The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Federal Reserve Bank of Chicago Second Quarter 2002 persp ives 2 The center restored: Chicago’s residential price gradient reemerges 12 Location trends of large company headquarters during the 1990s 27 Post-resolution treatment of depositors at failed banks: Implications for the severity of banking crises, systemic risk, and too big to fail 42 Following the yellow brick road: How the United States adopted the gold standard Economic perspectives President Michael H. Moskow Senior Vice President and Director of Research William C. Hunter Research Department Financial Studies Douglas Evanoff, Vice President Macroeconomic Policy Charles Evans, Vice President Microeconomic Policy Daniel Sullivan, Vice President Regional Programs William A. Testa, Vice President Economics Editor David Marshall Editor Helen O’D. Koshy Associate Editor Kathryn Moran Production Julia Baker, Rita Molloy, Yvonne Peeples, Nancy Wellman 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 Federal Reserve Bank of Chicago or the Federal Reserve System. Single-copy subscriptions are available free of charge. Please send requests for single- and multiple-copy subscriptions, back issues, and address changes to the Public Information Center, Federal Reserve Bank of Chicago, P.O. Box 834, Chicago, Illinois 60690-0834, telephone 312-322-5111 or fax 312-322-5515. Economic Perspectives and other Bank publications are available on the World Wide Web at http:Avww.chicagofed.org. Articles may be reprinted provided the source is credited and the Public Information Center is sent a copy of the published material. Citations should include the following information: author, year, title of article, Federal Reserve Bank of Chicago, Economic Perspectives, quarter, and page numbers. s? chicagofed, org ISSN 0164-0682 Contents Second Quarter 2002, Volume XXVI, Issue 2 The center restored: Chicago’s residential price gradient reemerges Daniel P. McMillen After a long period during which house prices were not affected by distance from Chicago’s central business district, values now decline by more than 8 percent per mile. Annual appreciation rates in house prices are higher in neighborhoods close to the city center with large minority populations, high concentrations of poverty, and many vacant homes in 1990. Location trends of large company headquarters during the 1990s Thomas Klier and William Testa This article documents changes in the spatial distribution of corporate headquarters of large U.S.-domiciled corporations during the 1990s. The authors find that the largest metropolitan areas continue to host a disproportionate share of headquarters, but there have been significant shifts toward cities with population between one and two million. Post-resolution treatment of depositors at failed banks: Implications for the severity of banking crises, systemic risk, and too big to fail George G. Kaufman and Steven A. Seelig Losses from bank failures have significant adverse implications for bank stakeholders, as well as for the macroeconomy. This article examines the potential sources of such losses, in particular the losses that may occur after the date a bank is failed, and makes recommendations on how to minimize these losses. Following the yellow brick road: How the United States adopted the gold standard Francois R. Velde The United States, with some difficulty, adopted the gold standard in the late nineteenth century, thus pegging the dollar to the pound sterling and other currencies. Some have argued it was a mistake, others that it was inevitable. This article recounts the historical background and uses a model to shed light on the choices faced by policymakers of the time. The center restored: Chicagos residential price gradient reemerges Daniel P. McMillen Introduction and summary Income growth and the development of new methods of transportation have made the decentralization of American cities a long-standing and ongoing process. Higher income raises the demand for land and housing, which typically are less expensive farther from the city center. The development of horse car lines, subways and elevated train lines, and most importantly, the automobile and highway system facilitated the growth of more remote locations by making long commutes feasible. The trend toward residential decentralization is reinforced by employment decentralization. Suburban locations offer firms low land costs, ready access to interstate highways, and the availability of a skilled labor force of nearby residents. In many metropolitan