The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
SEPTEMBER/OCTOBER 1 9 9 0 ECONOMIC PERSPECTIVES A review from the Federal Reserve Bank of Chicago C irc u it b re a k e rs H ig h w a y c a p a c ity and e c o n o m ic g ro w th Contents C irc u it b r e a k e r s ............................................................................................... 2 Jam es T. M oser Threatened with greater regulation, the markets have installed various safety measures to forestall trading disasters H ig h w a y c a p a c ity and e c o n o m ic g r o w th ............................................................................................ 14 D avid A . A schauer Economic growth benefits from improvements in highway systems and, by extension, from all improvements in public investment in infrastructure ECONOMIC PER SPEC TIV ES SEPTEMBER/OCTOBER 1990 Volume XIV, Issue 5 Karl A. Scheld, Senior Vice President and Director of Research ECONOMIC PERSPECTIVES is published by the Research Department of the Federal Reserve Bank of Chicago. The views expressed are the authors’ and do not necessarily reflect the views of the management of the Federal Reserve Bank. Single-copy subscriptions are available free of charge. Please send requests for single- and multiple-copy subscriptions, back issues, and address changes to Public Information Center, Federal Reserve Bank of Chicago, P.O. Box 834, Chicago, Illinois 60690-0834, or telephone (312) 322-5111. Articles may be reprinted provided source is credited and The Public Information Center is provided with a copy of the published material. Editorial direction Edward G. Nash, editor, David R. Allardice, regional studies, Herbert Baer, financial structure and regulation, Steven Strongin, monetary policy, Anne Weaver, administration Production Nancy Ahlstrom, typesetting coordinator, Rita Molloy, Yvonne Peeples, typesetters, Kathleen Solotroff, graphics coordinator Roger Thryselius, Thomas O ’Connell, Lynn Busby, graphics Chris Cacci, design consultant, Kathryn Moran, assistant editor ISSN 0164-0682 C ir c u it b re a k e rs These safety mechanisms are triggered by rapid or heavy market changes, and they can have unintentional effects on the financial system Jam es T. M oser The “circuit breakers’’ that have gradually been added to financial markets since 1987 got their toughest test of the year yesterday. They passed. —Wall Street Journal, July 24, 1990. The limits “did exactly what they were sup posed to do,’’ he said. —New York Times, July 24, 1990, quoting a trader. age to system integrity. Activation of a circuit breaker intentionally imposes costs that are expected to be less than losses realized by exceeding the system’s capacity. Cost consid erations naturally focus on the value the in tended users can expect to obtain through their use of the system. Activation of circuit breakers can also have unintentional costs. These have two sources. First, activation of circuit breakers can lead to unanticipated convenience losses. For example, system engineers may under value some activities lost when a circuit breaker is activated. Therefore, system users with a financial stake in its operation have incentives to increase system capacity by al lowing increases in the activation levels of circuit breakers. It is these incentives that produce pressure to re-allocate financial re sources toward increased investment in the system. Thus, when private interests are in volved, the ability to re-allocate resources insures that unanticipated convenience losses will be infrequent and temporary. Second, costs are also incurred when un planned uses of the system are disrupted. Sys tem engineers focusing on anticipated uses will not incorporate the value of unanticipated uses into their circuit-breaker decisions. These value losses are recognized only when service interruption motivates increased investment by such users. When value losses fail to attract investment, system engineers are not moti- Press reports describing the markets’ en counter with circuit breakers on July 23, 1990, regarded them as successful. Their apparent criterion for success is the fact that the Dow Jones Industrial Average rose 60 points after encountering the circuit breaker. The experi ence from other markets suggests that circuit breakers do not usually produce dramatic price reversals. But, they do have effects. This article examines these effects. Circuit breakers are mechanisms used by management to control activity in capacityconstrained systems. The term circuit breaker originates in electrical engineering to describe a pre-set switch that shuts down electrical activity in excess of a system’s design capac ity. The activation level of the breaker reflects an ex ante decision on the capability of the system. Circuit breaker activation is inherently costly. The system engineer designing a cir cuit-breaker makes an ex ante choice between temporary loss of the use of the system and reductions in the likelihood of permanent dam James T. Moser is a senior economist at the Federal Reserve Bank of Chicago. 2 ECONOMIC PERSPECTIVES vated to include these losses in the circuitbreaker decision. I refer to these interests as public, to distinguish them from private inter ests that do lead to increased investment in the system. In financial markets, the intended effect of circuit breakers is to halt trading when activity levels threaten market viability. Earlier circuit breaker policy was determined within the affected market by parties having private rather than public interests in the activation of circuit breakers. Exchanges, responding to these interests, developed three separate circuit breaker mechanisms. Order-imbalance circuit breakers are intended to protect the interests of market makers in specialist markets. Volumeinduced circuit breakers are intended to protect the viability of back-office operations. Pricechange circuit breakers are intended to bring excessive volatility under control. Recent developments in financial markets have elevated the importance of public inter ests. Markets are increasingly characterized as inter-related. This inter-relatedness increases the importance of price information flowing between markets. This is particularly true between the stock markets and the markets for financial derivatives—options and futures. Futures exchanges have developed stan dardized contracts for a variety of financial assets. Value changes in these contracts are closely linked to developments in their related asset markets. Thus, asset prices serve the public purpose of determining gains and losses in futures contracts. Futures exchanges and their customers have benefitted from the price information generated by asset markets. Acti vation of circuit breakers interrupts this infor mation flow, decreasing the public value of the services rendered by asset markets. Futures markets offer a distinct set of services including opportunities to manage risks and additional routes to price discovery. Circuit breakers activated in these markets similarly disrupt these services, lessening their value. The stock-index futures contract illus trates this. Prices for these futures contracts are for hypothetical baskets of stocks. Thus, a single quote determines the price of the futures basket, whereas in the asset market cash prices must be aggregated to produce a cash index. In addition, daily settlement in the futures market is in cash, greatly simplifying order FEDERAL RESERVE BANK OF CHICAGO processing. The simplicity of stock-index futures contracts produces an ideal instrument for institutions to manage systematic risk lev els through simultaneous trading in asset mar kets and futures. Circuit breakers disrupt the normal synchronization of price changes be tween futures contracts and asset prices. This disruption amplifies the risk that gains realiz able in one market may be unavailable to offset losses in the other market. The current proliferation of derivativeasset markets with differing capacity con straints, combined with intensive intermarket trading, raises coordination issues that were less crucial in the past. Circuit breakers acti vated in one market now can affect several markets, not only the market in which they were activated. Thus, circuit breakers in fi nancial markets can influence public interests. These public costs are realized in two ways. First, circuit breaker interruption of private markets serves to shift trading into markets that remain open. Such interruptions initiate a chain of events that ultimately generates de mand for a lender of last-resort to supply li quidity to the financial system. Second, pricechange circuit breakers shift credit risk to gaining positions that implicitly extend credit to loss positions. Their creditworthiness may decrease the quality of exchange guarantees of performance. The next section describes the three types of circuit breakers. Then, I examine the his tory of circuit breaker activity. An analysis of the unintended result of price limits on liquid ity demands and the quality of nonperform ance guarantees follows. C la s s ific a tio n o f c irc u it b reakers Circuit breakers are of three types. Each addresses a different design-capability issue. The first, the order-imbalance circuit breaker, occurs in specialist markets. Inequalities in the number of buy and sell orders are balanced by specialists trading for their own accounts. These trades maintain orderly markets by smoothing short-run order imbalances. Sub stantial order imbalances increase the risk bom by specialists. This, in turn, jeopardizes or derly markets. The second type, volumeinduced, occurs when order processing be comes uneconomic. At low volume, order processing does not meet costs. High volume impedes the ability of the exchange to effec 3 tively process orders. Markets close when either volume effect pushes trading costs to uneconomic levels. The third type, the pricetriggered circuit breaker, closes markets when a given price level is reached. This last type originated in the futures markets. Such circuit breakers are called “ price limits.” Justifica tions of price limits are couched in terms of controlling “ excessive” volatility. O rd e r-im b alan ce c irc u it b reakers Stock markets activate order-imbalance circuit breakers at the request of a