined. T h e s e findings are consistent with those of s o m e earlier studies and emphasize the appropriateness of the OCC's stress on managerial c o m p e t e n c e in evaluating charter proposals. T h e results do not imply that factors exogenous to managerial decision making are unimportant. Nonetheless, the findings clearly show that b a n k survival d e p e n d s largely on factors directly under the bank management's control. This study points to the importance of factors endogenous to the decision-making process at i n d e p e n d e n t d e novo banks in determining their financial performance. Controlling operating costs and loan portfolio quality clearly are critical in the early stages. Two other factors—a bank's equity capitalization and its location in a state that allows statewide branching—are also This article has identified the factors, both bank-specific and exogenous, most critical in determining the probability of financial success in independent d e novo banks. Three bankspecific (internal) characteristics appear to b e the primary determinants of newly chartered banks' likelihood of achieving financial success: credit policy, measured by the bank's ratio of 22 Conclusions E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 net loan l o s s e s to total a s s e t s ; operating costs, indicated by t h e ratio of wages and salaries to total a s s e t s ; and t h e level of equity capitalization. Other factors, such a s loan portfolio c o m position and s t a t e branching laws, were found to b e l e s s significant. Variables often c o n s i d e r e d i m p o r t a n t d e t e r m i n a n t s of d e novo banks' success—such a s market concentration and local e c o n o m i c conditions—evidence no consistent, significant effect in this analysis. While t h e s e findings are generally c o n s i s t e n t with previous research, two characteristics distinguish t h e s e conclusions. First, t h e u s e of t h e probit analysis t e c h n i q u e allowed various b a n k and market factors to b e ranked according to their relative importance in predicting financial s u c c e s s . S e c o n d , by c o m p a r i n g t h e financial performance of s a m p l e banks with similar, es- t a b l i s h e d banks in t h e s a m p l e banks s t a t e s ' of charter, p o s s i b l e b i a s e s d u e to regional e c o nomic differences were avoided. B e c a u s e this research included only s a m p l e banks surviving for t h e entire eight-year period, attributing its results to all d e novo b a n k s would b e i n a p p r o p r i a t e . Clearly, an a r e a worthy of further study is t h e financial performance of d e novo b a n k s that failed, that were merged with or acquired by other banks, or that m a d e acquisitions t h e m s e l v e s . With t h e proper c h o i c e of methodology, such research can provide e m pirical e v i d e n c e on t h e attributes important in determining t h e duration of newly c h a r t e r e d banks a s i n d e p e n d e n t entities a s well as t h e diff e r e n c e s b e t w e e n banks that fail during t h e start-up p h a s e and t h o s e that fail after b e c o m ing e s t a b l i s h e d . A Description of the Probit Technique To understand how the probit model works in this study, assume that there exists an index S(that measures de novo bank performance relative to established banks of similar size in the state of charter. S ( is chosen so that the higher its value is, the greater the likelihood that the bank will be financially successful. S,- is hypothesized to b e a linear function of bank-specific attributes and market and regulatory factors: Sj = a + pXj. (1) In addition, it is assumed that associated with each bank is a critical value S* of the index S(. such that: if Sj > S*, the bank is financially successful; if Sj < SJ, the bank is financially unsuccessful. (2) The probit model assumes that S* is a normally distributed random variable so that the probability that SJ is less than or equal to Sf- can b e computed from the cumulative normal probability function. The cumulative normal probability function is written as follows: By construction, the variable Pf. will lie in the (0,1) interval. To obtain the variable S(-, which is assumed to be linear in the unknown parameters, ¡3, the inverse of the cumulative normal function is appl ied to equation (3): Sj = F~] (Pj) = a + (iXj. (4) The probability P(- resulting from the probit model may b e interpreted as an estimate of the conditional probability that a bank will be financially successful. It is important to note that the parameters of the model are not necessarily the marginal effects analyzed in the general linear model. In general, - ^ = f { a + pXj)p, (5) where f (•) is the density function corresponding to the cumulative distribution F (•). These derivatives predict the effect of changes in one of the independent variables on the probability of being successful. For the probit model, f (•) is <}>(t), the standard normal density function. I7 FEDERAL RESERVE BANK O F ATLANTA Notes 1 2 Most previous studies of bank profitability determinants have examined one or more of the following issues: economies of scale, market concentration, and financial ratios associated with bank profitability. This study most closely approximates the third line of research as exemplified in studies conducted by Fraser and Rose (1972), Fraser, Phillips, and Rose (1974), Kwast and Rose (1982), Wall (1983), and Nejezchleb (1988). Studies by Alhadeff and Alhadeff (1976), Arshadi and Lawrence (1987), Austin and Binkert (1975), Huyser (1986), Martin and Sauter (1986), Tufts and Struck (1984), and Yeats, Irons, and Rhoades (1975) look at various aspects of new bank financial performance. More recent studies by Dunham (1989) and Rose and Savage (1989) examine new bank formation rates and market share accumulation, respectively. % e e Maddala (1986) and Pindyckand Rubinfeld (1976) for an in-depth look at the probit technique. 4 Note that the probit model is used in this article to classify the sample banks as successful based on observable characteristics. The statistical tests are descriptive and are not derived from a formal model of financial success. 5 Bank asset size is not included as a separate variable since the performance measure, ROA, involves the normalization of net income by asset size, thereby enabling the researchers to control for differences in bank size and the possible effects of economies of scale. The average asset size of the sample banks was $14.8 million in 1981 and $49 million in 1988. 6 Many of these variables have been used as determinants of financial success in previous studies of de novo bank performance. See, for example, Arshadi and Lawrence (1987), Fraser, Phillips, and Rose (1974), Gilbert and Peterson (1975), Gilbert (1984), Kwast and Rose (1982), Rhoades and Savage (1985), Rose and Savage (1984), and Rose (1977, 1988). 7 lt could b e argued that better-capitalized de novo banks are more likely to survive simply because they can finance their assets less expensively than can weakly capitalized de novo banks. While this argument certainly applies to larger banks that rely on purchased funds for asset growth, it does not typically apply to d e novo banks that rarely engage in the practice of liability management. h h e local market or primary trade area is defined as either the MSA or the non-MSA county where the sample bank is located. \ h e poorer the economic conditions in a given area, the fewer should be the number of new bank charter applications in that area. However, one would still expect a positive relationship at the margin between local economic conditions and financial success. l0 In 1981, 69 sample banks were classified as unsuccessful (ROA less than 80 percent of the median benchmark). In 1983, 1985, and 1988 the number of banks classified as unsuccessful was 89, 69, and 47, respectively. " T h e goodness-of-fit measures indicate the accuracy with which a model approximates the observed data (like the R2 in linear regression models). In the case of qualitative dependent variables, accuracy can be judged either in terms of the fit between the calculated probabilities and observed frequencies (percent of right predictions) or the maximum of the likelihood function. 12 Rose and Savage (1989) also conclude that liberal branching privileges for existing banks do not adversely affect market share accumulation by new banks. References Alhadeff, David A., and Charlotte P. Alhadeff. "Growth and Survival Patterns of New Banks, 1948-1970." Journal of Money, Credit, and Banking 8 (May 1976): 199-208. Arshadi, Nasser, and Edward C. Lawrence. "An Empirical Investigation of New Bank Performance." Journal of Banking and Finance 11 (March 1987): 33-48. Austin, Douglas V., and Christopher C. Binkert. "A Performance Analysis of Newly Chartered Banks." Magazine of Bank Administration 51 (January 1975): 34-35. Dunham, Constance R. " N e w Banks in New England." Federal Reserve Bank of Boston New England Economic Review (January/February 1989): 30-41. Fraser, Donald R., R. Wallace Phillips, and PeterS. Rose. "A Canonical Analysis of Bank Performance." Journal of Financial and Quantitative Analysis 9 (March 1974): 287-95. Fraser, Donald R„ and PeterS. Rose. "Bank Entry and Bank Performance." Journal of Finance 27 (March 1972): 65-78. Gilbert, Gary G., and Manferd O. Peterson. "The Impact of Changes in Federal Reserve Membership on Commercial Bank Performance." Journal of Finance 30 ()une 1975): 713-19. 24 Gilbert, R. Alton. "Bank Market Structure and Competition." Journal of Money, Credit, and Banking 16 (November 1984): 617-45. Huyser, Daniel. "De Novo Bank Performance in the Seven Tenth District States." Federal Reserve Bank of Kansas City Banking Studies ( 1986) : 13-22. Kwast, Myron, and lohn T. Rose. "Pricing, Operating Efficiency, and Profitability Among Large Commercial Banks." Journal of Banking and Finance 6 (March 1982): 233-54. Maddala, G.S. Limited-Dependent and Qualitative Variables in Econometrics. New York: Cambridge University Press, 1986. Martin, Larry W„ and Diane R. Sauter. "Analysis of High and Low Performance De Novo Banks." Bank Administration 62 (April 1986): 34, 36, 40. Nejezchleb, Lynn A. "Some Evidence on the Nature of Bank Profitability Problems." Paper presented at the Eastern Economic Association Meetings, Boston, Massachusetts, March 10-12, 1988. Office of the Comptroller of the Currency. News Release. October 2, 1980. E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 Pindyck, Robert S., and Daniel L. Rubinfeld. Econometric Models and Economic Forecasts. New York: McGrawHill, 1976. Rhoades, Stephen A , and Donald T. Savage. "The Viability of the Small Bank." The Bankers Magazine 168 d u l y / August 1985): 66-72. Rose, John T. "The Attractiveness of Banking Markets for De Novo Entry: The Evidence for Texas." Journal of Bank Research 7 (Winter 1977): 285-93. "New Independent Bank Entry in an Era of Financial Deregulation." Review of Research in Banking and Finance 4 (Fall 1988): 45-59. Rose, JohnT„ and Donald T. Savage. " D e Novo Entryand Performance: Bank Holding Companies versus Independent Banks." Journal of Bank Research 15 (Summer 1984): 95-107. FEDERAL RESERVE B A N K O F ATLANTA Rose, lohn T„ and Donald T. Savage. "New Independent Banks and Market Share Accumulation in a Deregulated Environment." Unpublished paper, 1989. Tufts, Roger W., and Peter L. Struck. "The Performance of New Banks Under the Revised OCC Chartering Policy." Office of the Comptroller of the Currency Doc. #0125B, September 17, 1984. Wall, Larry D. "Why Are Some Banks More Profitable? A Statistical Cost Analysis." Federal Reserve Bank of Atlanta Economic Review 68 (October 1983): 44-51. Yeats, Alexander J., Edward D. Irons, and Stephen A. Rhoades. "An Analysis of New Bank Growth." Journal of Business 48 (April 1975): 199-208. I7 Measuring Interstate Migration William ). Kahley t's hard to b e a t warm weather and jobs. An attractive climate and rapid employment growth help explain why states like Florida and Georgia attracted so many new residents in the 1980s that their population grew by about 30 percent and 18 percent, respectively, while the nation's grew only by 11 percent. Measuring state-to-state migration and understanding its causes are important b e c a u s e the effects of t h e s e population flows are significant for both the receiving and sending areas. I mists but also developers, bankers, utility companies, and, perhaps most of all, state and local planners and policymakers—can tap a variety of information sources to monitor and understand the size, direction, and composition of migration. For policymakers, this information is essential to weighing inmigration's c o s t s and b e n e f i t s and formulating appropriate policy responses. Unfortunately, data on migration from state to state are often not available on a timely basis. The Census of Population is the b e s t source of information, but it is available only once every 10 years. T h e 1990 census, now being compiled, will not b e available until 1991 on computer tapes and 1992 or 1993 in print. In addition, the statistical s e r i e s on migration entail various m e a s u r e m e n t p r o b l e m s that users must b e aware of before drawing conclusions and inferences. 