areas, the traditional city center retains a strong job core in spite of the trend toward decentralization. Central business districts tend to specialize in high-skill, high-wage service jobs. Such jobs attract young professionals who enjoy city living and do not want to incur the long commute required from the suburbs. Re-gentrification of neighborhoods near the city center may take place as older housing is converted to modern condos and apartments to serve these households. The subsequent rise in housing prices provides cities with much-needed new revenue from property taxes. In this article, I document the restoration of Chicagos city center from 1983 to 1998. Using a sample of single-family homes that each sold at least twice during the sample period, I find that prices rose far more rapidly near the city center than at the edge of the Chicago city limits. In the early 1980s, house prices increased with distance from the city center. In contrast, house prices declined by nearly 8 percent with each additional mile from the city center by the end of the 1990s. 2 The rapid growth of house prices in the city center has costs as well as benefits. Existing residents may find themselves forced to move when they can no longer afford what now are prime locations, and those who remain may not like the new character of the neighborhood. New residents may demand better provision of costly services, such as schools and police protection. Secondary effects will occur in other neighborhoods as displaced former residents move elsewhere. Nevertheless, the overall effect of a resurgent central city housing market is likely to be positive. Increased property tax revenues can more than pay for the new services, generating a surplus than can be used elsewhere in the city. New households attract stores and restaurants that in turn attract more residents. Just as urban decay can generate a flight to the suburbs, urban revitalization can generate additional growth that benefits the entire city. Historical trends in Chicago Chicago was a highly centralized city at the beginning of the twentieth century. Figure 1 shows that the City of Chicago then accounted for 81.5 percent of the population of the six-county region that today defines the Chicago metropolitan area.1 Chicagos population peaked at 3.6 million in 1950, when it accounted for 69.9 percent of the metropolitan areas residents. The citys population then fell steadily up until 1990, while the rest of Cook County and the five collar counties grew rapidly. In 1998, Chicagos 2,802,079 residents accounted for 36.0 percent of the metropolitan areas residents, while the rest of Cook County and the collar counties accounted for 30.7 Daniel P. McMillen is a professor of economics at the University of Illinois at Chicago and a consultant to the Federal Reserve Bank of Chicago. 2Q/2002, Economic Perspectives 31, 1998. The data are obtained from tax records and reflect actual transaction Chicago regional population prices. In order to construct an index that 19902020 controls for housing quality, I restrict the sample to repeat sales. When houses are 10,000,000 not remodeled between sales, the average Collar change in prices provides an estimate of a Suburban Cook 8,000,000 constant-quality house price index. ObserChicago vations are not included in the final sam6,000,000 ple if the building size, lot size, or the number of stories changes between sales 4,000,000 dates. I also discarded a small number of observations without addresses or sales dates. The final sample includes 2,000,000 52,972 transactions. The Chicago metropolitan area is 0 1900 '10 '20 '30 '40 '50 '60 '70 '80 '90 '98 '10 '20 somewhat unusual in that very little inforSource: Northeastern Illinois Planning Commission. mation is available on house sales other than price and location. Although the City of Chicago collects information on percent and 33.2 percent, respectively. However, the lot size, building area, age, and the number of stories, 1990s saw a reversal of the City of Chicagos 40-year no information is available on such common varidecline in population, as its population increased by ables as the number of rooms or the presence of air 18,353, or 0.6 percent. The Northeastern Illinois Planning Commission expects the FIGURE 2 trend to continue, with Chicagos populaChicago community area population change tion rising to 3,007,025 in 2020. 