specialist. The specialist asks for a suspension of trade in an individual stock when an order imbalance occurs. In these cases, suspension gives the specialist time to determine a market-clearing price based upon information obtained off the exchange floor. Following the price determi nation period, the market re-opens with the specialist taking a position at the newly deter mined price. The purpose of this circuit breaker is to protect the specialist from large losses. Order imbalances were a problem in both the 1987 and 1989 breaks. In 1987, selling pressure at the October 19 opening prevented trading in 140 of the NYSE-listed stocks in the S&P 500 during the first half hour. In 1989, openings on October 16, 1989 were similarly delayed. (Most stocks were reported not opened during the first fifteen minutes of trad ing. Beginning 8:45 CT, stocks began opening and trading was reported at 9:15.) In both cases, the intended effect of the order-imbal ance circuit breakers was to protect specialists from losses incurred by purchases in declining markets. Activation of these circuit breakers have unintentional effects. Trading halts in individ ual stocks create uncertainty about the correct level of the aggregate indexes. This, in turn, tends to be reflected in the futures contract. As a result, the futures contract becomes more likely to encounter a price limit. (On October 16, 1989, it did hit the open limit—5 points down.) When a price limit is reached futures trading stops, shifting some trades to the stock exchange. These trades tend to aggravate any existing order-imbalance problems. low trading levels, breakeven costs are not met. The determination of exchange trading hours recognizes that the fixed costs of opera tions must be covered by revenues generated from trading activities. Exchanges schedule closings based on expectations that additions to these fixed costs will not be adequately compensated. Thus, daily closes can be con strued as activation of a circuit breaker. Trading volume can surpass the ability of exchange back offices to process the paper work required to document executed trades. When this happens, the effectiveness of order processing is reduced, producing additional costs as the need for correcting orders rises. These costs are expected to rise with trading volume. With these additional costs, exchange operations can become uneconomic and the exchange closes. In 1968, the stock exchanges instituted a temporary four-day week during the last half of the year, closing on Wednesdays to increase the time available to process paperwork. Heavy volume in the period prior to the fourday week had led to increases in errors execut ing orders. These midweek closings insured that each five-trading-day delivery period included at least one nontrading day, allowing the back offices to catch up. More recently, heavy volume appears to have complicated the order-matching activity of the specialists. Stock trading volume during the 1987 price break surpassed the ability of specialists to match orders. As a result, execu tions were not timely and the ticker lagged current trades. Changes instituted after 1987 substantially increased the capacity of the exchange to process orders to an estimated one billion shares daily. However, the trading suspensions that resulted from the volume on October 13, 1989, suggest a lower capacity. At the rate of trading in the last hour of Octo ber 13, daily volume would have been just over 703 million shares. Volume on October 16, 1989, the heaviest day of trading since October 1987, was 416 million shares. After processing the overhang from the previous Friday (most stocks were trading by 10:15), stock trading proceeded smoothly all day. P ric e -lim it c irc u it b re akers V o lu m e -in d u c e d c irc u it b reakers The cost-effectiveness of order processing depends on the level of trading volume. At 4 In futures markets, price limits restrict trading to a band of prices generally symmetri cal above and below the previous trading day’s ECONOMIC PERSPECTIVES settlement price.1 The stated goals of circuit breaker policies have historically been to con trol volatility. More recently, price limits on stock index contracts have been set to coordi nate price movements in the cash and futures markets. Price limits serve as market-closing rules because: 1) short trades (sales of futures contracts) are not offered on up-limit days—the market clearing price is higher; and 2) long trades (buying futures contracts) are not offered on down-limit days—the mar ket clearing price is lower. Historically, the rules committees of the futures exchanges incorporated price limits into trading rules in response to threatened regulatory intervention. That pattern suggests that price-triggered circuit breakers would not exist without potential regulatory intervention. Past c irc u it b re a k e r e xp erien c e This history of the price-limit form of circuit breaker demonstrates that price limits appear to resolve “ political” volatility.2 The imposition of price limits, an apparent impedi ment to the price discovery purpose of futures exchanges, coincides with threats to the inde pendence of the exchanges. Rather than face increased regulatory oversight and lose their ability to resolve disputes internally, the ex changes accommodated pressures for regula tion by self-imposing price limits. Early h is to ry o f p rice lim its The earliest occurrence of a price limit in futures trading was at the Dojima exchange in Japan during the early 18th century. Settle ment in the koku “ small futures” contract for rice was determined by the average of the previous three days’ forward-closing prices. If this price deviated by more than a fixed amount from the cash price for rice, all con tracts were either reversed out or delivered. This effectively discontinued trading in the contract by eliminating all futures positions. Also, the futures price was tied to the cash market, avoiding the potential criticism that futures trading caused problems in cash mar kets. Imposition of the rule came during a time when rice markets were described as “ deteriorating.” Deteriorating markets are often characterized by price volatility. FEDERAL RESERVE BANK OF CHICAGO The first instance of a price limit rule in the United States came during the First World War. On February 1, 1917, Germany an nounced that its submarines would sink all ships found in the major Atlantic shipping lanes. Cotton prices for May delivery on the New York Cotton Exchange closed down by a record of over five cents a pound. By the following Monday, however, the market had recovered to within one and one-half cents of the earlier price. In subsequent weeks, futures prices continued to be extremely volatile. The threat of attacks on shipping continued to run down prices as traders feared lost access to the European markets. Cotton prices rose as mar kets responded to news of potentially large purchases of cotton for military uniforms. Congress responded by supplying flat-rate three percent war loss insurance—a substantial discount from the then-current Lloyd’s of London quote of ten percent. Cotton prices reached an all-time high following the intro duction of this subsidized insurance. The futures exchanges trading cotton responded to this volatility in two ways. On June 20, 1917, the British Board of Trade closed down cotton futures trading and the New York Cotton Exchange increased margin requirements. Separately, the U.S. govern ment requested a price limit on the cotton contract. On August 22, 1917, a three-cent price limit was imposed. This limit remained in effect for the duration of the war. Interest ingly, there is no record of a limit day during this period. Also during the First World War, the Food Administration froze prices on wheat to pre vent profiteering in that commodity. This action closed down trading in wheat futures at the Chicago Board of Trade. However, other grain prices were not frozen and their corre sponding futures contracts traded freely. Since these grains are partially substitutable for wheat, government policy regarding wheat induced volatility in other grains. Futures prices for these commodities reflected this volatility, attracting the attention of the Food Administration. As a result of this scrutiny, the Board of Trade instituted a two-cent per day price limit on the oat contract and the New York Mercantile Exchange introduced a threecent per day price limit on soy bean oil con tracts. These price limits were removed once 5 trading in wheat futures resumed after the war. Price limits were formalized in 1925 at the Chicago Board of Trade. The 1925 Annual Report reported a modification to all contracts allowing the Board of Directors to set pricechange limits of five percent of the preceding day’s average closing price, following a tenhour notice period. (For comparison, a five percent limit on wheat in today’s wheat con tract comes to 18.9 cents per bushel. The present limit is twenty cents per bushel.) De termination of an emergency was left to the Board. Nevertheless, price limits retained their temporary character, to be used only in emergency situations. Direct federal intervention in agricultural markets during peace time began in the early 1930s under the authority of the Agricultural Adjustment Act. The Federal Farm Board, attempting to maintain prices in spite of large supplies of wheat, opened long futures con tracts in May 1931 and 1932 wheat. Uncer tainty about government policy (including complaints that officials were manipulating prices in their own interests) increased the frequency of emergency use of price limits. (A recent proposal by Robert Heller makes similar use of futures markets. He argues the current policy of supplying liquidity during a market break disrupts monetary pol icy. Instead he suggests the Fed supply liquid ity directly by taking long futures positions. His use of futures contracts is reasoned from the same basis as the Federal Farm Board policy of six decades ago—both approaches avoid the problem of the federal government holding and disposing of assets. The experi ence of the 1930s suggests that careful consid eration should be given to the problem of contract expiration.)3 Passage of the National Industrial Recov ery Act in July 1933 opened the way