1 With proper caution, however, those interested in population movements from state to state—not only demographers and econo- Inmigration can raise incomes, fuel job growth, stimulate demand for housing and other types of construction, and generally promote economic development. 2 A rapid population influx can also result in congestion, pollution, and increased need for public transportation, schools, and other infrastructure investment that creates added burdens for state and local governments. Thus, knowing the age profile of migrants, for example, is useful in determining future capacity n e e d s for schools, hospitals, and retirement homes. Details regarding migrants' assets or income sources, skills, and education are likewise helpful in planning for community support systems. The author is an economist in the regional section of the Atlanta Fed's research department. He thanks Amy Bailey for research assistance. This article reviews and evaluates the strengths and weaknesses of the various interstate migration information sources. It also applies this analysis to southeastern data to demonstrate what available figures reveal about migration in 26 E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 and out of t h e s e six states—Alabama, Florida, Georgia, Louisiana, Mississippi, and Tennessee— during the 1980s. 3 In pointing out the lags, weaknesses, and shortcomings of certain data series, no criticism or j u d g m e n t of t h e r e s p o n s i b l e organization is intended. Availability problems, for e x a m p l e , are virtually unavoidable. T h e United States is a nation of great geographic expanse with no state border controls or national register for citizens. Hence, collecting migration information is expensive. Gathering detailed data at the national level raises costs further, as d o e s compiling figures about migration patterns in particular regions, states, or localities. For these reasons the U.S. government obtains migration data on an infrequent basis and from samples rather than the population as a whole. Federal agencies also collect more detailed information for the nation than for individual states or metropolitan areas. The n e e d to rely on samples introduces certain statistical problems. The magnitudes of sampling and nonsampling errors affect the accuracy of migration estimates. Chance variability in estimating occurs when surveying only a portion of the population; in other words, the estimated number will differ depending on the sample drawn from the surveyed population. Nonsampling errors occur for numerous reasons: r e s p o n s e s may b e incomplete, leaving gaps in information; categories might b e imprecise; p e o p l e may interpret questions in different ways; and s o m e of those surveyed are likely to b e unable or unwilling to respond correctly or to recall all the information sought. Recording and coding the data collected also introduce potential errors, and mistakes can b e made while processing data. Despite t h e s e problems, existing data measurement sources can yield careful users a surprising amount of information about migration at the state level. Researchers can u s e certain statistical methods to adjust migration data for s o m e measurement problems and to understand better what the data indicate. While timeliness and level of detail remain difficult issues, one can trace noteworthy patterns and shifts in migration by examining not only the decennial Census of Population but also the U.S. Census Bureau's Current Population Survey and its yearly s t a t e population e s t i m a t e s from t h e Components of Population Change. Internal Revenue Service data also can b e used to produce migration estimates, as can information from the private sector, such as moving company shipments. Before reviewing t h e s o u r c e s of migration information released between the population c e n s u s e s conducted in 1980 and 1990, as a backdrop to t h e s e series this article briefly discusses the migration information that can b e obtained from the 1980 census. Table 1 summarizes key attributes of the data sources, including their frequency, sample size or migration universe, and level of geographic and demographic detail. Migration Data Sources 1 9 8 0 Census of Population. The U.S. government, by constitutional mandate, has undertaken a census of population every 10 years since 1790. Since the early 1900s the U.S. Commerce Department's Bureau of the Census has produced the census. In the 1980 census (and in the one just conducted) the source of migration information is a question regarding tenure of r e s i d e n t s : r e s p o n d e n t s were a s k e d whether they were living in the s a m e house five years earl ier, and, if not, in what U.S. state, county, city or town, or foreign country they 1 ived earl ier. T h e survey also asked whether their residence was within or outside the city/town limits. Census takers use the answers to t h e s e questions about residence in 1975 and 1980 to determine the U.S. population's geographic mobility. For example, a 1980 resident of Atlanta, Georgia, who responded that she lived in York, Pennsylvania, five years earl ier would have b e e n counted as an outmigrant from Pennsylvania and an inmigrant to Georgia. Three of the "supplementary reports" series publish migration data from the 1980 census. Statistics are also available from computer t a p e files. From several perspectives the Census of Population is the b e s t single information source on population mobility although even it has certain shortcomings. Census migration data generally are more comprehensive, give greater information a b o u t migrant characteristics, and offer more geographic detail. The decennial census counts the overall number of migrants during a I7 F E D E R A L RESERVE B A N K O F ATLANTA fixed t i m e period and provides gross a s well a s net migration streams among states. However, t h e s e data are b a s e d on a sample, a l b e i t a very large one, and thus are s u b j e c t to t h e kind of sampling error discussed above. In addition, s o m e inaccuracy arises from t h e five-year period covered by t h e mobility question. P e o p l e who moved away from and t h e n returned to t h e s a m e s t a t e within t h e 1975-80 period would not appear in t h e tally of interstate migrants, nor would t h o s e who m o v e d after 1975 b u t t h e n d i e d b e f o r e April 1, 1980. Meanwhile, o t h e r s who moved two or m o r e t i m e s during t h e period would b e c o u n t e d only once. Aside from t h e s e p r o b l e m s with t h e 1980 census, a major historical g a p in migration data exists b e c a u s e t h e r e is no standard r e f e r e n c e or reliable s o u r c e of p r e c i s e net migration estim a t e s by age, sex, and race at t h e s t a t e and county level during t h e 1970s, nor will t h e r e b e such a b e n c h m a r k s e r i e s for t h e 1980s. T h e decennial c e n s u s d o e s not ask a b o u t all migration s i n c e t h e previous census, and t h e amount of migration c a n n o t b e d e d u c e d from c e n s u s data for, say, 1970 and 1980. As o n e prominent r e s e a r c h e r s t a t e s t h e p r o b l e m , " T h e higher level of coverage in t h e 1980 c e n s u s c o m p a r e d with t h e 1970 c e n s u s prohibits strict application of previous (estimating) methods, and t h e r e is now no satisfactory m e t h o d for adjusting subnational areas for differential undercount in t h e two c e n s u s e s . " 4 In other words, b e c a u s e t h e n u m b e r s for population stocks are not strictly c o m p a r a b l e , t h e p r e c i s e amount of migration flow s u g g e s t e d by c o m p a r i n g t h e s t o c k s is wrong. Counting black men has proven particularly difficult. Census takers also find it hard to e n u m e r a t e illegal residents, t h e h o m e l e s s , and t h o s e who distrust government. Population e s t i m a t e s for geographic areas containing high concentrations of t h e s e groups may b e s u b j e c t to substantial error. A final shortcoming of c e n s u s migration data results from its relative infrequency. S i n c e c e n sus taking is a decennial event, significant shifts in migration s t r e a m s b e t w e e n c e n s u s e s can g o unnoticed or, m o r e likely, b e very much underor overestimated. Current Population Survey. A ready source of i n t e r c e n s a l information on migration is t h e Current Population Survey. Each m o n t h t h e Current Population Survey, a division of t h e U.S. 28 Census Bureau, c o l l e c t s labor force data for t h e noninstitutional civilian population from a sample of 57,000 households. Every March interviewers ask supplementary migration questions. As in t h e Census of Population, information on mobility is derived from answers to q u e s t i o n s on r e s i d e n c e o n e year b e f o r e t h e survey d a t e and t h e geographical location of t h e respond e n t ' s current r e s i d e n c e . T h e survey's estimation p r o c e d u r e extrapolates weighted s a m p l e results to various age, sex, and race categories. Data from t h e annual surveys are p u b l i s h e d in s e r i e s P-20 of t h e Current Population Reports; microdata c o m p u t e r files are also available for each survey beginning with 1968. In s o m e ways Current Population Survey data may b e superior to t h o s e from t h e Census. T h e Current Population Survey staff is permanent, whereas this year t h e Census Bureau e m p l o y e d temporary workers—more than 450,000 in 1990— to h e l p t a k e t h e c e n s u s . Among t h e s e , 200,000 enumerators visited t h e h o m e s of p e r s o n s who did not return their questionnaires. T h e smaller size of t h e t h e Current Population Survey operation also facilitates closer surveillance and control. Moreover, t h e yearly frequency of t h e Current Population Survey m a k e s t h e s e data m o r e s u i t a b l e for studying and analyzing t i m e trends and cyclical m o v e m e n t s in t h e statistics. Perhaps t h e major drawback of t h e Current Population Survey data inheres in t h e s a m p l e size: b e c a u s e only 57,000 h o u s e h o l d s are surveyed, details on geographic mobility are statistically val id only for t h e nation as a whole and its four major c e n s u s regions—the Northeast, Midwest, South, and West. Moreover, t h e range of migration e s t i m a t e s , within which 95 out of 100 survey s a m p l e s , say, would fall, is quite large. For example, t h e Census South's n e t gain from migration in 1986-87 is e s t i m a t e d at 279,000 plus or minus 193,000, and t h e Midwest's 111,000 e s t i m a t e d n e t outmigration is not statistically significant. Another s h o r t c o m i n g is that t h e data, while available yearly, are p u b l i s h e d with a c o n s i d e r a b l e lag—usually a b o u t two years and s o m e t i m e s longer. T h e March 1988 survey results will not b e p u b l i s h e d until late this year b e c a u s e t h e C e n s u s Bureau staff has b e e n working on t h e 1990 Census. Data on c o m p u t e r t a p e are available with a lag of a b o u t o n e year. 5 Components of Population Change. A s o u r c e of m o r e geographically d e t a i l e d yet still freE C O N O M I C R E V I E W , M A R C H / A P R I L 1990 Table 1. Sources of Migration Data Source Geographical Detail Flows Personal Characteristics Migration Universe/Sample Frequency Census migration data are reported in: State of Residence in 1975 by State of Residence in 1980 (PC80-S1 -9); Residence in 1975 for States, by Age, Sex, Race, and Spanish Origin (PC80-S1-16); Gross Migration for Counties: 1975 to 1980 (PC80-S1-17); and Vol. 2, Subject Reports: Geographical Mobility for States and the Nation and Mobility for States and the Nation and Mobility for Metropolitan Areas (PC80-2-2C). All from the Bureau of the Census. United States, region, division, state, and county In-Out-Net Depends on report; may include age, sex, race, Hispanic origin, marital status, education, labor force status, occupation, and income. Sample is 10 percent of census responses. Decennial starting in 1940. Most recent is 1980. Migration data from March Current Population Surveys are from micro-data files or from the published reports in the Current Population Reports series: Mobility of the Population of the United States: March 1986 to March 1987 (P20, no. 430) from the Bureau of the Census. United States, region In-Out-Net Age, sex, race, Hispanic origin, education, labor force status, income, occupation. Sample is 57,000 households from civilian noninstitutional households and members of the Armed Forces. Annual since 1948, but data are not available for 1972-75 or 1977-80. Most recent is 1986-1987. Estimates of net migration for states and counties are prepared annually by the Bureau of the Census and published in the P-25 and P-26 series of Current Population Reports. See, for example, State Population Estimates by Age and Components of Change: 1980 to 1989, Current Population Reports, series P-25, no. 1058 (Washington, D.C.: U.S. Government Printing Office, 1990). United States, region, division, state, and county Net None Estimate for resident U.S. population derived from administrative records and statistical techniques. Annual since 198081. Most recent is 1988-89. Individual Master File, Internal Revenue Service, U.S. Department of Treasury. State and county In-Out-Net None U.S. income tax filers Annual since 1980; also 1970-73; 197576; 1976-77; 197879. Most recent is 1986-87. Allied Van Lines and United Van Lines. State In-Out-Net None Company customers Annual since 1983 for Allied and since 1979 for United. q u e n t data on migration is t h e Census Bureau's annual e s t i m a t e of t h e c o m p o n e n t s of population change—births, deaths, and net migration— for states, metropolitan areas, a n d c o u n t i e s as of July 1. Net migration, along with total population e s t i m a t e s , is derived via t h e averaging of e s t i m a t e s from two m e t h o d s . O n e is t h e "comp o s i t e " procedure, which u s e s vital statistics (information on births and deaths), school enr o l l m e n t figures, a n d ratio-correlation t e c h niques in making t h e e s t i m a t e s . T h e s e c o n d m e t h o d is a " c o m p o n e n t " p r o c e d u r e that relies on administrative records such a s individual i n c o m e tax returns t o e s t i m a t e internal migration and immigration reports to calculate n e t immigration from a b r o a d . 6 Even though t h e C o m p o n e n t s of Population Change d o e s not m e a s u r e net migration at t h e s t a t e and local level directly, this s e r i e s is useful for estimating migration flows among states, large cities, and counties—areas for which t h e Current Population Survey d o e s not provide yearly e s t i m a t e s . For example, o n e can u s e t h e C o m p o n e n t s of Population Change to c o m p a r e metropolitan area growth rates and t h e c o n tribution of migration (internal and international) to an area's population growth. Also, t h e c o m p o n e n t s data for s t a t e s are available on a m o r e timely b a s i s than are t h e Current Population Survey data: e s t i m a t e s for t h e year ending July I, 1989, were available on D e c e m b e r 31, 1989, six m o n t h s after t h e r e f e r e n c e date. A major drawback of C o m p o n e n t s of Population Change data is their q u e s t i o n a b l e accuracy. Annually t h e Census Bureau p u b l i s h e s a revised e s t i m a t e for t h e cumulative net migration figure since 1980. T h e C e n s u s Bureau d o e s not publish data that give t h e yearly residual net migration e s t i m a t e s c o n s i s t e n t with t h e cumulative numb e r that is p u b l i s h e d in t h e Current Population Reports series, although t h e s e are provided upon r e q u e s t . While t h e agency considers t h e yearly data u n a c c e p t a b l e for publication, noting that it " c a n n o t d e f e n d t h e u s e of any individual n u m b e r s , " it d o e s say that " t h e annual figu r e s may b e indicative of g e n e r a l migration trends."7 Still, t h e C o m p o n e n t s of Population Change data have other major failings. They d o not provide information on gross inmigration and outmigration flows, nor d o they give any information on migrant characteristics. Also, t h e 1989 30 e s t i m a t e s are t h e most q u e s t i o n a b l e in t e r m s of accuracy s i n c e t h e population b a s e , or " h e a d count," to which estimating t e c h n i q u e s were a p p l i e d was nine years old. B e c a u s e of error accumulation, t h e further t h e annual derived e s t i m a t e s are from t h e b a s e year, t h e worse they become. Internal Revenue Service Data. While t h e C e n s u s Bureau is t h e main government s o u r c e of migration statistics, t h e Internal R e v e n u e Service (IRS) has s t a t e - t o - s t a t e migration flow data for m o s t years since 1970 and county-to-county data for s e l e c t e d t i m e p e r i o d s b e t w e e n 1978 and 1986. T h e s e statistics, g e n e r a t e d from t h e IRS data by t h e Census Bureau, are not as current as t h e C o m p o n e n t s of Population Change series, but they give m o r e detail. Information c o l l e c t e d c o m e s from administrative records of "One can use the Components of Population Change to compare metropolitan area growth rates and the contribution of migration (internal and international) to an area's population growth." income tax returns filed with t h e IRS. T h e migration e s t i m a t e s were d e v e l o p e d by matching t h e Social Security n u m b e r of primary taxpayers and comparing a d d r e s s e s on returns in c o n s e c u t i v e years. T h e total e x e m p t i o n s claimed by taxpayers with t h e s a m e Social Security n u m b e r b u t different a d d r e s s e s indicates t h e n u m b e r of p e r s o n s moving. An advantage of this s e r i e s c o m p a r e d with t h e other intercensal migration data s o u r c e s is that it provides p l a c e - t o - p l a c e migration flows b a s e d on t h e a d d r e s s e s supplied on individual income tax returns. Thus, it is p o s s i b l e t o identify which a r e a s are sending many or few p e o p l e to a given p l a c e and which are receiving many or few p e o p l e from another place. Information also is available at t h e county level. However, certain limitations of t h e s e data c a u s e IRS e s t i m a t e s to differ from actual migration flows. First, p e o p l e a r e e x c l u d e d if their E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 income is below the required minimum for filing. Second, the methodology assumes that all the p e o p l e listed as exemptions on tax returns move with the tax filer. Finally, the residences of taxpayers who file through the office addresses of their bankers, lawyers, or accountants will not b e accurately reflected in the data. Overall, IRS data underestimate the actual total migration flow. Another major shortcoming of IRS data is that information for 1985-86 and 1986-87 was determined to b e inaccurate for s e l e c t e d states, including Florida and Georgia. In response, migration flows b a s e d on the IRS data have b e e n reestimated by the Census Bureau for t h e s e years using a revised estimation methodology. 8 The new methodology also is being used to estimate migration flows for 1987-88. Unfortu- "[Moving company shipmentsl data do not accurately indicate the actual magnitudes of migration streams but are still valuable because they can point to changes in the size or direction of state migration." nately, the revised data were not available as of March 1990, further reducing the timeliness of the estimates. Moving Company Shipments. Two o t h e r sources of migration data c o m e from the private sector. Moving companies such as Allied Van Lines and United Van Lines r e l e a s e information on the number of state inbound and outbound shipments to each state handled by their companies each year. Both Allied and United claim in news releases that their company data on interstate household goods shipments are representative of the nation's mobility patterns. In 1989 the number of such shipments was 165,000 for Allied and more than 172,000 for United. Because of their timeliness moving company shipments data can offer valuable information insofar as they accurately reflect U.S. migration patterns at the state level. T h e s e data provide the most current information generally avail- a b l e that deals with interstate migration; data on shipments for all of 1989 b e c a m e available just one month into 1990. However, moving company data have certain drawbacks. First, it is not possible to determine the accuracy of the data as a measure of migration. To a degree, the flows reflect particular movers' market shares in different states. It is encouraging, though, that the series are highly correlated with each other and with the Current Population Survey data when aggregated to the census region level. In general, correlations of t h e s e data with the other available information series on migration tend to show higher associations for the inmigration and outmigration series than for the net migration series (see the box on page 32). Another shortcoming is that moving company shipments data reflect only a portion of total interstate migration b e c a u s e many migrants move t h e m s e l v e s . In particular, p e o p l e with lower incomes or seeking employment are more likely to undertake their own move, whereas those with higher income jobs (or their employers) hire moving companies. Movers' data also may b e procyclical. If jobs are plentiful, more households and companies are likely to hire movers. When j o b s a r e scarce, families forgo this e x p e n s e and move themselves. Given the limitations noted, t h e s e data do not accurately indicate the actual magnitudes of migration streams but are still valuable b e c a u s e they can point to changes in the size or direction of state migration. Although moving company s h i p m e n t s data do not provide information about place-to-place migration or the characteristics of those moving, several other data sources have the s a m e gap. Moving company shipments statistics are not the only source of migration data in the corporate sector. Records regarding customers or clients of utility companies, such as new hookups by electric power companies and billing address changes, can provide valuable insights into shifts and migration gains at the local level. Similarly, changes of addresses among credit card or banking account holders could yield additional information concerning migration. Data like t h e s e may b e especially helpful at the metro or county level in making migration estimates on a timely basis. For example, new hookups figures might b e used to estimate POPI7 FEDERAL RESERVE B A N K O F ATLANTA Associations a m o n g t h e Data S e r i e s How closely do the different migration estimates discussed in the text conform to each other? Except for moving company data, it is not possible to compare migration estimates directly because the measures differ in their timing and coverage of interstate migration. Nevertheless, as measured by the Pearson correlation coefficient, the different sources of migration tend to move in tandem. The correlation coefficient, denoted by r, is a summary number that quantifies the strength of the association between two variables. It is defined as n r = E (x,. - X) (y f i ^ j y) (N — l ) S x S y where N is the number of cases; Sx S y are standard deviations of the two variables; X, V are sample means; and X/( Vf are sample observations. The absolute value of r indicates the strength of the linear relationship. The largest possible absolute value is one, which means that a change in X determines exactly an increase in Y (and vice versa); a value of zero indicates no linear association. Calculation of correlation coefficients for the Allied Van Lines and United Van Lines interstate household shipments data for each of the seven years from 1983 through 1989 yielded the following coefficients: for inbound shipments the value was .98 for each year except 1983's value of .97; values ranged from .93 to .96 for outbound shipments and from .67 to .91 for net shipments. Statistically, these associations are very strong, suggesting that the two data series tell much the same story. Calculation of correlation coefficients for other pairwise comparisons are shown in the accompanying table. Overall, these association patterns suggest that the different data series, though not perfectly aligned, tend to be strongly associated. The weakest associations are between the moving company data and the Components of Population Change data in 1988 and 1989. The Components of Population Change data show outmigration for Alabama in the 1987-88 and 1988-89 period, whereas moving company data show net inmigration for 1987, 1988, and 1989. The results of the 1990 Census may eventually determine which estimate was correct. (Note, too, that the Components of Population Change data