19902000 The distribution of population is also changing within the City of Chicago. Figure 2 shows the growth rates in population between 1990 and 2000 across the 77 community areas that comprise the city. Community areas on the Far South Side lost population over the decade. In contrast, the city center grew rapidly. The Loop added 4,434 residents, which is a growth rate of 37.1 percent. The Near North Side grew from a population of 62,842 to 72,811, or 15.9 percent. The Near South Side had a growth rate of 39.3 percent, adding 2,681 residents over the decade. The growth near the city center is significant because it reverses many years of decline, and it has Percentage change –100.0 to –10.0 occurred in some of the most expensive –10.0 to –4.5 areas of the city. –4.5 to –1.7 FIGURE 1 Data To analyze trends in housing prices in the City of Chicago, I use a data set that includes all transactions of single-family homes that sold at least twice during the period from January 1, 1983, to December Federal Reserve Bank of Chicago –1.7 to 0.4 0.4 to 5.2 5.2 to 14.0 14.0 to 20.0 20.0 to 100.0 0 1 2 3 Miles Source: U.S. Census of Population and Housing, 2000. 3 TABLE 1 Descriptive statistics, house sales 0–18 miles 0–6 miles 6–9 miles 9–12 miles 12–18 miles 106.983 (105.901) 136.343 (191.882) 97.411 (55.544) 109.558 (109.216) 95.785 (61.940) Log sales price 11.404 (0.582) 11.404 (0.859) 11.363 (0.510) 11.487 (0.487) 11.290 (0.621) Distance from CBD (miles) 8.759 (2.561) 4.718 (1.019) 7.678 (0.793) 10.229 (0.801) 13.296 (0.955) Lot size (square feet) 4,158.77 (1,500.40) 3,345.72 (1,698.24) 3,958.15 (901.86) 4,469.70 (1,546.43) 4,963.86 (2,061.32) Building area (square feet) 1,208.37 (658.51) 1,276.34 (674.09) 1,205.27 (395.94) 1,189.92 (915.16) 1,189.40 (382.15) 6.121 (2.326) 8.807 (2.192) 6.519 (1.864) 4.811 (1.767) 5.290 (2.100) More than one story (%) 16.58 21.66 13.92 17.51 16.75 Number of observations 52,972 7,572 21,248 18,264 5,888 Sales price ($1,000) Age (years in tens) Notes: Means are followed by standard deviations in parentheses for the continuous variables. CBD is central business district. conditioning.2 Table 1 presents descriptive statistics for the available variables. Over the full sample, sales prices average $106,983 over 198398. Average prices are much higher for houses near the city center: The average price for houses located within six miles of the city center is $136,343, compared with $95,785 for houses located more than 12 miles from the center. Lot sizes are smaller near the city center, with average lots of 3,345.72 square feet within six miles of the center versus 4,963.86 square feet in the most distant areas of the city. Building areas do not differ much across locations, but the housing stock is much older on average near the city center (88 years, compared with 61 years for all houses in the sample). The traditional center of Chicago is the intersection of State and Madison streets in the Loop. For the sample of house sales, distances range from 0.27 to 16.73 miles, with an average of 8.76 miles. Estimated house price indexes Figure 3, panel A plots the averages of the natural logarithms of house sales prices, calculated separately for each quarter from 1983 to 1998. I calculated separate price indexes for four intervals of distance from the Chicago city center, 06 miles, 69 miles, 912 miles, and 1218 miles. During the early 1980s, average 4 house prices were much lower for the 0-6 mile interval than for any other interval. Though average prices rose over time for all distance intervals, the rate of appreciation was much more rapid in the interval closest to the city center. By the end of the 1990s, average prices were much higher in the area surrounding the city center than in any of the other intervals. Although figure 3, panel A shows a clear tendency toward the return of Chicagos center, simple averages are not the best way to construct price indexes. The composition of house sales may change systematically over the business cycle and by location. For instance, it is possible that only expensive homes remain in demand near the city center when the economy slows, which would tend to overstate the rate of price appreciation near the city center during economic downturns. Such changes in housing composition violate the spirit of a house price index, which is supposed to represent the rate of price appreciation for homes whose quality is not changing over time. A better measure then is a constant-quality price index. Constant-quality price indexes Two econometric methods are commonly used to construct constant-quality price indexes. The hedonic approach is based on a straightforward regression of house sales prices on housing characteristics, which 2Q/2002, Economic Perspectives