for trade associations to enforce price stabilization agreements, with the federal government act ing both as architect and enforcing partner. Application for these partnerships was made through the National Recovery Administration (NRA) with the agreements chartered through Executive Orders by President Franklin Roosevelt. The agreements came in the form of codes for fair competition. Grain price volatility continued to be high after the Farm Board ceased its price manipu 6 lations. The drought of the period and uncer tainty about government policy were contrib uting factors. This high volatility led to De partment of Agriculture pressure in July 1933 for a fair-trade agreement among the grain exchanges. Pressure on the exchanges to com ply came in the form of a proposal by the Agricultural Adjustment Administration and the Grain Futures Administration that would have empowered the Secretary of Agriculture to modify and enforce trading rules at the futures exchanges. The proposed authority included limits on individual trading, limits on daily price changes, and margin setting. The futures exchanges complied with the request and Executive Order No. 6648, entitled “Code of Fair Competition for Grain Exchanges and Members Thereof’, was signed by President Roosevelt on March 20, 1934. The agreement, implemented the next day, included price limits which could not be exceeded, but did permit exchanges to set limits below the pre scribed maximums. The Supreme Court ruling in the Schechter Poultry Corporation case on May 27, 1935, declared the NRA codes unconstitutional. Following the Schechter decision, Congres sional hearings began on the Commodity Ex change Act to broaden the scope of federal regulatory powers over the futures exchanges. These powers had previously been lodged within the Grain Futures Administration. Con gressional discussion indicated the proposed Act would institutionalize the defunct NRA codes. To thwart increased regulation, the Chi cago Board of Trade incorporated permanent price limits on all its contracts. (At the same time, the Board of Trade also eliminated trad ing of options on futures, then called “priviledges [sic].” These were also targeted in Congressional hearings.) The action began the use of price limits as a standard contract fea ture. The Commodity Exchange Act later passed specifying only regulatory review, rather than expanded powers, over contract details—including price limits. C irc u it b re akers in th e 1 98 0s In 1982, futures contracts on stock indexes were introduced. The initial contracts, keep ing with standard practice, were introduced with price limits. However, for the first time since the 1930s, these limits were dropped on ECONOMIC PERSPECTIVES objections from New York stock trading inter ests. In late 1984, price limits were dropped on all International Monetary Market contracts for foreign exchange. The movement away from price limits continued until the market break of October 1987, when price limits were instituted on the S&P 500 contract. Three of the six commis sions studying issues of the market break rec ommended significant regulatory changes. With regard to price limits the recommenda tions differ substantially. The Brady Commis sion recommended coordinated trading halts. While no specific method was proposed, the Commission indicated that price limits should be considered among the possible mecha nisms. The NYSE “Katzenbach” study group said that price limits will not resolve market break issues. Their proposals focused on in creasing the cost of trading to prevent specula tion. They specifically proposed requiring delivery of stocks on stock-index futures contracts-increasing the cost of trading futures. The SEC study recommended against price limits on stock-index contracts. The SEC proposal suggested optional delivery of stock on index contracts, again increasing the cost of trading futures. After the 1987 break, price limits were imposed on stock-index futures. The stated reason for these limits was to synchronize futures and cash prices. In 1988, the S&P 500 contract traded with a level-determined price limit. At levels below 275, the limit was 15 index points ($7500 per contract); between 275.05 and 325, the limit was 20 index points ($10,000 per contract); and, above 325, the limit was 25 index points ($12,500 per con tract). Initial margins on these contracts were $15,000, twice the pre-break amount. In addi tion, a five-point limit was established at mar ket opening. On reaching an opening limit, trading is suspended for two minutes and re opened at a new opening level. The opening limit rule holds only for the first ten minutes of trading. The 1987 market break also led to intro duction of price-triggered circuit breakers on the New York Stock Exchange. After a fall of 25 points in the Dow Jones Industrial Average (DJIA), the Sidecar program re-prioritized orders, giving priority to small (less that one million dollars) orders. After a decline of 250 FEDERAL RESERVE BANK OF CHICAGO points in the DJIA, the stock market would be closed for one hour. After a 400-point decline in the DJIA, the stock market would be closed for two hours. In addition, the DOT (Desig nated Order Turnaround) program would be shut down after a 50-point decline in the DJIA. The m ini-crash o f O c to b e r 1 9 8 9 Recalling that the intent of these circuit breakers is to synchronize cash and futures markets, the events of October 13, 1989, pro vide a gauge for the usefulness of circuit breaker mechanisms. The evidence suggests that price limits did not synchronize these markets and may have routed dynamic-hedge trades into the stock market. At 1:43 (CDT) negotiators announced the failure of financing for the proposed UAL buyout. The announcement sent the stock and index-futures markets into a steep decline. At 2:00 the DJIA was down 55 points. This cor responds to a 7.3 point drop in the S&P. Seven minutes later, the S&P futures contract hit its limit— 12 points down. With futures trading suspended, the DJIA at 2:30 was down 114.76 points or roughly 15.3 S&P points. At 2:30, the futures contract reopened, but closed again fifteen minutes later—down 30 points. At the close of trading (3:00), the DJIA was down 190 points, or 25.3 S&P points. Quotes from the stock market clearly lagged behind those from the futures market. The circuit breakers do not appear to have kept prices in line. Trading volume was affected by the cir cuit breakers. Figure 1 shows NYSE volume for half-hour intervals for 10/13/89 and 10/16/ 89. Volume at 1:30 on 10/13 was 125.52 million shares for the day, or 12.55 million shares per half-hour interval. The market response to the UAL announcement in the 1:30-2:00 interval increased volume to 17.48 million shares, or 39 percent above the aver age prior to 1:30 but still less than two of the previous half-hour intervals. At 2:07 CDT futures trading was sus pended for the remainder of the half-hour period. Minutes later Chicago Board Options Exchange (CBOE) closed without re-opening. Volume during that period was 45.86 million shares, 265 percent above the average and more than twice the busiest previous period. During the last half-hour of trading, volume was 396 percent above the average—nearly 7 FIGURE 1 Stock volume levels: October 1989 millions of shares four times the busiest period before 1:30. This trading might be explained as a response to new information and, therefore, independent of the incidence of circuit breakers but evi dence in the Index futures pit suggests more was involved. After limits were hit in the S&P 500 pit for the second time, a limited number of sell orders were executed at the limit price despite the disadvantageous price obtained there. Further, at the official 3:15 close of index trading, 2,000 sell orders worth $330 million were said to be outstanding. The pattern of selling in the Index pit indicates that traders were searching for reliable executions. The closest available substitute to selling stockindex futures is the sale of stock holdings. Thus, the substantial increase in stock volume can be related to the incidence of the CME circuit breaker. (See Figure 2). The consequence of the volume increase may have been an increased difficulty in keep ing up with the flow of orders. Heavy selling after 1:30 CDT produced suspensions in ten stocks with seven not re-opening. This sug gests that stock markets were unable to handle the increased volume. Finally, the evidence from the 1989 price break reveals three weaknesses. First, volume increases after price limits were encountered suggest these circuit breakers routed trades from the futures markets to the stock markets. 8 This is a serious concern. There is good evi dence from the 1987 break that order imbal ances are positively correlated with price changes. Policies tending to exacerbate the order-imbalance problem are likely to increase price volatility during price swings encoun tered in the future. Second, both the price lags reflected in the DJIA and the suspensions in stock trading indicate that the circuit breakers did not keep prices in line. Third, taking the $330 million overhang in the futures market to be intended to cover stock positions of institu tional traders, at least one-third of a billion dollars went unhedged. N e w c irc u it b reakers in place After the 1989 market break, price limits were revised. The following describes current limit procedures for the S&P contract. The five-point opening limit is retained. After the opening interval and at all levels of the index, current levels are: On a 12-point drop in the index prior to 2:30 PM (Central Time), trading is suspended for thirty minutes; on a 20-point drop in the index prior to 1:30 PM, trading is suspended for one hour; on a 30-point change (up or down), trading is suspended until 50 percent of S&P stocks (by capitalized value) are open for trading. The NYSE also revised its circuit breakers to restrain program trading. After a 30-point drop in the DJIA, incoming orders are routed ECONOMIC PERSPECTIVES into the Sidecar for fifteen minutes. After a 75-point drop trading orders are Sidecar’d for thirty minutes. In addition, the CME rejects incoming S&P 500 contract orders after a 12point drop in the S&P. The emphasis on drops clarifies the pur pose of recent