include immigrants.) Pearson Correlation Coefficients for Migration Data CPS Current Population Survey (CPS) IRS MCS. MCS, net = .77 in = .94 out = .75 net = .89 in = .92 out = .89 net = .79 in = .92 out = .87 net = .77 1983-89 .12-.91 1983-89 .41-.87 .92-,98 .92-.98 Components of Population Change (COP) net = .77 Internal Revenue Service (IRS) in = .94 out = .75 net = .89 1980-81-1985-86 .81-.92 Moving Company Shipments (Allied Van Lines) (MCS A ) in = .92 out = .89 net = .79 1983-89 .12-.91 ,92-.98 Moving Company Shipments (Allied Van Lines and United Van Lines) (MCSA+U) in = .92 out = .87 net = .77 1983-89 .41-.87 .92-.98 * Net comparisons 32 COP* 1980-81-1985-86 .81-.92 only. E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 ulation growth, using historical ratios for average household size and households per hookup. O n e could then calculate residual migration using vital statistics data on births and deaths by applying the "demographic equation": M = (p, - p 0 ) - (6 - D), where M represents net migration, p ( and p 0 stand for population in the current (P|) and b a s e (p0) period, B equals births, and D, deaths during the period. In other words, net migration equals total population change less the natural increase over the time interval. This methodology, as might b e e x p e c t e d , entails s o m e difficulties. For example, average household size has b e e n declining, and this decline would need to b e reflected in the ratio analysis. Aside from this technical issue and certain others, hookups data are proprietary and thus analysts and r e s e a r c h e r s may not have access to them, particularly if the electric utility company d o e s not regularly process and summarize them. Even if available, such data could b e c u m b e r s o m e or expensive to prepare for detailed analysis. Nonetheless, even raw data on hookups or building contractor advance orders for hookups could h e l p to i n d i c a t e whether migration is augmenting or draining an area's population growth. Case Study: Shifting Migration Patterns in the Southeast Information from the migration sources reviewed above can help construct the likely patterns of southern migration in t h e 1980s. In doing so, it is tempting to use the numbers to support various hypotheses that claim to explain why particular patterns appear. However, using these statistics in this way is risky. Whether migrants follow jobs or vice versa, for instance, is an unresolved issue among researchers d e s p i t e their u s e of s o p h i s t i c a t e d statistical techniques. What r e c e n t Census of Population d a t a show. The Census of Population provides the most accurate measures of mobility and Change in mobility over time, notwithstanding its shortcomings, which are probably inherent or very costly to remedy. When new census mobility data b e c o m e available, researchers in s o m e instances are surprised, but more often t h e s e statistics simply confirm trends and patterns already suggested by observation or other data. For example, comparison of interregional migration patterns using 1970 and 1980 census data revealed several significant but e x p e c t e d changes in southern migration patterns: 9 • net migration losses, which had b e e n typical throughout the twentieth century for most southern states, were reversed in the 1970s, and for the first time interregional migration added to the S o u t h s population; • experienced workers were becoming a more d o m i n a n t force in t h e S o u t h s migration gains; • the South gained workers in all the broad o c c u p a t i o n a l c a t e g o r i e s in t h e l a t e 1970s, whereas mostly white collar workers and skilled craftsmen flowed to the South during the late 1960s; • the South benefited from a North-to-South "brain drain" as a result of migration; • migration to t h e South h e l p e d lower t h e incidence of southern poverty; and • migration started adding more women than men to the South's population. Eight years have passed since the 1980 census data, from which t h e s e conclusions were drawn, b e c a m e available. Naturally, one wonders whether or not t h e s e patterns persisted in the 1980s and if they intensified or abated. The answers to t h e s e q u e s t i o n s were r e c o r d e d when the Census of Population was taken in the spring and summer. However, as mentioned above, the process of tallying and refining the statistics will take several years. Meanwhile the other sources of migration information discussed in this article can paint a preliminary picture of s o u t h e r n migration t r e n d s as they developed in the 1980s. What r e c e n t CPS interregional d a t a indicate. The net migration gain by the 16 states plus the District of Columbia that make up the Census South region slipped in the 1980s. 1 0 This increase declined from 1,986,000 p e o p l e in the 1975-80 period to an estimated 1,898,000 in the 1980-85 period. The yearly gain then dropped to a margin of 35,000 from 1985 to 1986 (not statistically significant) before rising to 279,000 b e t w e e n 1986 and 1987. Migration from the I7 FEDERAL RESERVE B A N K O F ATLANTA Northeast to the South slowed sharply in the second half of the 1975-85 period, while outmigration from the Midwest to the South increased. Between 1980 and 1985, for the first time ever, the South registered a net gain (roughly 60,000) from population exchange with the West. Over the next two years the South resumed its loss in exchange of residents with the West, but it continued to gain from the Northeast. It lost 117,000 in population exchange with the Midwest in the 1985-86 period but gained a similar number from 1986 to 1987. It is not the purpose of this article to explain why these and other migration patterns emerged or changed. 1 1 Nevertheless, the broad changes noted are consistent with major shifts in national e c o n o m i c and demographic determinants of migration flows. For example, as the large baby-boom generation ages, passing the time when they are most likely to migrate, population movement is slowing nationally. Also, as strengths and weaknesses have shifted from o n e e c o n o m i c s e c t o r to another, s t a t e s and regions most d e p e n d e n t on particular industries such a s energy or manufacturing have experienced attendant transitions in migration patterns. Net migration by blacks to the South also d r o p p e d b e t w e e n t h e 1975-80 and 1980-85 periods—from 108,000 to 84,000, or by about one-fifth. This percentage drop was much larger than the 4 percent decline for whites. However, the black n e t migration rate apparently has begun rising again in the 1985-90 period. The S o u t h s net addition of blacks was 18,000 b e tween 1985 and 1986 and 148,000 from 1986 to 1987. Black n e t migration a c c o u n t e d for over half t h e region's total migration gain during t h e s e two years. What C o m p o n e n t s of Population Change migration d a t a reveal. Migration patterns in several southern states have shifted markedly between the 1970s and the 1980s (see Table 2). The Components of Population Change data indicate that the South had gained an estimated 1.2 million fewer p e o p l e from migration in the 1980s in the period up to nine months before the 1990 Census was taken. Maryland, Virginia, North Carolina, and Georgia posted sharp increases, and all of the Atlantic coastal states gained population from migration in the 1980s. However, five states—West Virginia, Ken34 tucky, Mississippi, Oklahoma, and L o u i s i a n a experienced net outmigration during the 1980s after posting net gains in the 1970s. Examination of net migration yearly estimates for southern states reveals that sharp declines developed among energy-dependent states as oil prices dropped. Texas, for example, started the d e c a d e with a net migration gain of 431,000 in 1981 -82 but lost 122,000 residents to outmigration in 1987-88. Net migration a d d e d 91,000 residents to Oklahoma between 1981 and 1982, whereas 55,000 p e o p l e moved out from 1986 to 1987. Housing and office markets in t h e s e and other states plunged as the previously positive net migration stream reversed direction. Louisiana, the most energy-dependent southeastern state, experienced a shift in net migration from a 36,000 gain between 1981 and 1982 to an 89,000 loss from 1986 to 1987. T h e outflow apparently then stabilized, as losses held to 78,000 and 64,000 in the 1987-88 and 1988-89 periods, respectively. Altogether, between July 1983 and July 1989 Louisiana lost 308,000 people from migration, or 7 percent of its 1980 population base. Other southeastern states have had different migration experiences, according to the Components of Population Change data. The population influx to Florida and Georgia, which had b e e n fairly strong from 1980 to 1983, accelerated significantly in the July 1983-July 1987 period. Net inmigration then d r o p p e d s o m e w h a t in Florida and even more in Georgia over the next two years. 1 2 T e n n e s s e e ' s pattern was similar to Florida's and Georgia's but much smaller in magnitude. Alabama's net migration reversed twice in the 1980s, with gains posted in the 198387 period and losses at the beginning and end of the d e c a d e . However, the yearly figures for net migration over the d e c a d e in Alabama were small, ranging only from -10,000 to +11,000. Mississippi experienced net outmigration throughout the d e c a d e and especially during the last three years. What IRS d a t a show. Yearly p a t t e r n s of southeastern migration in the 1980s shown by the Internal Revenue Service data are similar to those indicated by the Components of Population Change data. However, IRS statistics reveal s o m e additional information about state-tostate flows (see Table 3): E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 Table 2. Southern Net Migration (in thousands) April 1970April 1980 West Virginia Maryland Monthly Average April 1980July 1989 Monthly Average 71 0.59 -143 -1.29 -36 -0.30 187 1.68 -6 -0.05 40 0.36 -164 -1.37 -60 -0.54 Virginia 239 1.99 377 3.40 North Carolina 278 2.32 351 3.16 South Carolina 210 1.75 159 1.43 Georgia 329 2.74 526 4.74 Delaware District of Columbia Florida South Atlantic 2,519 20.99 2,549 22.96 3,440 28.67 3,986 35.91 -1.05 Kentucky 131 1.09 -117 Tennessee 297 2.48 122 1.10 97 0.81 8 0.07 Alabama 31 0.26 -82 -0.74 East South Central 556 4.63 -69 -0.62 -0.14 Mississippi Oklahoma 230 1.92 -16 Arkansas 184 1.53 14 0.13 Louisiana 100 0.83 -226 -2.03 Texas 1,481 12.34 1,087 9.79 West South Central 1,995 16.63 859 7.75 South Total 5,991 49.93 4,776 43.04 Source: Computed by Federal Reserve Bank of Atlanta from Bureau of the Census, U.S. Department of Commerce, Components of Population Change, series P-25, various years. Alabama. At t h e beginning of t h e d e c a d e Florida, Texas, and Louisiana were t h e t h r e e largest gainers from e x c h a n g e with Alabama. Toward t h e e n d of t h e d e c a d e , though, Georgia and T e n n e s s e e had r e p l a c e d t h e two energyd e p e n d e n t s t a t e s as t h e biggest gainers from Alabama. In a startling turnaround Louisiana actually lost m o r e from migration e x c h a n g e with Alabama than all but o n e other state. It is likely that many of t h o s e who moved to Alabama from Louisiana were return migrants who had lost their j o b s in Louisiana's d e p r e s s e d economy. Florida. Texas was t h e biggest gainer from migration e x c h a n g e with Florida at t h e begin- ning of t h e d e c a d e but b e c a m e a net l o s e r when t h e price of oil started to drop in 1982. New York consistently ranked as t h e biggest contributor to Florida's migration gain, b u t several other large northern s t a t e s c o n t r i b u t e d significant n u m b e r s too. When Floridians left t h e state, they were m o s t likely to g o to Georgia, which drew m o r e and m o r e of its southern neighbors as t h e d e c a d e progressed. Georgia. Indeed, t h e biggest contributors t o Georgia's yearly net migration gain switched during t h e d e c a d e from Michigan to Florida. Generally, though, Georgia gained population from e x c h a n g e with its southern neighbors and I7 FEDERAL RESERVE BANK O F ATLANTA Table 3. Southeastern States' Net Migration Exchange* State Alabama Florida Georgia Louisiana Mississippi 1980-81 1981-82 1982-83 1984-85 1985-86 -2,948 -2,906 -532 GA FL NC -3,733 -2,429 -498 GA FL TN -3,858 -1,283 -168 692 732 1,185 LA MS IL 894 1,153 1,395 MS LA IL 1,265 1,666 1,838 GA NC AZ -3,619 -1,326 -209 GA NC DC -4,619 -1,997 -310 GA NC AZ -7,279 -1,550 -177 20,425 20,500 32,407 MI OH NY 18,269 22,358 36,573 IL OH NY 17,685 21,826 37,450 NJ OH NY 23,666 25,981 56,631 TX HI AK -290 -282 -258 DC MT ND -69 -43 -27 -14 -8 DC AK -32 -20 2,360 2,684 2,951 MI IL OH 2,451 2,622 3,674 IL FL OH 3,194 3,619 