include location and the date of sale in addition to standard house features such as living area. The hedonic price index is simply the set of predicted house prices for a house with given characteristics, constructed at varying target dates.3 The other common method for constructing constant-quality price indexes is the repeat sales method.4 The repeat sales approach is less vulnerable to missing variable bias than the hedonic approach because it estimates the rate of price appreciation from houses that sell at least twice during a sample period. If houses have not been remodeled between sales, then the change in prices across sales dates provides a measure of the rate of appreciation that is not contaminated by the effects of unobserved housing characteristics. Details on these estimation procedures are provided in the appendix. By confining the sample to houses that sell at least twice during the sample period, the repeat sales estimator ignores the information provided by homes that sell only once. In addition to this potential inefficiency, the repeat sales estimator may be subject to sample selection bias if the sample of repeat sales homes differs systematically from the overall housing market. These problems are not likely to be serious in our sample, which covers a long period. In an active housing market, the set of houses that sell at least twice over 16 years is not likely to differ much from the overall stock of houses in the city. Figure 3, panel B shows price indexes constructed by the hedonic method for various distance intervals. The representative house is 60 years old, has a single story, 1,200 square feet of living space, and a 4,200 square foot lot. The results are quite similar to the simple averages shown in panel A. Prices start out lowest in the interval closest to the city center, but this area has the most rapid rate of price appreciation over time, so that it has the highest prices at the end of the 1990s. Prices appreciated least rapidly in the most distant region, which is 1218 miles from the city center. For the full sample of 018 miles, figure 3, panel C compares the price indexes calculated using the hedonic and repeat sales approaches. The indexes are similar, although the repeat sales estimator shows a slightly lower overall rate of price appreciation. Estimated city-center gradients Although dividing the sample into four distance intervals provides a useful illustration of the effects of distance from the city center on house-price appreciation rates, it is based on the unrealistic assumption that prices change discretely across intervals while remaining constant within them. A more conventional approach is to use distance from the city center as an Federal Reserve Bank of Chicago FIGURE 3 Chicago house price indexes A. Average quarterly sales prices log sales price 12.25 11.75 11.25 0–6 miles 6–9 miles 9–12 miles 12–18 miles 10.75 10.25 1983 ‘85 ‘87 ‘89 ‘91 ‘93 ‘95 ‘97 year of sale B. Hedonic sales price indexes log sales price 12.25 11.75 11.25 0–6 miles 6–9 miles 9–12 miles 12–18 miles 10.75 10.25 1983 ‘85 ‘87 ‘89 ‘91 ‘93 ‘95 ‘97 year of sale C. Hedonic and repeat sales price indexes index for log sales price 2.00 1.50 1.00 Repeat sales Hedonic 0.50 0.00 -0.50 1983 ‘85 ‘87 ‘89 ‘91 ‘93 ‘95 ‘97 year of sale 5 FIGURE 4 City center price gradients A. Hedonic city center gradients: 95% confidence intervals log sales price 0.050 0.025 Upper bound 0.000 -0.025 Lower bound -0.050 -0.075 Estimated CBD gradient -0.100 1983 ‘85 ‘87 ‘89 ‘91 ‘93 ‘95 ‘97 year of sale B. Hedonic and repeat sale city center gradient indexes log sales price 0.025 0.000 Repeat sales -0.025 -0.050 -0.075 Hedonic become significantly negative by the beginning of the 1990s. By 1998, house prices are estimated to decline by more than 7 percent with each mile of distance from the city center. Figure 4, panel B presents the estimated index of repeat sales city-center gradients, along with the implied hedonic index, which is calculated by subtracting the estimated first-quarter gradient from the hedonic index. The repeat sales index shows a less rapid decline in the gradient because missing variables that help produce the sharp rise in house prices near the city center are correlated with distance to the city center, leading to a downward bias in the hedonic gradient term. Nonetheless, the repeat sales index also shows a significant decline in the gradient over time as areas near the city center regain their popularity and increase sharply in price. Figure 5 provides an alternative view of the sequence of events. The hedonic