price-linked, circuit-breaker policies. They do not resolve cash flow prob lems for the futures exchanges—else limits would be imposed on the upside as well. Nor do they control volatility—for the same rea son. They do shield the futures exchanges from the criticism that futures trading pulls down stock prices. C irc u it b reakers and th e m a rk e t fo r liq u id ity Liquidity is the relative ease of matching buy and sell orders at recently observed prices. Sellers can always obtain liquidity by lowering offers to sell. The difference between the price they obtain and the previously observed prices they expected can be construed as the cost of liquidity. Buyers recognize that for some assets these costs may be high. Thus, their offers to buy incorporate the risk of en countering a high liquidity cost on the eventual sale of the asset. Buyers respond by adjusting bids downward. Markets respond by organizing to keep liquidity costs low. They accomplish this through efficient matching of buy and sell orders backed up by methods to handle any order imbalances that may arise. T h e m a rk e t-m a k in g a c tiv ity Market making refers to the activity of matching buy and sell orders. In specialist markets such as the stock exchanges, orders to buy or sell arrive at a central post, are matched up by a specialist, and are posted as transac tions. The specialist’s order book is unbal anced when the number of buy and sell orders at the most recent price are unequal. When these order imbalances occur, exchange rules require the specialist to trade for his own position—buying in a declining market or selling in a rising market. Since these trades are aimed at re-balancing the order book, they may be loss trades for the specialist; that is, buying above the correct market price in a declining market or selling below in a rising market. These trades produce a balance of buy and sell orders and fulfill the specialist’s re FEDERAL RESERVE BANK OF CHICAGO sponsibility of maintaining an orderly market. To facilitate this role, dealers have exchangerequired capitalization and minimum invento ries for their stock listings. Under an interest-rate targeting policy, the Fed acts as a marketmaker in markets directly linked to reserve assets. Reserve policy effects credit levels so that a stable monetary policy depends on a stable market for reserves. To maintain this stability, the Fed acts as a spe cialist in reserves—both buying and selling to prevent order imbalances. Links b e tw e e n fin a n c ia l m a rk e ts The Federal Reserve is affected by circuit breakers because markets for stocks, bonds, and futures contracts are fundamentally linked through the payments system and the market for reserves. To see this, consider the problem faced by the specialist after a steep decline in stock prices. In the process of buying stock to maintain an orderly market, losses have been encountered. In addition, inventories of stock, generally purchased on margin, have been marked down and require additional financing. Summing the financing needs of many special ists after declines of the magnitude experi enced during the breaks of 1987 and 1989, one will generally observe a large increase in the demand for loanable funds. Institutions sup plying funds to specialists respond by selling short-term Treasury securities to meet reserve requirements. Thus, the demand shock in the loanable funds market tends to destabilize markets for Treasury securities—orders to sell Treasury securities exceed buy orders. Shocks to the loanable funds market are also felt as the margin accounts of mark-tomarket assets are adjusted. Dynamic hedge trades in a declining market increase demand for Treasury securities placed in the initial margin accounts of long and short futures positions. Long and short positions marked to market add further shocks as losing positions sell Treasuries to generate funds required to cover calls for variation margin and winning positions invest cash balances in Treasuries. Over a period of time these shocks will net out. Nevertheless, lack of synchronicity in duces short-term swings in the supply of li quidity. Combining with these separate effects, stock-market specialists encountering losses from their market-making activities are seek- 9 FIGURE 2 ------- — — ------------- — Stock market DJIA 2753.28 UAL 285 \ 8:30 am _ DJIA 2736.68 1:00 pm ______— DJIA 2793„ 1:43 pm 2702.87 . 1:54. pm - _ , 2:00 pm _______ __________________ s,_________ ______________ The UAL break: Trading on October 13,1989 Volum e m oderate at 117.46 m illio n shares fo r the day. Trading in UAL halted. Stock prices begin decline. S & P futures ______ Negotiators announce their failure to obtain financing fo r proposed UAL buyout. Market opens,trends down. Futures market 7 ” 8WM"1V »..■■■«------------------ --------- UAL DJIA dow n 57 points (from previous close). D ro pco rre -( sponds to a 7.3 point drop in S & P 500 contract. Volume i last hour 25.54 m illion shares S & P futures 358.00 351.27 2:00 pm — -s—_sMJ jng funds in a market subjected to volatile levels of liquidity. In its capacity as reserve specialist, the Federal Reserve supports liquid ity by maintaining a balance between buy and sell orders for reserves and Treasury securities. This activity prevents short-run order imbal ances from wringing liquidity from the system. The credit-demand shock from specialists’ needs for funds are supported as the Fed adds reserves to the system. Importantly, Fed policy must first distin guish between the real and monetary compo nents of these shocks in the market for re serves. Facilitating the liquidity demands of a financial shock need not have real effects. Liquidity can be increased through purchases of Treasuries. Once the short-term credit needs of the payments system subside, reserve levels can then be reduced. These financial shocks can be identified by sharp market de clines accompanied by volume and order balancing problems. The timing of this credit accommodation requires consideration. Circuit breakers interfere with trades needed to generate liquidity. The appropriate time for the Fed to begin the supply of liquidity is at the point when trading halts create an imbal ance of buy and sell orders for reserves and Treasury securities. Policy considerations for mixed realmonetary shocks differ. In these events, the Fed must consider both the need for credit accommodation through its order-matching activity and its monetary policy which is im plemented through reserve-level choices. For example, liquidity operations after the 1987 break produced significant decreases in short term rates. Reserves were left in the system after October 1987, giving permanence to the October liquidity operations. The 1989 break was followed by reports that the Fed would supply liquidity as in 1987. These reports were later disavowed. However, open market operations on October 16, 1989, did effect a modest temporary increase in the reserve base. 10 ECONOMIC PERSPECTIVES In terms of the effect on credit markets, the expectation of increased credit availability is key. Interest rates fell on October 16, evidenc ing market anticipation of increased purchases of Treasuries. Interest rates rose shortly after wards as the Fed’s response and its disclaimers became known. Identification of a real component to a shock suggests consideration of a less tempo rary adjustment to the reserve level. Accom modation of temporary liquidity needs facili tates the allocation of capital. Failure to ac commodate liquidity tends to hinder the re allocation of capital, delaying recovery from the real shock. Once temporary liquidity needs are met, reserve policy should then focus on real, relatively permanent aspects. E ffe c t o f c irc u it b reakers Circuit breakers alter the effect of a mar ket move on credit markets by altering cash flows. Trading halts have three effects on the flow of funds. FEDERAL RESERVE BANK OF CHICAGO First, amounts marked to market based on prices recorded when trading was halted do not reflect market values—short positions record less gain in a price decline, long positions record less loss. Provided trading halts are synchronized across markets, amounts marked to market for related securities are similar causing no excess demand for loanable funds. Unsynchronized trading halts, on the other hand, tend to produce asymmetry in that losses and gains on related assets are unequal. The liquidity needs produced by losses incurred in one market cannot be covered by recognition of gains in related markets. Thus, nonsynchronized trading halts increase the demand for loanable funds and shift liquidity trades into those markets which remain open. This places greater stress on these markets. Consider, for example, a specialist hedg ing his equity position with a short futures position in an index contract. A trading halt in the index futures contract can result in equity 11 losses exceeding gains on futures. In an unre stricted market, the futures position, in this example, would generate needed cash to cover the financing needs of the stock position. A halt in futures trading reduces the flow of funds to the specialist, increasing dependance on borrowed funds. Inventory financing needs that cannot be met with gains from the futures contract must be covered by increased borrow ing. Second, positions not marked to market are affected like marked-to-market accounts. The difference is one of form, not result. Gains and losses on stocks or options are real ized by unwinding the position. Circuit break ers halt trading and prevent unwinding of contracts. This restricts access to invested balances required to cover losses, realized elsewhere, increasing the demand for credit. For example, traders holding UAL on 10/ 13 attempted to sell out after the 1:43 an nouncement. The halt in trading of that stock initiated a search for close substitutes for UAL stock. The nearest substitutes were other take over stocks and transportations, particularly airlines. Order books for these stocks quickly became unbalanced. Three Big Board stocks halted trading temporarily: USAir Group, Delta Air Lines, and Philips Industries. Seven Big Board stocks halted and remained closed for the day: UAL, AMR, BankAmerica, Walt Disney, Capital Cities/ABC, Philip Morris, and Pacific Telesis. Sales