4,321 TN FL OH 4,492 4,619 5,438 IL OH FL 5,384 6,398 7,279 -6,374 -1,090 -630 CA TX MS -4,830 -1,424 -705 TX FL GA -4,470 -2,095 -977 TX FL CA -7,467 -2,963 -2,277 TX FL CA -10,441 -3,378 -3,315 MI AL IL 2,262 2,290 2,300 MI OH IL 1,106 1,107 1,399 OH IN IL 462 491 587 MI OK IN 234 313 591 ND IA IN 97 156 222 -2,868 -1,624 -758 TX LA OK -3,654 -1,349 -587 TX FL CA -1,910 -962 -532 TX FL GA -2,067 -1,491 -824 TX GA FL -2,652 -1,511 -1,455 TX FL AL -1,956 -1,641 1,265 602 661 1,230 AL IN TN 853 1,229 2,942 LA TN IL 705 1,134 1,152 IN LA IL 256 455 618 IN IL LA 333 671 1,756 TN IL LA 599 834 2,243 -4,029 -4,008 -1,792 TX FL LA -6,398 -2,766 -2,290 FL GA TX -2,374 -1,848 -1,639 OH IL Ml 764 1,072 1,611 IL TN MI 1,074 1,180 1,216 IN IL MI 740 965 1,900 TX GA LA -7,078 -1,909 -1,556 TX GA CA -11,835 -2,038 -1,522 GA NC TX -2,422 -1,382 -995 Ml OH NY 20,490 20,781 48,320 OH MI NY 21,347 21,735 40,307 OH MI NY TX LA AZ -2,430 -752 -284 TX LA OK -4,050 -842 -603 NY OH MI 2,792 2,824 3,050 NY Ml OH TX CO OK -4,027 -461 -203 TX OK CO MI AR AL 1,728 1,729 1,792 TX LA FL MI TN IL FL TX LA 1983-84 GA FL NC Ml IN IL Hl VT — — — — continued on next page from the large northern states. Between 1984 and 1985 Georgia posted net losses only with Hawaii (14 people) and Vermont (8) and t h e s e were not statistically significant; from 1985 to 1986 the only losses were with the District of Columbia (32) and Alaska (20). The Peach State's experience with Texas has b e e n similar to that of Florida, losing population at the beginning of the d e c a d e and then gaining later. IO IO N- CM T- M t (D N O l ß r h- TCO t T1t T-" I I < _i U U- ! O Z Louisiana. T h e Pelican S t a t e started the d e c a d e with net losses to only eight states; by 1985-86, it was a net gainer with only ten states, and its biggest gain was only 222 p e o p l e with Indiana. Typically, Texas drew more Louisiana residents than any other states; its gain rose from 4,000 in 1980-81 to over 10,000 in 1985-86. Mississippi. About as many states lost population to Mississippi as gained from it. However, net outmigration to states like Florida, Texas, Alabama, and Georgia e x c e e d e d its largest gains from other states. Interestingly, Louisiana was the second biggest gainer with Mississippi between 1980 and 1981 at 1,624 but the biggest contributor (2,243) to Mississippi's gain from 1985 to 1986. Like Alabamians, s o m e of t h e s e migrants undoubtedly were returning to their state of birth after having b e e n lured to Louisiana during the energy boom of the 1970s. S É CM CM LO CO O o O) co co I— (D N CM TF CO CO CO •T" CM' 1-" T-" CM" CO" I I I < _J o o ¡1 z X >. O d IO CM ^ CD CO r (Or CO" CM" R~ I I I £ io co co co Tf s t r- o CM CM CO < _| o O Ü- Z T CO i - cn co CD o co O) 05 CM" CM" T-" I ! I < _J O u. O) CM CO A> CM h- 0> CO N-" CM" T-" I I I X O d 2 CM CM 05 CO CO CM CO < x P 2 CD RT ^ O) T" I CM CO CO T-" I IO TO CM R-" I CO T CO CM CM O t in co CM" CM" CO" X _J hiTO fi g ir .Ol ¡5 E ® Tennessee. The Volunteer State has consistently contributed population to Florida and Georgia while gaining from Michigan, Illinois, Ohio, and Kentucky. On balance, the number of states from which T e n n e s s e e has b e e n a net gainer increased during the 1980s. However, migration d o e s not appear to have b e e n nearly as important a phenomenon in T e n n e s s e e as it has b e e n in Florida and Georgia (positively), or Louisiana (negatively). What moving company shipments d a t a suggest. Allied's headline on its 1989 data r e l e a s e reads, "Allied Van Lines' 'Magnet States' Survey Confirms U.S. Mobility Patterns toward U.S. Coastlines." United's headline states: "Oregon, Pacific Northwest Area Lead United Van Lines' 1989 Migration Patterns Study." Both releases discuss details of states' 1989 experiences, describing a similar, though not identical, picture of migration streams. (See the map on page 38 for a visual summary of combined Allied Van Lines and United Van Lines shipment data.) If accurate, t h e s e data provide encouraging news to southeastern states (see Table 4). For I7 FEDERAL RESERVE B A N K O F ATLANTA Map. Moving Company Shipments (Allied Van Lines and United Van Lines) (percentage inbound for selected states) 66.2 D.C. Top Ten Inbound Bottom Ten Inbound Source: S e e Table 1. Table 4. Combined Moving Company Shipments Net Migration Estimates State 1983 1984 1985 1986 1987 1988 1989 Alabama -221 -195 -290 432 395 420 456 Florida 8,465 9,237 9,846 9,944 10,462 8,875 9,561 Georgia 2,678 3,894 4,180 4,086 3,755 2,745 2,708 -1,096 Louisiana 182 56 -760 -1,661 -1,577 -959 Mississippi -97 -202 -241 -124 -319 -270 -209 Tennessee 169 225 303 1,448 1,815 914 1,143 Southeast 11,176 13,015 13,038 14,125 14,531 11,725 12,563 Source: C o m p u t e d b y Federal Reserve B a n k of Atlanta f r o m Allied Van Lines, Inc., a n d United Van Lines, Inc., data. Alabama net outmigration in t h e 1983-85 period gave way to net inmigration in each of t h e next four years. Florida and Georgia both e n j o y e d rising net inmigration in t h e 1983-86 period and then diminishing population inflows over t h e next two years; n e t inmigration may have sta38 bilized in Georgia during 1989 and r e s u m e d its growth in Florida. Net outmigration from Louisiana and Mississippi continued, b u t t h e amounts d e c l i n e d in t h e 1988-89 period c o m p a r e d to 1987. Meanwhile, net inmigration to T e n n e s s e e has b e e n higher in t h e 1986-89 period than it ECONOMIC REVIEW, MARCH/APRIL 1990 was earlier in the 1980s. For the region as a whole, migration gains i n c r e a s e d during the 1983-87 period, dropped in 1988, and stabilized in 1989. Directions for Future Migration Research As additional information about migration b e c o m e s available from the intercensal sources reviewed above and the 1990 Census of Population, several aspects of migration should elicit further research. With more and b e t t e r information, policymakers and planners will probably s e e k more accurate assessments of migration's impacts on income, educational n e e d s , and labor force skills. States that experienced reversals of migration patterns during t h e 1980s should b e especially i n t e r e s t e d in such research since it appears that past planning was often b a s e d on simple extrapolations and the implicit assumption that current economic conditions, particularly prosperity induced by high oil prices, would continue indefinitely, as would inmigration. In addition, future research ought to p r o b e the determinants of migration, including the effect on migration decisions of employment opportunities (or the lack thereof) and other economic conditions in both the sending and receiving areas. Conclusions This survey of the various data series on U.S. migration and review of what the data indicate about southeastern migration is revealing. First, a surprising amount of statistical information is available between the detailed decennial censuses—probably the b e s t source of information. A review of t h e alternative intercensal series indicates that each o n e offers certain advantages. The Current Population Survey data give the most detail about movements of major population groups, such as blacks, from one large region of the country to another. Components of Population Change estimates are c o m p r e h e n s i v e and fairly current. Internal Revenue Service data offer place-to-place migration estimates, and moving company shipm e n t s data provide t h e m o s t current direct estimates of interstate migration. Against the backdrop of information from the 1970 and 1980 Censuses of Population, t h e s e sources indicate several significant changes in s o u t h e a s t e r n migration patterns during t h e past d e c a d e . The Southeast continued to gain population from net inmigration. Florida and Georgia registered an acceleration from already strong levels. T e n n e s s e e ' s gains were positive, and Alabama's performance, though mixed, s e e m e d to b e strengthening at the end of the decade. In contrast, Louisiana and Mississippi suffered significant migration losses during the 1980s, though improvement s e e m e d evident in the past few years. The d e c a d e of the 1990s is sure to bring new surprises. As events unfold and state economies strengthen or weaken in response, migration patterns will continue to change. With new insights into the causes and c o n s e q u e n c e s of migration b a s e d on solid empirical tests using the extensive data available from this year's census, the burdens and benefits of swelling and sharply receding migration streams can b e managed better. I7 FEDERAL RESERVE B A N K O F ATLANTA Notes 1 In a nation of 250 million people who are highly mobile, census takers are bound to miss some people. Based on post-1980 census surveys and reviews of vital statistics, the Census Bureau estimates that the U.S. population on April 1, 1980, was 1.4 percent higher than the 226.5 million it reported. The undercount for 1970 was estimated at 2.9 percent; in 1960 it was 3.3 percent. 2 See, for example, J.D. Kasarda, M.D. Irwin, and H.L. Hughes, "The South is Still Rising," American Demographics 8 ()une 1986): 33-40. Other research, reviewed by M.). Greenwood in " H u m a n Migration: Theory, Models, and Empirical Studies," journal of Regional Science 25, no. 4 (1985): 521-45, suggests that more is known about the determinants of migration than about its consequences, but that even the causal linkages have not yet b e e n well established. t h r o u g h o u t this article Southeast refers to these six states, which are wholly or partly in the Sixth Federal Reserve District. 4 Larry Long, Migration and Residential Mobility in the United States (New York: Russell Sage Foundation, 1988), 284. The Census Bureau has published three components of population change for states during the decade of the 1970s: net migration, natural increase, and "error of closure." The last represents the difference between the Census Bureau's Components of Population Change estimate for April 1, 1980, and the 1980 Census of Population count. It amounted to over 4.7 million persons for the nation. These estimates appear in Bureau of the Census, Estimates of the Population of States: 1970 to 1983, Current Population Reports, series P-25, no. 957 (Washington, D.C.: U.S. Government Printing Office, 1984). 5 The Census Bureau p o i n t s out in an a p p e n d i x to its Geographic Mobility that two other current surveys conducted by the bureau give additional migration information. The American Housing Survey, conducted since 1973, contains residential m o b i l i t y data for selected metro areas, and the Survey of Income and Program Participation, begun in 1984, is a series of annual panel surveys of a 40 national sample of individuals who are reinterviewed every four months for a period of two and a half years. 6 Both of these methodologies are quite complex. A more detailed description of the estimating methodologies can be found in the Census Bureaus's Current Population Reports, series P-25, no. 957. I 7 This statement is stamped on the unpublished table on yearly residual migration that the Census Bureau sends to researchers upon request. 8 Detailed information on processing the IRS data can be obtained in an unpublished Census Bureau document titled Use of Federal Tax Returns in the Bureau of the Census, Population Estimates and Projections Program. 9 F o r a detailed discussion of changing southern migration patterns, see William ). Kahley, "Migration: Changing Faces of the South," Federal Reserve Bank of Atlanta Economic Review 67 (June 1982): 32-42. ^The states are Delaware, Maryland, Virginia, West Virginia, North Carolina, South Carolina, Georgia, Florida, Kentucky, Tennessee, Alabama, Mississippi, Arkansas, Louisiana, Oklahoma, and Texas. II For a detailed discussion and explanation, and also for projections of southern migration in t h e 1990s, see William ). Kahley, "interregional Migration: Boon or Bane for the South?" Federal Reserve Bank of Atlanta Economic Review 74 (lanuary/February 1989): 18-34. 12 State agencies and some state-affiliated institutions, such as the Bureau of Business and Economic Research (BBER) at the University of Florida, also produce migration estimates. These are prepared using similar techniques. For Florida in the 1980s, estimates prepared by the BBER show a pattern similar to that shown by the Census Bureau's Components of Population Change data. However, the BBER migration estimates have been 50,00055,000 higher over the past several years. BBER estimates for the 1980s, in thousands, are as follows: 1980,476; 1981, 333; 1982, 237; 1983, 181; 1984-85, 609; 1986, 336; 1987, 341; 1988, 327; and 1989, 326. E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 Book Review The Separation of Commercial and Investment Banking: The Glass-Steagall Act Revisited and Reconsidered by George ). Benston. New York: Oxford University Press, 1990. (Also published in England by The Macmillan Press, London, 1990.) 