again shifted; first to index futures, then to a broad range of stocks after the 15-point limit halted futures trading. These first two effects of price limits derive from restrictions on investor access to liquidity. During market breaks when liquid ity is most valuable, circuit breakers reduce the number of routes available for private resolution of liquidity needs. This tends to increase demand for a source of last resort to supply liquidity—a role many expect to be taken up by the Fed. A third effect derives from responses to price uncertainty as clearinghouses re-consider prudential margin levels. The trade halt pro duced by a circuit breaker creates uncertainty about the market’s actual volatility. Since margin levels are determined in response to estimates of price volatility, risk-averse clear inghouses are forced to estimate margin needs 12 on a worst-case basis. This will tend to in crease the margin levels required by prudent clearinghouses. Recognizing that further losses to customer accounts may be substan tial, initial and variation margin levels are increased to prevent losses from spilling over from customer accounts into clearing-member accounts. This effect tends to decrease the supply of loanable funds by increasing use of Treasury securities to meet margin obligations. C re d it ris k due to loss fin a n c in g The previous comments on circuit break ers emphasize problems induced by disruption to the flow of funds. Circuit breakers can also be viewed as shifting credit risk. Failure to record the full loss amount in a marked-tomarket account implicitly extends an interestfree loan for a portion of the loss amount to the losing position. The amount of this loan is the difference between the amount marked to the settlement price and the amount marked to the true market price. The loan is extended to losing positions from gaining positions. The amount of credit extended by these loans can be considerable. To illustrate, I will use the October 1989 market break. Taking the true futures price to be roughly the 10/16/ 89 opening, the true settlement price for the 117, 202 December contracts outstanding should have been 323.85. The difference between the actual settlement of 328.85 and the estimate of the true settlement is 5 S&P points. The amount of credit implicitly ex tended to short positions over the weekend of 10/14-10/15 was, therefore, $58.6 million or 3.3 percent of the value marked to market on 10/ 13.4 To gauge the risk to the financial system, we need to recall that futures clearinghouses provide performance guarantees for contracts trading on their affiliated exchanges. The quality of these guarantees depends on the amount of potential loss relative to equity. As potential losses increase relative to a fixed level of equity, the possibility of default rises, diminishing the quality of any guarantees. Book equity balances for the CME at the end of 1988 were $79.3 million. The $58.6 million implicitly lent to short positions is 73.9 per cent of book equity. The full implication for contract perform ance guarantees is not known. Threats to these guarantees will tend to shift trading away from ECONOMIC PERSPECTIVES the futures markets as the perceived quality of the guarantees declines. This tends to increase the credit needs of specialists operating in stock markets, requiring increases in reserves to meet these demands. Thus, the significance of the credit balance implied by price limits bears investigation. The policy issue is the viability of contract performance guarantees provided by the futures exchanges. C onclusion This article examines the effects of circuit breakers on the stability of the financial mar kets. Circuit breakers are classified into three types based on capacity issues: volume-trig gered circuit breakers halt trading when vol ume exceeds order-processing capacity; orderimbalance circuit breakers halt trading when orders to buy or sell threaten the viability of the specialist; and price-limit circuit breakers halt trading when price changes are regarded as excessive. The history of price limits sug gests they are introduced when futures ex changes are threatened with greater regulatory oversight. This paper argues that circuit breakers re duce access to markets. This reduces the abil ity of markets to resolve needs for liquidity. Second, price limits extend credit to loss posi tions in futures and options markets. Since clearinghouses guarantee contract perform ance, these guarantees may be threatened by large credit balances. On several recent occasions, circuit break ers have proved of some value in market crises. But it must be remembered that their value is not costless, nor their benefits without limit. FOOTNOTES *A notable exception to price-limit symmetry is found in the stock-index contracts. These are discussed in the next section. 2This term is from Joseph A. Grundfest, Commissioner of the Securities and Exchange Commission. ^The Heller proposal is in Heller, R., “ Have the Fed Sup port Stock Market, Too,” Wall Street Journal, October 27, 1989, p. A 14. An analysis of the 1930s experience is in FEDERAL RESERVE BANK OF CHICAGO Moser, James T., “ Public Policy Intervention Through Futures Market Operations,” forthcoming in the Journal of Futures Markets. 4 The 10/16 open price is probably too high. Opening contract prices on that date encountered the CME open limit of five points, preventing realization of a lower price on the S&P contract. Thus, our estimate may significantly underestimate the amount of credit extended. 13 H ig h w a y c a p a c it y and e c o n o m ic g ro w th The quality and quantity of highway transportation systems have a direct bearing on economic growth— good roads are good business D avid A . A schauer To the commuter struggling along the clogged freeways of southern California, this sta tistic must seem unlikely: the average auto commute in Los Angeles County took only 22 minutes in 1985. Even more unlikely: that time was shorter than 1980’s average, 23.7 minutes. After decades of increasing traffic and looming gridlock, how could these daily pilgrimages have be come shorter? One answer is suggested by Peter Gordon, Associate Dean of the School of Urban and Regional Planning at the University of South ern California. Gordon and his colleague, Harry Richardson, say that the highly devel oped freeway system in the Los Angeles area has allowed business and industry to further decentralize, often locating (or relocating) along the freeway system. It is this shift that has helped to shorten the commuter trips. Four minutes or so a day per worker may not seem like much. But it adds up to nearly two full working days a year per worker, in a working-age population of some 5.4 million. And industry’s intelligent use of the freeway system has other benefits, such as shorter de livery and pick-up times. The concepts and empirical evidence contained in this article support the idea that transportation infrastructure plays an important role in the process of regional economic growth. While it is common for economists to argue that investment is a key determinant of productivity growth and economic develop ment, it is often the case that the particular investment chosen for analysis is quite limited in scope. Indeed, public investment in infra structure capital—streets and highways, mass transit, airports, water and sewer systems, and the like—is typically left out of growth discus sions, at least at the level of national, aggre gate analysis.1 Only a relatively small number of studies have sought to establish the importance of infrastructure investment to private sector productivity and income growth. In a series of papers, I have developed a framework (Aschauer 1988, 1989a, 1989b, 1989c) with three basic empirical implications: 1) That infrastructure capital carries a positive mar ginal product in a private-sector neoclassical production technology; 2) That infrastructure capital is complementary to private capital and is capable of enhancing the marginal product of private capital; and 3) That infrastructure investment is likely to spur private investment in plant and equipment. The empirical results contained in those papers are in broad confor mity with the underlying framework. Holz-Eakin (1989) and Munnell (1990) come to nearly the same conclusions using slightly different empirical approaches or sample periods. Similarly, Garcia-Mila and McGuire (1987) establish a contemporaneous, positive link between the stock of highways and per capita output. Based on the results of 14 ECONOMIC PERSPECTIVES When he w rote this article, David A. Aschauer was a senior econom ist at the Federal Reserve Bank of Chicago. He is now the Elmer W. Campbell profes sor of economics at Bates College, Lewiston, Maine. these studies, one might be convinced that by ignoring public capital stocks the relationship between investment and economic growth is misspecified and potentially underestimated. Still, legitimate questions may be raised about the results in the aforementioned papers. For instance, the estimates in Aschauer (1989a) seem to suggest a marginal productiv ity of public capital in private production which is “too high.” The elasticity of private sector output with respect to public capital is approximately the same as that with respect to private capital while the public capital stock is approximately one-half the size of the private capital stock. This implies a marginal product of public capital which is approximately twice as large as that of private capital. Perhaps, it may be argued, the correlation between the public capital stock and private sector produc tivity is merely evidence of economic causa tion running in the reverse direction—from productivity through per capita output and, in turn, through tax revenues to the demand for public capital. This article develops an alternative esti mation strategy in order to establish the direc tion of causation from highway investment to economic growth. Specifically, this article searches for a connection between the level of highway capacity and the growth rate of per capita output. The following section lays out the conceptual approach. The next section contains a description of the data and a discus sion of empirical results. The article con cludes by offering some suggestions for future research. highway capacity regardless of the level of vehicle density; additions to the highway stock reduce travel time and, thus, increase traffic flow and the associated transportation services. The production technology can exhibit a posi tive, a flat, or a negative marginal product of density, however, depending upon whether density is below, at, or above a certain critical level, vd*, which is typically termed the “ bot tleneck point” for the highway stock.*2 The production function is depicted as the funda mental diagram of traffic in Figure 1. For a given level of highway capacity, the produc tion of transportation services increases with vehicle density up to the bottleneck point, and declines with further increases in density.3 A number of empirical studies, such as Fare, Grosskopf, and Yoon (1982), have confirmed this relationship for isolated locales. FIGURE 1 Fundamental diagram of traffic transportation services C o n c e p tu al issues The conceptual analysis centers on the linkages among highway capacity and the production of transportation services, private sector investment, and economic growth. Transportation services are taken to be “ pro duced” by a simple neoclassical technology 1) t. = f(vdj, hi.) ? + where t. = transportation services (measured as a flow of vehicles per time period) in a par ticular locale j; vd. = vehicle density (meas ured as vehicles per mile of highway); and hi. = highway capacity (measured as miles of highway). The production technology is char acterized by a positive marginal product of FEDERAL RESERVE BANK OF CHICAGO This article links the level of highway capacity to a measure of economic growth across localities. I argue that the return to productive activity (apart from transportation services) in any place is positively related to the level of transportation services, measured as a flow of vehicles per time period. Thus, I postulate the rate of return function 2) r. = r(t.) = r(vd., hi) + ? + so that the return from production, r, in locale j depends on the degree to which the highway 15 3) Dk. = g (r.-p ) where Dk. = the growth rate of the physical capital stock per person in locale j. Finally, non-transportation output per person is assumed to be related to the accumu lated capital stock per head according to a Cobb-Douglas production function augmented by a common rate of exogenous technological growth and a “catch-up” factor whereby total factor productivity in any given local is al lowed to converge on that of other, leading locales. Following Dowdrick and Nguyen (1989), this allows us to write the growth rate of per capita output in the form 4) Dyj = a0 + a^y^O) + a2*DL where y.(0) is the initial level of output in locality j; and where < 0, and a2 > 0. Com bining Equations (2), (3), and (4) yields the growth relationship between output, vehicle density, and highway capacity 5) Dyj = y(yj(0), vd., hi.) — ? + so that output growth will be negatively re lated to the initial level of output; positively or negatively related to vehicle density (depend ing on whether vehicle density has passed the bottleneck point); and positively related to highway capacity. The logic of this approach is quite simple. An increase in the stock of highways for a given locale generates a higher return to local, productive activity by raising the level of transportation services available to producers. This higher return to production, in turn, stimulates private investment in these produc tive facilities. The increased investment car ries with it higher growth in output and income for the particular locale.4 Increased productivity, of course, is not the only possible mechanism by which infra structure in general, or highways in particular, might affect the rate of economic growth. Murphy, Shleifer, and Vishny (1989) suggest 16 that investment in infrastructure (in this case, a railroad) may result in lower production costs to a number of economic sectors. These “ex ternal effects” of the investment allow for a multiplicity of equilibria, so that infrastructure spending can generate a “big push” to a higher level of output.5 In a recent paper, Romer (1989a) con structs a model in which technological change evolves endogenously as the result of profit maximizing investment behavior by imper fectly competitive firms. Knowledge is only partly excludable so that the aggregate produc tion function for final goods exhibits increas ing returns to scale. This nonconvexity in the production set allows for steady-state growth in per capita income. Romer shows how mar ket power is necessary for the growth in knowledge to be a result of a response to mar ket incentives; without imperfect competition, total output is less than would be required in payment of all inputs according to their mar ginal productivities. From the perspective of the current article, the key result of his model is that the rate of growth of a particular econ omy depends directly on the degree to which it is integrated with other economies. Such integration allows access to a larger stock of human capital which, in turn, raises invest ment in knowledge or technological improve ment and boosts growth. Sokoloff (1989) offers support for the Romer model. Sokoloff utilizes 19th century United States county-level data to show that the introduction of water transportation (canal construction or river dredging) sparked a sharply higher rate of patenting in those coun ties adjacent to the transportation system. Presumably, such counties displayed a higher rate of economic growth as well. Clearly, one could argue that similar effects would be ex pected from the development or improvement of a highway transportation system. In a model that also admits the possibility of increasing returns to scale, and steady-state growth in per capita income, Barro (1989a) shows how the rate of economic growth can affected by the size of a government sector, larger government raises economic growth to the extent that it raises the marginal productiv ity of private capital but lowers economic growth to the extent that the associated higher rate of taxation discourages productive activ > 8T stock is congested and on the magnitude of the highway stock. The level of capital accumulation in a particular locale, in turn, is dependent upon the gap between the return to productive activity, r., and the economy-wide cost of capital which we denote p. Hence, we have ECONOMIC PERSPECTIVES ity. In a companion empirical paper, Barro (1989b) presents evidence that suggests that governments optimize in their choice of the size of the government sector relative to the economy. In particular, he finds that eco nomic growth is inversely related to “unpro ductive” government activity (such as govern ment consumptions spending) and weakly positively related to “ productive” government activity (such as nonmilitary public invest ment). D a ta and e m p iric a l resu lts In this article, I use data on real per capita income growth and measures of highway ca pacity and quality across the contiguous fortyeight states during the period 1960 to 1985. As the focus of the study is on the longer term relationship between the transportation infra structure and economic growth, the data on per capita income growth are sample averages of underlying annual observations. The basic highway capacity variable is measured as the total existing road mileage, inclusive of urban and rural roadway, in a given state relative to the square mileage of the state over the period 1960 to 1985. The separate importance of the urban and rural road systems to per capita income growth will also be investigated. In these data, urban refers to census places with a minimum popu lation of 5,000. The basic highway quality variable is the percent of highway mileage of deficient quality in 1982; such road surface carries a Present Serviceability Rating (PSR) of 2.5 or less for interstate highways and of 2.0 or less for other categories of roadway (other arterial and collector roads).6 In order to assess the degree to which the transportation system is congested, a highway usage variable must be employed. The vari able chosen for that purpose in this article is vehicle density, expressed as total vehicle registrations (cars, trucks, and motorcycles) per highway mile over the period 1960 to 1985. Of course, this measure of vehicle den sity will be inaccurate to the extent that ve hicles registered in a particular state are oper ated in other states. The basic relationship to be investigated is a linearized version of Equation (5): 6) Dy. = b0 + b,*y.(0) + b2*vd. + b3*hi. + b4*pqj + c*d, + e FEDERAL RESERVE BANK OF CHICAGO where Dy. = per capita income growth in state j; yj(0) = initial (1960) level of per capita in come (in logarithms), vd. = logarithm of ve hicle density; hi. = logarithm of highway capacity; pq. = pavement quality; and d. = dummy variables for the Northeast, Midwest, and West regions of the United States as de fined by the Census Bureau. As the primary focus of this article is on the relationship between highway capacity and economic activity, the above equation is esti mated without explicit consideration of the separate effects of vehicle density and pave ment quality. Table 1 contains results of esti mating this simpler equation by ordinary least squares (OLS) and weighted least squares (WLS) methods. Column 1 reports OLS re sults including all regional dummy variables. As is shown, there is a significant tendency for states’ economies to converge toward a com mon level of per capita income. Specifically, the coefficient estimate of -1.38 on initial income implies that a one-standard-deviation reduction in the initial level of the logarithm of per capita income results in a faster rate of income growth of .28 percentage point during the period 1960 to 1985. Notably, the central proposition of this article—that economies with a superior surface transportation infra structure will benefit through higher productiv ity and per capita income growth—achieves empirical confirmation. The coefficient esti mate of .22 on the highway capacity variable indicates that a one-standard-deviation in crease in the logarithm of highway capacity induces a . 