272 pages. $29.95. o n g r e s s p a s s e d t h e Banking Act of 1933 to r e s t o r e o r d e r to t h e b a n k i n g industry in t h e wake of financial distress during t h e late 1920s and early 1930s. T h e portion of t h e law authored by S e n a t o r Carter G l a s s a n d R e p r e s e n t a t i v e Henry S t e a g a l l — widely referred to as " t h e Glass-Steagall Act"— limited commercial banks' participation in t h e securities b u s i n e s s . T i e s b e t w e e n c o m m e r c i a l and investment banking had allegedly led t o such a b u s i v e b a n k p r a c t i c e s a s granting unsound loans to shore up investment affiliates, pressuring c u s t o m e r s to invest in securities, and investing in excessively risky long-term securities. C During t h e 1980s c o m m e r c i a l b a n k s h a v e urged regulatory a g e n c i e s and t h e c o u r t s t o i n t e r p r e t G l a s s - S t e a g a l l l e s s stringently and have l o b b i e d Congress to repeal t h e act. On t h e other hand, trade groups representing t h e s e curities and investment industries have fiercely c o n t e s t e d a t t e m p t s to relax restrictions or rescind t h e act. They argue that permitting banks additional securities powers would invite a repetition of t h e p r o b l e m s of t h e 1920s and 1930s. O p p o n e n t s of Glass-Steagall have c o u n t e r e d t h a t t h e r e is no e v i d e n c e t h a t c o m m e r c i a l banks' securities affiliates c a u s e d t h e c o l l a p s e of t h e banking system. While acknowledging that numerous other a b u s e s occurred during t h e 1920s, banks point out that most of t h e s e q u e s t i o n a b l e practices were legal at that t i m e but are now illegal. In a r e c e n t turnaround, t h e Securities Industry Association (SIA) e n d o r s e d a plan to allow commercial banking organizations to expand their securities activities (see Robert G u e n t h e r 1989). Congress, however, a p p e a r s unlikely to repeal Glass-Steagall promptly. G e o r g e J. Benston s t e p s into t h e m i d d l e of this d e b a t e with his latest book, The Separation of Commercial and Investment Banking: The Glass-Steagall Act Revisited and Reconsidered, which c o n s i d e r s a variety of arguments for and against revising or repealing Glass-Steagall. T h e b o o k ' s main contribution is its careful reexamination of t h e 1930s' congressional record to a s s e s s t h e e v i d e n c e of alleged bank a b u s e s during that period. Benston—John H. Harland Professor of Finance, Accounting, and Economics at Emory University in Atlanta—rejects almost all claims of a b u s e a s not being s u p p o r t e d by specific e x a m p l e s or empirical evidence. T h e I7 FEDERAL RESERVE BANK O F ATLANTA b o o k a l s o reviews c o n t e m p o r a r y a r g u m e n t s against granting additional security powers, inc l u d i n g c l a i m s that b a n k s ' security affiliates would exploit federal protection of commercial banks. An appendix reprints a survey by William M. Isaac and Melanie L. Fein (1988) of permiss i b l e security activities carried out by banking organizations. Sources of Information Many r e c e n t analyses of Glass-Steagall c i t e secondary s o u r c e s such a s Ferdinand Pecora (1939) to d o c u m e n t a b u s e s prior t o t h e act. Benston d e v o t e s an entire c h a p t e r to a discussion of s o u r c e s on a b u s e s in t h e 1920s, following t h e r e f e r e n c e s of prior authors back to t h e original sources—six congressional hearings and reports, along with t h e Securities and Exchange Commission's (1940) report, Study of Investment Trusts and Investment Companies. The congressional r e f e r e n c e s are: (1) 1932 hearings on foreign b o n d s a l e s held by t h e S e n a t e Comm i t t e e on Finance, (2) 1931 and 1932 hearings held by S e n a t o r Glass b e f o r e a s u b c o m m i t t e e of the S e n a t e Committee on Banking and Currency, (3) t h e 1932 report of t h e Glass s u b c o m m i t t e e hearings, (4) c o m m e n t s m a d e on t h e House and S e n a t e floors during May 1932 and May 1933 as r e p o r t e d in t h e Congressional Record, (5) t h e 1933 and 1934 Stock Exchange Practices (SEP) Hearings of t h e S e n a t e C o m m i t t e e on Banking a n d Currency (frequently c a l l e d t h e P e c o r a Hearings, after t h e c o m m i t t e e ' s c o u n s e l ) , a n d (6) t h e 1934 report of t h e S E P Committee. He also u s e s "scholarly articles and reports" that provide analysis, especially of financial data. Evidence That Securities Affiliates Caused Bank Failures T h e m o s t serious charge raised against commercial banks affiliated with securities firms in t h e 1920s was that such l i n k a g e s l e d to t h e c o l l a p s e of t h e banking system and thus contributed to t h e start of t h e Great Depression. Benston e x a m i n e s this issue from several perspectives. 42 T h e Glass S u b c o m m i t t e e Hearings (1931, 1063-64) list a variety of ways in which b a n k s can a d v e r s e l y affect their affiliates. B e n s t o n n o t e s , though, that this o f t e n - c i t e d list refers to potential p r o b l e m s and not necessarily to ways in which b a n k s had actually b e e n endangered. A review of t h e hearings, Benston claims, yields little e v i d e n c e that banks with securities affiliates were m o r e likely to fail or that they c a u s e d larger d e p o s i t o r l o s s e s when they did fail. 1 O n e c a s e in which a bank clearly failed a s a result of its affiliates was that of t h e Bank of t h e United States. Its affiliates' primary purpose, however, was to hold real e s t a t e ; t h e only s e curities affiliate that c a u s e d p r o b l e m s was engaged in buying t h e bank's stock. Furthermore, two of t h e bank's l e a d e r s went to prison for using b a n k a s s e t s for their personal gain. T h e s o l e lesson to b e learned from t h e Bank of t h e United S t a t e s ' failure, according t o Benston, is that bank affiliates may b e used as a m e a n s of hiding b a d loans from examiners. Prohibiting banks from having securities affiliates, he points out, cannot eliminate such practices. Benston also e x a m i n e s t h e hypothesis that large b a n k s ' s e c u r i t i e s a f f i l i a t e s d a m a g e d smaller banks by pushing securities on them. S e n a t o r Glass (Congressional Record 1932, 9887) a s s e r t e d that a significant portion of bank failures was c a u s e d by small banks' investment in long-term securities. B e n s t o n finds no support for this c h a r g e in t h e s u b c o m m i t t e e ' s hearings. I n d e e d , in t h e few c a s e s in which securities investments were discussed, t h e evid e n c e b e f o r e t h e s u b c o m m i t t e e s u g g e s t e d that i n v e s t m e n t portfolios w e r e n o t a significant c a u s e of failure. Evidence That Banks Committed Other Abuses B e n s t o n c o n s i d e r s a variety of a l l e g a t i o n s concerning abusive activities by banks' securities affiliates. Of t h e s e charges, t h o s e in t h e testimony of Charles E. Mitchell, p r e s i d e n t of National City Bank, and Albert Wiggin, presid e n t of Chase National Bank, b e f o r e t h e Pecora Hearings had t h e most influence on Congress's decision to p a s s t h e Glass-Steagall Act. T h e s e E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 [ [ [ men ran t h e two largest banks in t h e United S t a t e s at that t i m e . 2 Mitchell's a n d Wiggjn's testimony. B e n s t o n evaluates various authors' claims that Mitchell's and W i g g i n s testimony at t h e Pecora hearings e x p o s e d a variety of a b u s e s . He remarks that, although many of t h e s e alleged a b u s e s are not relevant to t h e separation of commercial and investment banking, they e n g e n d e r e d significant public r e s e n t m e n t in 1933, when t h e country was still in t h e throes of d e p r e s s i o n . S o m e of the q u e s t i o n a b l e practices not a d d r e s s e d by Glass-Steagall were excessive salaries and b o n u s e s as well as tax avoidance and, in s o m e c a s e s , p o s s i b l e tax evasion. Other charges raised at t h e hearings relate to p r o b l e m s that are potentially serious but not unique to commercial banks' investment affiliates, according to Benston. Moreover, B e n s t o n claims, a careful scrutiny of t h e record reveals most of t h e s e charges are not s u p p o r t e d . For example, Vincent Carosso (1970,330-31) a l l e g e s that National City Corporation (the investment affiliate of National City Bank) sold bonds issued by Minas G e r a e s (a Brazilian state, now Minas Gerais) with i n a d e q u a t e disclosure. O n e s p e cific allegation is that National City withheld particularly d a m a g i n g information a b o u t t h e s t a t e uncovered by o n e of t h e corporation's officials. B e n s t o n reports that this information c o n s i s t e d merely of a description of conditions under former l e a d e r s of Minas G e r a e s at least 10 years b e f o r e t h e s t a t e issued t h e b o n d s . S o m e of t h e alleged a b u s e s r e p o r t e d at t h e Pecora Hearings are unique to securities firms affiliated with commercial banks. If substantiated, t h e s e charges might justify legislation banning any relationship b e t w e e n t h e two types of financial firms. Yet, after careful examination, Benston c o n c l u d e s that t h e accusations fail to s u p p o r t G l a s s - S t e a g a l l r e s t r i c t i o n s . For instance, Benston c o n s i d e r s Carosso's (1970, 333) allegation t h a t National City Bank sold b a d loans to its investment affiliate in an a t t e m p t to disguise bad banking practices from t h e bank's shareholders. This charge would b e serious if the b a n k had u s e d t h e s a l e to hide t h e l o s s e s from bank examiners, but Carosso acknowledges that e x a m i n e r s had previously criticized t h e loans. Furthermore, Benston states, t h e bank's shareholders had b e e n informed a b o u t t h e loan b e f o r e its sale. Although t h e bank did not notify s h a r e h o l d e r s of t h e sale, t h e sale itself had no effect on t h e value of their shares. An owner of a given p e r c e n t a g e of National City Bank s h a r e s was required to own t h e s a m e p e r c e n t a g e of National City Corporation. National City mana g e m e n t claims that it m a d e t h e transfer to increase t h e bank's portfolio's liquidity. In any c a s e , m a n a g e m e n t thought t h e loans were g o o d at t h e time of t h e transfer. Trust D e p a r t m e n t s and Affiliated Investm e n t Companies. A final area of a l l e g e d a b u s e that Glass-Steagall purported to correct was t h e c o n f l i c t s of i n t e r e s t t h a t c o m m e r c i a l b a n k s affiliated with securities and investment firms may e x p e r i e n c e . David Silver (1987) alleges numerous a b u s e s by b a n k - s p o n s o r e d investment c o m p a n i e s . Tracing Silver's r e f e r e n c e s to supporting e v i d e n c e from t h e SEC Reports (1940, 1942), Benston d i s p u t e s s o m e of Silver's allegations but c o n c e d e s that s o m e e v i d e n c e exists to support other charges. In particular, Benston finds that b a n k s occasionally u s e d investment company funds to support activities of special value, such a s purchasing b a n k loans. Such activities by b a n k - s p o n s o r e d investment c o m p a n i e s could c a u s e substantial l o s s e s to investors if banks forced t h e investment c o m p a n i e s to purchase a s s e t s at above-market values. Benston finds no e v i d e n c e , however, that t h e s e practices resulted in l o s s e s to investment c o m p a n i e s or their shareholders. Moreover, he argues, t h e SEC Reports found far m o r e a b u s e s involving i n d e p e n d e n t investment c o m p a n i e s than among b a n k - s p o n s o r e d c o m p a n i e s . B e n s t o n a l s o d i s p u t e s o t h e r e v i d e n c e of a b u s e s involving investment affiliates, such as a U.S. S u p r e m e Court s t a t e m e n t in Investment Company Institute v. Camp (401 U.S. 617, 633 |1971|): "Congress had b e f o r e it evidence that security affiliates might b e driven to unload e x c e s s i v e holdings through t h e trust departm e n t of t h e s p o n s o r b a n k . " (The e m p h a s i s is B e n s t o n ' s . ) Although t h e h e a r i n g s p r o d u c e d e v i d e n c e that o n e trust d e p a r t m e n t b o u g h t securities from its affiliate and that s o m e other