13 of a percentage point increase in the growth rate of per capita income. The finding that the stock of highways is an important contributor to economic growth parallels the results of recent empirical re search by Romer (1989b). Romer focuses on the importance of human capital—measured by the level of literacy of the population—for economic growth across countries. In regres sions similar to those in Table 1, he finds a significant positive relationship between hu man capital and per capita output growth. He also finds that human capital is positively correlated with private investment in plant and equipment. According to the conceptual analysis above, a similar connection between highways and investment would be expected. The results in Column 1 of Table 1 indi cate that, apart from initial per capita income 17 TABLE 1 Per capita income growth and highway capacity (Dependent variable: Dy^ method 1 2 3 4 5 6 7 8 9 10 OLS OLS OLS OLS WLS WLS WLS WLS WLS WLS sq. rt. of y(0) sq. rt. of y(0) level of y(0) level of y(0) log of y(0) log of y(0) constant -6.53 1.53 -6.92 1.10 -6.94 1.45 -7.69 1.08 -7.18 1.15 -7.94 1.08 -7.47 1.20 -8.19 1.09 -6.84 1.09 -7.61 1.08 VjlO) -1.38 .25 -1.44 .19 -1.48 .24 -1.59 .18 -1.49 .20 -1.64 .19 -1.54 .21 -1.69 .19 -1.43 .18 -1.58 .18 hi. .22 .10 .26 .06 .27 .09 .30 .06 .26 .06 .30 .06 .25 .06 .31 .06 .26 .06 .30 .06 PQj __ __ -009 .004 -009 .003 -37 .11 -31 .08 mw -27 .11 ne <.01 .13 — -07 .12 — w -0.8 .15 — -11 .14 — R2 .61 .63 .66 SER .26 .25 .24 -25 .08 __ -010 .003 __ -011 .003 __ -008 .003 -26 .08 -32 .08 -26 .08 -33 .08 -25 .09 -31 .08 .67 .39 .49 .32 .46 .69 .73 .24 .26 .23 .26 .23 .25 .24 Variable definitions in appendix. and highway capacity, only the Midwest re gion has a growth rate of per capita income statistically different from that of the South, which is used as a benchmark in this Table. Column 2 reestimates the basic equation, drop ping the Northeast and West regional dummy variables. As was to be expected, ther adjusted coefficient of determination improves margin ally upon this alteration and only minor im pacts on the individual coefficient estimates can be discerned. Column 3 includes a measure of pavement quality in the regression equation to determine the separate effect of pavement quality on productivity and income growth. Here, a onepercentage-point erosion in pavement quality induces a reduction of per capita income growth equal to .009 of a percentage point per year. The point estimates of the coefficients on initial income and on highway capacity are left relatively undisturbed and, as before, the Northeast and West regional dummies are statistically insignificant. Column 4 elimi nates the latter dummy variables and exhibits nearly the same results for the remaining coef ficients. As estimation is being undertaken over a cross-section of states, there is some presump tion that the error structure may not be homoskedastic. Accordingly, Table 1 also con tains the results of various generalized leastsquares estimations using a variety of weight ing series. Columns 5 and 6 use the square root of initial per capita income as a weighting series; columns 7 and 8 use the level of initial per capita income; and columns 9 and 10 use the logarithm of initial per capita income. Only the results with the Midwest regional 18 ECONOMIC PERSPECTIVES ing the relationship between highway quantity and quality variables and per capita income growth by two-stage, least-squares methods. Instruments chosen for estimation are the ini tial 1960 stock of highway mileage, initial 1960 vehicle registrations, initial 1960 popula tion, new road mileage financed with federal aid highway funds during 1980, seasonal heat ing degree days, and the number of local gov ernmental units in 1982. The reasoning behind the choice of certain instruments, such as ini tial highway capacity, initial vehicle registra tions, and initial population require no expla nation. New road mileage financed through federal grants is taken as exogenous to individ ual states and is expected to be correlated with highway capacity and quality. Heating degree days is a measure of temperature extremes and is expected to be correlated with pavement quality. Finally, the extent to which a state’s governmental decision-making is concen trated, measured by the number of local gov ernmental units, arguably will affect its ability to collect and disburse funds for the purpose of dummy are presented; as in previous equa tions, the Northeast and West regional dum mies carried little “explanatory” power. In every case, the rate of growth of per capita income is significantly related to highway capacity and pavement quality; further, the quantitative values of the coefficient estimates remain within a small interval of the original unweighted estimates. Of course, one should be concerned about the potential for simultaneity bias in the esti mated coefficients contained in Table 1. For instance, it may be argued that a portion of the positive correlation between highway capacity and per capita income growth is simply due to the fact that high income growth states are likely to be states with adequate resources to invest in additional highways. Similarly, states with such resources would be in a posi tion to undertake appropriate maintenance expenditures in order to avoid an erosion of pavement quality over time. To address the possibility of such simulta neity bias, Table 2 exhibits results of estimat TABLE 2 Per capita income growth and highway quantity and quality (Dependent variable: Dy.) method 1 2 3 4 5 6 7 8 TSLS TSLS WTSLS WTSLS WTSLS WTSLS WTSLS WTSLS sq. rt. of y(0) sq. rt. of y(0) level of y(0) level of y(0) log of y(0) w eight log of y(0) constant -6.92 1.10 -8.28 1.30 -7.17 1.15 -8.53 1.27 -7.45 1.20 -8.70 1.24 -6.83 1.09 -8.18 1.31 Vj(0) -1.44 .19 -1.71 .23 -1.48 .20 -1.76 .23 -1.53 .21 -1.79 .22 -1.43 .18 -1.69 .23 .26 .06 .33 .07 .25 .06 .34 .07 .25 .06 .34 .07 .26 .06 .33 .07 hr PQ, __ -016 .008 — -018 .008 — -019 .007 — -015 .008 mw -25 .08 -36 .10 -25 .08 -37 .10 -26 .08 -38 .09 -25 .09 -35 .10 R2 .63 .64 .39 .42 .31 .38 .69 .70 SER .25 .25 .26 .25 .26 .25 .25 .25 Instrum ent list: y^O), hi(0), v(0), |3.(0), newhi.(0) , hddj( govu. FEDERAL RESERVE BANK OF CHICAGO 19 highway construction and maintenance. As before, only results from estimating with a dummy variable for the Midwest region are displayed; inclusion of other regional dummies does not affect the conclusions in any impor tant way. Column 1 of Table 2 shows that the basic relationship between highway capacity and economic growth is not reflective of a reverse causation from per capita income growth to highways. The point estimate of the effect of highways on economic growth remains the same as with ordinary least squares regression, and there is no change in the standard error associated with the coefficient on highway capacity. The results contained in Column 2 reflect an increase in the quantitative relation ship between pavement quality and economic growth, with a near doubling of the relevant coefficient estimate. However, the associated standard error increases by a large amount, with the result that the relationship between pavement quality and per capita income growth is of somewhat diminished statistical significance. Nevertheless, the negative rela tionship between deficient highway mileage and economic growth still remains at roughly the 5% significance level. Columns 3 and 4 repeat the estimation utilizing weighted twostage least squares, with the square root of initial per capita income as a weighting series; the point estimates are similar to those in Col umns 1 and 2 with some improvement in the statistical importance of pavement quality. Columns 5 and 6 make use of initial per capita income as a weighting series; the only discern ible difference in results is a further increase in the importance of the pavement quality vari able. Finally, Columns 7 and 8 use the loga rithm of initial per capita income to weight the observations; in this case, the statistical asso ciation between pavement quality and per capita income growth is attenuated and returns to that obtained in Column 2. Highway capacity may be acting as a proxy for some other variable that may be of direct and primary importance to economic growth. One such variable might be the de gree to which the economy of a state is geo graphically concentrated; perhaps highly ur banized states exhibit higher per capita income growth due to the compact nature of the par ticular state’s economy. Table 3 allows one to dismiss the validity of this particular argu 20 TABLE 3 Per capita growth and urbanization (Dependent variable: Dy.) 1 2 3 4 method OLS TSLS OLS TSLS constant -9.41 1.71 -13.27 2.96 -10.53 2.16 -12.71 2.92 VjlO) -1.84 .26 -2.44 .46 -1.86 .25 -2.17 .37 .31 .06 .38 .09 .31 .06 .37 .08 P4j -009 .003 -023 .010 -010 .003 -022 .009 urbj .004 .003 .012 .006 .307 .204 .433 .249 mw -28 .08 -34 .12 -29 .08 -37 .11 R2 .67 .53 .68 .58 SER .23 .28 .23 .27 hi. Instrum ent list: see Table 2. ment. As can be seen, urban density—meas ured by the raw percentage of total population living in standard statistical metropolitan areas in Columns 1 and 2 and by its natural loga rithm in Columns 3 and 4— is, at best, only marginally significant and does not attenuate the strength of the basic relationships between highway capacity, highway quality, and eco nomic growth. V e h ic le d en s ity and e co n o m ic g ro w th According to the discussion in the theo retical section, an economy with an overbur dened highway system—one with traffic den sity beyond the bottleneck level—will have lower traffic volume and, as a result, lower productivity and per capita income growth. Thus, if during the period under investigation there existed chronic underinvestment in high way capacity across states, one would expect to find a negative relationship between vehicle density—measured as the logarithm of vehicle registrations per highway mile—and per capita income growth. The results contained in Table 4 allow one to gauge the adequacy of the highway capital stock across states. ECONOMIC PERSPECTIVES TABLE 4 Adequacy of highway capital stock (Dependent variable: Dy.) 1 2 3 4 5 6 7 8 method OLS WLS WLS WTSLS TSLS WTSLS WTSLS WTSLS weight _ sq. rt. of y(0) level of y(0) log of y(0) sq. rt. of y(0) level of y(0) log of y(0) log of y(0) constant -8.00 .47 -8.30 1.45 -8.61 1.42 -7.90 1.48 -8.80 1.80 -9.03 1.71 -9.19 1.62 -8.70 1.83 V,(0) -1.63 .22 -1.68 .22 -1.74 .22 -1.62 .23 -1.78 .29 -1.82 .27 -1.85 .26 -1.76 .29 .28 .10 .27 .10 .27 .10 .28 .10 .30 .11 .30 .11 .30 .11 .30 .11 pqj -009 .003 -010 .003 -011 .003 -008 .003 -016 .008 -018 .008 -020 .007 -016 .008 vdj .027 .086 .033 .085 .038 .084 .026 .089 .041 .097 .043 .096 .044 .094 .040 .097 mw -29 .086 -30 .10 -30 .08 -29 .10 -33 .08 -34 .09 -35 .09 -33 .10 R2 .66 .47 .45 .72 .62 .40 .36 .69 SER .24 .24 .24 .24 .25 .25 .25 .25 hi. Instrum ent list: see Table 2. Upon scanning the results of Table 4, one finds no evidence of a chronic shortage of highway capacity across states over the entire period 1960 to 1985. The point estimate of the effect of higher vehicle density on per capita income growth is uniformly statistically insig nificant regardless of the method of estimation (ordinary least squares, weighted least squares, two-stage least squares, and weighted twostage least squares). Furthermore, the esti mated relationship between highway capacity and economic growth and that between pave ment quality and economic growth remain nearly the same as when the vehicle density variable was omitted from the basic empirical specification. U rb an versus ru ral h ig h w a y c a p ac ity A natural question is whether urban or rural roads are of greater quantitative and/or statistical importance in determining economic growth across states. Table 5 allows for a FEDERAL RESERVE BANK OF CHICAGO decomposition of the initial stock of highways into urban (SSMA) and rural (non-SSMA) mileage. The first Column of Table 5 indi cates that both the urban and rural components are quantitatively and statistically important determinants of economic growth, with rural roads having the larger effect. One should note that the diminished statistical significance of the relationship between highways and per capita income growth to a large degree is due to the collinearity between urban and rural highway mileage; the correlation between the two variables across states is .59. Indeed, dropping each of the rural and urban compo nents in turn—as in Columns 2 and 3—leaves significant importance for the remaining high way capacity measure, with individual point estimates of .17 (urban) and .40 (rural) and associated standard errors of .04 (urban) and .09 (rural). Column 4 combines the two com ponents of the highway stock by weighting 21 TABLE 5 Urban and rural highway capacity (Dependent variable: Dy.) 2 3 constant -7.35 1.09 -6.81 1.08 -7.80 1.120 -7.37 -7.53 1.04 1.06 -1.54 .20 -1.42 .19 -1.65 .19 -1.54 -1.61 .18 .18 hir. .24 .12 .40 .09 — — h ia .10 .05 — .17 .04 — h it — -- — .34 .07 — — -- < o 1 hita. — 4 5 — .26 .05 areaj -.27 .08 -35 .07 -13 .04 PPj -008 .003 -008 .003 -008 .003 mw -36 10 -42 .10 -25 .09 -36 .09 -31 .08 R2 .68 .66 .66 .69 68 SER .23 .24 .24 .23 23 -26 .05 — -008 -009 .003 .003 according to the coefficient estimates in Col umn 1 and then summing the two separate components. The coefficient estimate is highly statistically significant. Finally, Col umn 5 takes the highway capacity measure in Column 4 and normalizes by the surface area of the state. The coefficient estimate can be compared with that of Table 1, whereupon it is seen that this measure of highway capacity bears a stronger statistical association with per capita income growth than did the original, simpler measure. C onclusion This article develops a simple model in which the government sector of a particular jurisdiction can influence the rate of growth of output in that locale. A higher level and better quality of highway capacity expands transpor tation services and, in so doing, raises the marginal product of private capital. The higher marginal product of capital induces higher investment in physical capital and growing per capita incomes and output. Local governments can thereby exert an important influence on the rate of economic growth within their own locality. In future research, it would be interesting to expand on the theme of this article by look ing at the relationship between other measures of infrastructure—water and sewer systems, airports, mass transit, etc.—and local eco nomic growth. Along with existing results on the importance of public capital to metropoli tan production, such as contained in Eberts (1988), such evidence would give an improved indication of the importance of the services of government capital to the development and performance of state and local economies. APPENDIX Data description and sources Dy = average annual growth of per capita in come (1972$) from 1960 to 1985. SAUS, various issues. hiu = logarithm of total existing urban road mileage, average over 1960 to 1985. SAUS, various issues. hit = logarithm of weighted sum of hir and hiu. y = logarithm of level of per capita income (1972$). SAUS, various issues. area = logarithm of square miles of surface area. SAUS. p = logarithm of population, average over 1960 to 1985. SAUS, various issues. hita = hit-area. hi = logarithm of total existing road mileage, average over 1960 to 1985. SAUS, various issues. pq = percent of highway mileage of deficient quality in 1982 (PSR < or = 2.5 for interstate highways, PSR < or = 2.0 otherwise). HS 1982, Table HM63. hir = logarithm of total existing rural road mileage, average over 1960 to 1985. SAUS, various issues. 22 ECONOMIC PERSPECTIVES v = total vehicle registrations, average over 1960 to 1985. SAUS, various issues. vd = logarithm of vehicle registrations per highway mile, average over 1960 to 1985. SAUS, various issues. urb = percent of total population residing in standard metropolitan statistical areas in 1970. SAUS, 1977 Table 17. newhi = new road mileage financed with fed eral aid highway funds in 1980. HS 1980, Table FA1. hdd = seasonal heating degree days (60_ base). SAUS, 1982-83, Table 378. govu = number of local governmental units in 1982. SAUS, 1988, Table 452. FOOTNOTES 'For example, consider the following statement by Richard Bartel (1989): “ ...some economists tend to think of invest ment in narrow terms—private spending on business plant and equipment. We often forget about additions to the stock of public infrastructure—spending on roads, bridges, mass transportation, airports, waterways, water supply, waste disposal facilities, and other public utilities. a link between general infrastructure capital (inclusive of but not confined to highways), the rate o f return to private capital, and the level of private investment in nonresidential equipment and structures. . 2See McDonald and d’Ouville (1989). T h e U.S. Department of Transportation’s “ PSR is a numerical value between zero and five reflecting poor pavement condition at the lower end and very good pave ment condition at the higher values.” Highway Statistics (1982), p. 108). 3See McDonald and d’Ouville (1988). 4These conceptual results are consistent with the empirical results in Aschauer (1988) and (1989b), which established sFor related arguments, the reader is referred to Rostow (1960) and Rosenstein-Rodan (1961). REFERENCES Aschauer, David A., 1988, “ Government spending and the falling rate of profit,” Eco nomic Perspectives, Vol. 12, pp. 11-17. _____, 1989b, “ A cross-country study of growth, saving, and government,” unpub lished. _____, 1989a, “ Is government spending pro ductive?,” Journal of Monetary Economics, Vol. 23, pp. 177-200. Dowdrick, Steven and Duc-Tho Nyugen, 1989, “ OECD comparative economic growth 1950-85: catch-up and convergence,” Ameri can Economic Review, Vol. 79, pp. 10101030. _____, 1989b, “ Does public capital crowd out private capital?,” Journal of Monetary Economics, Vol. 24, pp. 171-188. _____, 1989c, “ Public investment and pro ductivity growth in the Group of Seven,” Economic Perspectives, Vol. 13, No. 5, pp. 17-25. Bartel, Richard D., 1989, “ Introduction,” Challenge, September/October, pp. 2-3. Barro, Robert J., 1989a, “ Government spending in a simple model of endogenous growth,” Rochester Center for Economic Re search, Working Paper, No. 186. FEDERAL RESERVE Eberts, Randall, 1988, “ Estimating the con tribution of public capital stock to metropoli tan manufacturing production,” Federal Re serve Bank of Cleveland, Working Paper. Fare, R., S. Grosskopf and B. Yoon, 1982, “A theoretical and empirical analysis of the highway speed-volume relationship,” Journal of Urban Economics, Vol. 12, pp. 115-121. Garcia-Mila, Teresa and Terese McGuire, 1987, “ The contribution of publicly provided inputs to states’ economies,” unpublished. Holz-Eakin, Douglas, 1989, “ A note on the infrastructure crisis,” unpublished. BANK OF CHICAGO 23 McDonald, John F. and Edmond L. d’Ouville, 1988, “ Highway traffic flow and the uneconomic region of production,” Re gional Science and Urban Economics, Vol. 18, pp. 503-510. _____, 1988, “Optimal road capacity and the uneconomic region of production,” unpub lished. Munnell, Alicia, 1990, “Why has productivity growth declined? Productivity and public investment,” AJew England Economic Review, January/February, pp. 3-22. Murphy, Kevin M., Andrei Shleifer, and Robert W. Vishny, 1989, “ Industrialization and the big push,” Journal of Political Econ omy, Vol. 97, pp. 1003-1026. Romer, Paul M., 1989a, “ Endogenous tech nological change,” National Bureau of Eco nomic Research, Working Paper, No. 3210. 24 _____, 1989b, Human capital and growth: theory and evidence, National Bureau of Eco nomic Research, Working Paper, No. 3173. Rostow, Walt W., 1960, The stages of eco nomic growth: A non-Communist manifesto, Cambridge University Press, London. Rosenstein-Rodan, 1961, “ Notes on the the ory of the big push,” in Economic develop ment for Latin America, Howard S. Ellis and Henry C. Wallich (eds.), St. Martin’s Press, New York. Sokoloff, Ken, 1989, “ Inventive activity in early industrial America: Evidence from patent records, 1790-1846,” Journal of Economic History, Vol. 48, pp. 813-850. Statistical Abstract of the United States, vari ous issues (U.S. Government Printing Office, Washington, D.C.). ECONOMIC PERSPECTIVES ECONOMIC PERSPECTIVES BULK RATE Public Information Center Federal Reserve Bank of Chicago P.O. Box 834 Chicago, Illinois 60690-0834 U.S. POSTAGE PAID CHICAGO, ILLINOIS PERMIT NO. 1942 D o N o t F orw ard A d d ress C o rrec tio n Requested R eturn P o stag e G u a ra n te ed FEDERAL RESERVE BAN K OF CHICAGO
@FedFRASER