trust d e p a r t m e n t s may have m a d e such purc h a s e s , Benston says t h e record also suggests that m o s t b a n k s did not buy securities from their affiliates. Furthermore, B e n s t o n ' s review uncovered no e v i d e n c e that investment affilia t e s " u n l o a d e d " securities on trusts at a b o v e market prices. I7 FEDERAL RESERVE B A N K O F ATLANTA While B e n s t o n ' s analysis to this point appears sound, his evaluation of banks' investm e n t affiliates suffers from two minor weakn e s s e s . First, a careful reading of t h e S u p r e m e Court's opinion and Benston's analysis shows that t h e two are not necessarily in conflict. T h e court referred to e v i d e n c e that securities affiliates "might b e driven to unload" (emphasis added) securities on t h e trust affiliate. T h e court did not claim to have e v i d e n c e that such an action had occurred. S e c o n d , B e n s t o n c i t e s e v i d e n c e suggesting that investment affiliates in general, and City Bank Farmers Trust in particular, did not buy securities from their affiliates. However, evid e n c e not cited by B e n s t o n suggests that City Bank Farmers Trust Company, affiliated with National City Bank, may have bought securities from National City Corporation, t h e bank's s e curities affiliate. Carosso (1970, 332) n o t e s that Mitchell t e s t i f i e d at t h e Branch, Chain, and Group Banking Hearings (1930) that National City's trust affiliate would not buy from t h e securities affiliate without e x p r e s s permission from t h e maker. Mitchell (p. 1972 of t h e hearings) s t a t e d , however, that individuals who sought a trust a g r e e m e n t were shown " t h e advantages" of granting permission. This testimony i m p l i e s that s o m e trusts m a d e purc h a s e s from t h e investment affiliate, but t h e implications are never directly a d d r e s s e d at t h e s e hearings. In t h e G l a s s S u b c o m m i t t e e Hearings (1931), as q u o t e d by Benston, Mitchell testified that t h e trust affiliates' policies prohibited purchases through the investment affiliate, e v e n if such purchases were authorized by t h e trust a g r e e m e n t . Nonetheless, a careful review of t h e s e hearings d o e s not indicate when this policy took effect. Thus, contrary to Benston's conclusions, t h e record neither confirms nor disproves t h e trust affiliate's purchase of securities from National City's investment affiliate. Overproduction of Financial Securities Aside from bank failures and abusive practices s u p p o s e d l y e n g e n d e r e d by t h e affiliation of commercial and investment banking, s o m e 44 p r o p o n e n t s of G l a s s - S t e a g a l l b e l i e v e d that such r e l a t i o n s h i p s would n e c e s s a r i l y c a u s e commercial bankers to relax their normal caution. Benston e x p l o r e s this argument a s d e v e l o p e d by S e n a t o r Bulkley in a frequently c i t e d 1932 s p e e c h . 3 Bulkley claimed that b a n k s m a d e excessive margin loans to support their securities affiliates and that investment banking affiliates encouraged overproduction of securit i e s to k e e p t h e i r s a l e s f o r c e s busy. T h e s e c h a r g e s s u g g e s t that c o m m e r c i a l b a n k s and their securities affiliates contributed to t h e 1929 stock market crash by increasing alreadyexcessive speculation during t h e 1920s through t h e excessive u s e of margin loans and t h e sale of q u e s t i o n a b l e securities. B e n s t o n n o t e s that s u b s e q u e n t e c o n o m i c analysis d o e s not support Bulkley's claim that s p e c u l a t i o n f u e l e d by b a n k margin l e n d i n g c a u s e d or increased t h e severity of t h e D e p r e s sion. He a l s o p o i n t s out that G l a s s - S t e a g a l l d o e s not prohibit banks from making margin loans. Benston a d d s that e c o n o m i c activity and s e c u r i t i e s underwriting r e m a i n e d d e p r e s s e d long after bank securities affiliates were eliminated. If S e n a t o r Bulkley's s e c o n d charge were valid, then returns on bank-underwritten securities should have b e e n significantly worse than returns on securities underwritten by other firms, according to Benston. He n o t e s that exact c o m parison is difficult b e c a u s e t h e relevant returns are t h e u n o b s e r v a b l e ex a n t e e x p e c t e d distribution of returns rather than t h e o b s e r v e d ex post realized returns. However, c o n c e d i n g this l i m i t a t i o n , B e n s t o n c i t e s s t u d i e s by Terris Moore (1934) and G e o r g e Edwards (1942), both of which suggest that issues originated by c o m mercial b a n k s ' affiliates actually had slightly b e t t e r ex post returns than issues from indep e n d e n t organizations. Explanations for the Passage of Glass-Steagall Given that B e n s t o n ' s analysis suggests little e c o n o m i c rationale for t h e adoption of GlassSteagall, t h e q u e s t i o n arises a s to why t h e act was p a s s e d . In c h a p t e r 6 B e n s t o n d e v e l o p s t h e hypothesis that t h e act resulted from a c o m E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 [ bination of s p e c i a l i n t e r e s t trade-offs and banks' desire to avoid even worse legislation. Chapter 11 reviews a number of other factors that may have contributed to the passage of the act. Chapter 6 reports that banks' securities affiliates were increasing their share of the new issue market in the 1920s, originating over 50 percent of new bonds in 1929. Benston (pp. 13435) attributes this growing market share to two factors: (1) banks' "apparent positive reputations among their customers, which provided them with goodwill that they could transfer to securities products" and (2) economies of scale and s c o p e that could b e realized by combining commercial and investment banking. Bank affiliates' gain in market share provided unaffiliated securities firms with a strong incentive to s e e k a legislative ban on affiliation. Benston discusses Jonathan R. Macey's (1984) argument that, since Glass-Steagall could not b e justified on the basis of its public benefits, the actual motive behind the act must have b e e n to protect special interests. Benston is sympathetic to Macey's proposal but suggests that commercial banks had several reasons for not resisting the legislation. First, banks' securities affiliates had b e c o m e unprofitable by 1933. In addition, banks were already dealing with public outrage resulting from the large number of bank failures, the losses on investments made through bank securities firms, and the charges of wrongdoing by bankers, e s p e cially by Charles Mitchell. In fact, Benston avers, Winthrop W. Aldrich, the new p r e s i d e n t of Chase National Bank, probably lent support to Glass-Steagall to d i s t a n c e himself and the Rockefeller interests from the controversial activities of Chase's prior administration. S o m e of the explanations offered by Benston may h e l p to explain t h e p a s s a g e of GlassSteagall. However, Macey's claim that specialinterest theory must b e true b e c a u s e the publ ic did not benefit from the act is weak even if one is skeptical of the public interest justifications. The relevant issue, which is not discussed by Benston, is whether Congress believed GlassSteagall to b e in the public interest, not whether it actually was. The available evidence from congressional s p e e c h e s and newspaper editorials suggests that many believed that the act was indeed in the public interest. 4 Chapter 11 begins with a discussion of Senator Glass's long-held beliefs that banks' loans should b e limited to businesses' short-term credit n e e d s and that margin lending was a perversion of the banking system. Along with this explanation, Benston considers the failure of the Bank of the United States as well as the banking system's collapse in the early 1930s as reasons for eliminating bank securities affiliates and "punishing" banks. A structural weakness of the book is chapter 11 's recap of the reasons for passage of GlassSteagall presented in chapter 6. The division of explanations into two widely separated chapters interrupts t h e flow of the argument and causes an unnecessary repetition of material. A more serious problem for the book is that s o m e of its explanations for t h e p a s s a g e of Glass-Steagall may b e construed as providing ammunition for those o p p o s e d to repealing the act. B e n s t o n , a strong a d v o c a t e for repeal, s e e m s not to recognize this problem. He argues that bank affiliates benefited from the public's c o n f i d e n c e in banks, which had maintained good service records. Consumers s e e m to have felt that bank affiliates would give better or at least more conservative investment advice than independent firms. Banks and affiliates used this goodwill to boost their market share in the 1920s. Whatever e d g e affiliates may have had, they were unable to offset the magnitude of the 1929 market crash. Yet the public appears to have judged banks and their affiliates harshly after the crash b e c a u s e of unrealistic expectations of their investment prowess. Today's proponents of Glass-Steagall cite the public's overconfidence in bank affiliates' investment advice in the 1920s as justification for banning bank securities affiliates. Supporters of the act argue that consumers will suspend normal caution toward new investments if they are recommended by banks and will therefore b e more likely to make inappropriate, excessively risky investments. While policymakers should consider ways to reduce the risk that s o m e consumers will place excessive confidence in banks' affiliates, this issue d o e s not necessarily justify a ban on securities affiliates. A total ban would r e d u c e c o m petition among investment firms and eliminate s o m e gains in c o n v e n i e n c e that c o n s u m e r s would obtain from one-stop financial shopping. I7 FEDERAL RESERVE B A N K O F ATLANTA Furthermore, while s o m e consumers may place unjustified confidence in their bank's securities affiliate, several factors suggest that this problem would b e limited. Investors of the 1980s are likely to b e more sophisticated than those of the 1920s due to the growth of business and financial reporting on television and in print media. Thus, the proportion of investors who would naively trust advice from a securities affiliate of their bank has probably dropped significantly since the 1920s. Additionally, experience will teach consumers that the quality of investment advice offered by banks is not necessarily superior to that offered by unaffiliated firms. 5 Moreover, bank affiliates may have a greater incentive to maintain consumer goodwill than unaffiliated firms. Unaffiliated firms that make excessive promises to an individual risk losing that individual's securities business. Bank affiliates that promise more than they can deliver may lose not only an investor's securities busin e s s b u t also that person's accounts at t h e affiliate bank. Benston would probably not disag r e e with any of t h e s e points, b u t his argument would have b e e n stronger had he dealt with this issue directly. Contemporary Justifications for Glass-Steagall Benston recognizes that other arguments exist for continuing Glass-Steagall even if the original rationale is questionable. O n e commonly raised objection to repeal is that allowing banks to have securities affiliates would substantially increase the risks borne by the Federal Deposit Insurance Corporation (FDIC) and the Federal Reserve System. This view holds that securities activities are inherently riskier than traditional commercial banking activities and that risks incurred by o n e affiliate will ultimately affect the operations of the entire banking organization. Moreover, according to current Glass-Steagall proponents, the government, through the FDIC and Federal Reserve, would bear most of the losses associated with the bank affiliates' failure. While c o n c e d i n g that d e p o s i t insurance should b e reformed b e c a u s e it creates a moral hazard problem—encouraging bankers to take 46 on more risk b e c a u s e insurance limits the potential losses to depositors—Benston provides several arguments suggesting that securities affiliates will not substantially increase the risk borne by the government. The Federal Reserve, he claims, already provides a safety n e t for securities markets, as demonstrated during the October 1987 stock market crash, and banks may already achieve any d e s i r e d risk level through traditional commercial banking activities. Benston contends that securities activities would not increase banks' risks except in circumstances involving inexperienced bankers who do not fully appreciate the hazards inherent in activities. In fact, he observes, in the 1930s banks with securities affiliates had a failure rate significantly lower than that of banks without affiliates. Benston notes several studies that suggest potential reductions in banks' risk due to diversification. 6 Other studies indicate that corporate stock underwriting is not particularly risky. Though Benston admits that the effect securities affiliates will have on commercial banking is uncertain, he summarizes the evid e n c e as suggesting that "there is no reason to b e l i e v e that banks would b e more likely to b e c o m e insolvent" or place a greater strain on the federal safety net. He also d o u b t s that securities affiliates will make banks significantly more profitable. Proponents of Glass-Steagall contend that bank-affiliated securities firms would constitute unfair competition for independent firms b e cause deposit insurance would lower affiliates' funding costs and their captive market would provide ready customers on whom they could unload securities. In response, Benston argues that the relevant issue is not bank affiliates' possible lower funding costs but the opportunity cost of using those funds, which is ind e p e n d e n t of deposit insurance. Countering the argument that securities affiliates would have captive markets in the affiliate bank, bankadministered trusts, and correspondent banks, Benston asserts that securities affiliates which overbid for securities will create losses for their parent whether they sell the securities in the market or to the bank. Hence, the p r e s e n c e of a bank affiliate provides no incentive to overbid. Bank-administered trusts would violate the law if they purchased overpriced securities from E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 their affiliates. He also points out that securities affiliates would not b e u n i q u e in facing p o t e n tial conflicts of interest. Many unaffiliated s e curities firms have mutual funds that can give rise to t h e s a m e situation. Finally, he argues that c o r r e s p o n d e n t banks would b e unlikely to unload overpriced securities on their r e s p o n d e n t banks b e c a u s e t h e market is too competitive. B e n s t o n also c o n s i d e r s t h e effect that repeal of Glass-Steagall would have on banks' c o n c e n tration of power. T h e claim that repeal would give t o o much power to banks is b a s e d on t h e assumption that banks could u s e their trustadministered stock holdings plus any mutual fund holdings to control nonfinancial firms. B e n s t o n finds no e v i d e n c e that b a n k s h a v e used their trust holdings to d o m i n a t e other corporations and s e e s no indications that securities privileges would b e u s e d manipulatively. The e c o n o m i s t reviews a related argument— that repeal of t h e act could r e d u c e concentration by expanding t h e limited n u m b e r of firms that currently d o m i n a t e t h e underwriting of corporate securities. Benston finds t h e e v i d e n c e on both s i d e s of this argument a m b i g u o u s and suggests that t h e b e s t way to d e t e r m i n e t h e effect of banks on t h e market is to repeal t h e act and s e e what h a p p e n s . 7 T h e last i s s u e c o n s i d e r e d by B e n s t o n is universal banking. Universal b a n k s can offer a full range of financial services, may b e owned by nonbanks, and can vote t h e s h a r e s of c o m p a n i e s whose stock they own. This c h a p t e r cons i d e r s t h e a r g u m e n t s for a n d a g a i n s t b o t h specialized banking, as now m a n d a t e d in t h e United States, and universal banking, a s practiced in West Germany. Benston c o n c l u d e s that the substantial advantages universal banking can provide would outweigh t h e few p r o b l e m s it might c r e a t e for t h e economy. Summary An understanding of t h e merits of t h e GlassSteagall Act is important to Congress a s it considers whether to repeal or revise t h e act, and to regulators and t h e courts as they interpret it. T h e most important contribution of The Separation of Commercial and Investment Banking: The Glass-Steagall Act Revisited and Reconsidered is its thorough review of t h e d o c u m e n tation c o n c e r n i n g a l l e g e d a b u s e s by c o m mercial banks' securities affiliates during t h e d e c a d e b e f o r e Glass-Steagall was e n a c t e d . No future deliberation on e x p e c t e d a b u s e s if t h e act is relaxed will b e c o m p l e t e unless it considers B e n s t o n ' s work. His discussion of contemporary issues in reforming Glass-Steagall is largely a review of t h e existing literature, b u t it provides valuable insights on t h e s e studies' strengths and w e a k n e s s e s . Perhaps t h e book's least valuable discussion, certainly in t e r m s of t h e current d e b a t e on Glass-Steagall, is its a t t e m p t t o explain why t h e act was p a s s e d . Any discussion of this type is necessarily speculative. Moreover, understanding t h e reasons for passing t h e act d o e s not n e c e s s a r i l y h a v e much b e a r i n g on d e c i d i n g what t h e b e s t policy would b e in t h e future. Overall, however, B e n s t o n ' s b o o k s h o u l d b e useful to policymakers in that it provides a careful review of t h e e v i d e n c e c o n s i d e r e d in e n a c t ing Glass-Steagall and a timely examination of r e s e a r c h on this l e g i s l a t i o n ' s i m p a c t on t h e banking industry. Larry D. Wall The reviewer is research officer in charge of the financial section of the Atlanta Fed's research department. He gratefully acknowledges comments on the origin of Glass-Steagall by Matthew Fink. FEDERAL RESERVE B A N K O F ATLANTA I7 Notes 'Benston cites further evidence from White (1986) that securities operations were not responsible for the wave of bank failures in the early 1930s. White performs a variety of tests and concludes that securities affiliates d i d not endanger banks. A d d i t i o n a l allegations arising from this period concern trust departments' relationships with investment affiliates and investment companies' (now called mutual funds) relationship with their commercial bank sponsors. ^Congressional Record (1932), 1931-32. See, for example, "Big Bankers' Gambling Mania,'' The Literary Digest, March I I , 1933, 11-12. ''Indeed, the issue of misplaced confidence would probably not exist at this time if the Glass-Steagall Act had 4 never been passed. The stock market crash of 1929 taught investors that even securities promoted by bank affiliates could suffer significant losses. 6 H e also points out that one important study suggesting that securities activities are riskier than t r a d i t i o n a l banking, Boyd and Graham (1988), misclassifies six of the firms labeled as securities firms. 7 Benston inaccurately states (in chapter 9) that "concerns for concentration of power by commercial banks over corporations d i d not arise before the passage of the GlassSteagall Act." In fact, such concerns had existed at least since the House of Representatives conducted the Pujo Hearings, which investigated the concentration of the control of money and credit (Pujo Report 1913). References Boyd, lohn H„ and Stanley L. Graham. "The Profitability and Risk Effects of Allowing Bank Holding Companies to Merge with Other Financial Firms: A Simulation Study." Federal Reserve Bank of Minneapolis Quarterly Review (Spring 1988): 3-20. Branch, Chain, and Group Banking Hearings. U.S. Congress. House. Branch, Chain, and Group Banking. Hearings before the House Committee on Banking and Currency. 71st Cong., 2d sess., 1930. Vol. 2, pt. 15. Carosso, Vincent. Investment Banking in America: A History. Cambridge, Mass.: Harvard University Press, 1970. Congressional Record. 72d Cong., I st sess., 2 May 1932 to 17 May 1932. Vol. 75., pt. 9., 9357-10494. Washington, D.C., 1932. ,73d Cong., I st sess., 12 May 1933 to 25 May 1933. Vol. 77, pt. 4, 3295-4326. Washington, D.C., 1933. Edwards, George W. "The Myth of the Security Affiliate." Journal of the American Statistical Association 37 (1942): 225-32. Foreign Bond Hearings. U.S. Congress. Senate. Sale of Foreign Bonds or Securities in the United States. Hearings before the Senate Committee on Finance. 72d Cong., 1st sess., 1931 and 1932. Glass Subcommittee Hearings. Operation of the National and Federal Reserve Banking Systems. Hearings before a Subcommittee of the Senate Committee on Banking and Currency. 71st Cong., 3d sess. 1931. pts. 1-6, VII. . Operation of the National and Federal Reserve Banking Systems. Hearings on S. 4115 before the Senate Comm ittee on Banki ng and Currency. 72d Cong., 1 st sess., 1932. Glass Subcommittee Report. Operation of the National and Federal Reserve Banking Systems. Report No. 584 to accompany S. 4412, 22 April 1932. Guenther, Robert. "Securities Group May Agree to Bigger Role for Banks." Wall Street journal November 29, 1989, CI, C19. 48 Isaac, William M., and Melanie L. Fein. "Facing the Future: Life Without Glass-Steagall." Catholic University Law Review 37 (Winter 1988): 281-346. Macey, lonathan R. "Special Interest Groups Legislation and the Judicial Function: The Dilemma of Glass-Steagall." Emory Law journal 33 (Winter 1984): 1-40. Moore, Terris. "Security Affiliate versus Private Investment Banker: A Study in Security Originations." Harvard Business Review 12 duly 1934): 480-82. Pecora, Ferdinand. Wall Street Under Oath: The Story of Our Modern Moneychangers. New York: Simon and Schuster, 1939. Pecora Hearings. U.S. Congress. Senate. Stock Exchange Practices. Hearings on S. Res. 56, S. Res. 84, and S. Res. 97 before the Senate Committee on Banking and Currency. 73d Cong., 2d sess., 1933 and 1934. Pujo Report. U.S. Congress, House. Report of appointed pursuant to House Resolutions investigate the concentration of control credit. House C o m m i t t e e on Banking 1913. the Committee 429 and 504 to of money and and Currency, SEC Report. U.S. Securities and Exchange Commission. Study of Investment Trusts and Investment Companies, Part 2: The Statistical Survey of Investment Trusts and Investment Companies. 76th Cong., I st sess., 3 lanuary 1939. H. Doc. 70, 1939a. . U.S. Securities and Exchange Commission. Study of Investment Trusts and Investment Companies: Commingled or Common Trust Funds Administered by Banksand TYustCompanies. 76th Cong., 2d sess., 23 September 1939. H. Doc. 476, 1939b. . U.S. Securities and Exchange Commission. Study of Investment Trusts and Investment Companies: Part 3, Chapters I and II, Abuses and Deficiencies in the Organization and Operation of Investment Trusts and Investment Companies. 76th Cong., I st sess., 3 May 1939. H. Doc. 279, 1940. E C O N O M I C R E V I E W , M A R C H / A P R I L 1990 U.S. Securities and Exchange Commission. Study of Investment Trusts and Investment Companies: Part 3, Chapter VII, Abuses and Deficiencies in the Organization and Operation of Investment Trusts and Investment Companies. 77th Cong., 1 st sess., 26 February 1941. H. Doc. 136, 1942. SEP Report. Stock Exchange Practices, Report of the Committee On Banking and Currency, pursuant to S. Res. 84 (72d Congress) and S. Res. 56 and S. Res. 97 (73d Con- gress). Washington, D C.: U.S. Government Printing Office, 1934. Silver, David. Written Statement of the Investment Company Institute Before the Senate Committee on Banking, Housing and Urban Affairs, 23 ]anuary 1987. White, Eugene N, "Before the Glass-Steagali Act: An Analysis of the Investment Banking Activities of National Banks." Explorations in Economic History 23. (1986): 33-55. I7 FEDERAL RESERVE B A N K O F ATLANTA ; - EFT Brochure Now Available In an effort to help consumers, educators, and others better understand how payments are increasingly made through electronic transfer of funds, the Atlanta Fed has published Electronic Funds Transfer: Making Payments Simpler. The eight-page brochure outlines various ways consumers and businesses can make use of electronic funds transfers, including direct deposit, preauthorized bill payment, and point-of-sale debit systems. It also explains the Fed's role as well as that of private sector participants in EFT networks like the Automated Clearing House. Finally, the pamphlet explores future directions in electronic payments. Up to 50 copies may be ordered free. Additional copies are available for $.15 each. (Educators are exempt.) To order, write the Public Information Department, Federal Reserve Bank of Atlanta, 104 Marietta Street, N.W., Atlanta, Georgia 30303-2713, or call 404/521-8788. • r i Please send me EFT Brochure. copy (copies) of the Atlanta Fed's Name. Address. City. L State. ZIP. J | g i Economic Review • Federal Reserve Bank of Atlanta 104 Marietta S t , N.W. Atlanta, Georgia 30303-2713 Address Correction Requested Atlanta, Ga