estimates are used to generate predictions for four datesthe second quarters of 1983, 1988, 1993, and 1998at varying distances from the city center. The representative home again is 60 years old, with a single story, 1,200 square feet of living space, and a 4,200 square foot lot. In 1983, house prices rose with distance from the center. House prices are not affected significantly by distance from the center in 1988. By 1993, prices are much higher near the center than in distant locations. Prices simply appreciate in all locations between 1993 and 1998, maintaining the city center premium. Over the full 198398 period, prices do not increase significantly in the most distant locations, whereas they rise dramatically in the city center. -0.100 -0.125 1983 ‘85 FIGURE 5 ‘87 ‘89 ‘91 ‘93 ‘95 ‘97 year of sale Note: CBD is central business district. Shift over time in the sales price functions log sales price 12.50 1998 explanatory variable in the hedonic price function. The coefficient for this variablethe city-center gradientrepresents the rate at which prices change with each additional mile of distance from the city center. The distance variable can also be interacted with the explanatory variables for the repeat sales model to form an index of time-varying city-center gradients. The estimated hedonic city-center gradients and the associated 95 percent confidence intervals are presented in figure 4, panel A. The figure clearly illustrates the return of centralization to the City of Chicago. In the early 1980s, house prices increased with distance from the city center by a rate of about 2 percent per mile. The gradient fell throughout the 1980s and had 6 12.00 1993 11.50 1988 11.00 1983 10.50 0.0 2.5 5.0 7.5 10.0 12.5 15.0 distance from the CBD Note: CBD is central business district. 2Q/2002, Economic Perspectives Census tracts Confining the effects of location to discrete intervals or a single variable representing distance from the city center may obscure variation in appreciation rates across small geographic areas. Prices may move together in some city neighborhoods while they diverge greatly in others. It is difficult to estimate accurate price indexes for small tracts because some areas occasionally have only a few sales. However, McMillen and Dombrow (2001) show that price indexes can be estimated accurately for small samples when prices change smoothly over time. They use a Fourier expansion to estimate the time trend in house prices. In the remainder of this section, I use McMillen and Dombrows approach to estimate the rate of increase in house prices from 1990 to 1996 for 851 census tracts in the City of Chicago. I use a nonparametric estimator that uses the standard repeat sales estimator as its base. The estimator places more weight on nearby observations when constructing an estimate for a given geographic location. The target geographic locations are the midpoints of the 851 census tracts that are represented in the sample of repeat sales. All observations are used in constructing the estimated FIGURE 6 Annual house price growth rates 199096 Percentage change 0.000 to 3.690 3.690 to 4.340 4.340 to 4.727 4.727 to 5.570 5.570 to 6.475 6.475 to 7.086 7.086 to 7.593 7.593 to 10.000 0 1 2 3 Miles Federal Reserve Bank of Chicago price appreciation rate for a census tract, but the estimator places more weight on house sales from the target tract. The estimated price appreciation rates are illustrated in figure 6. As with previous results, the results in figure 6 show that housing prices grew much more rapidly near the city center than in more distant locations. Appreciation rates do not decline uniformly with distance, however. Growth rates do not decline as rapidly on the Near North Side as in locations that are comparable distances from the city center on the South and West sides of the city. The Far South Side has higher appreciation rates than comparable locations on the North Side. The Englewood area on the South Side is a pocket of no growth in the midst of moderate appreciation rates. Calculating the appreciation rates for 199096 allows us to match the housing data with data from the 1990 U.S. Census of Population and Housing to explain differences in appreciation rates across census tracts. Table 2 presents the regression results, along with descriptive statistics for the explanatory variables. The regression results imply that growth rates decline by .401 percentage points with each additional mile from the city center. House-price growth rates increase by 0.0392 percentage points when the percentage of African-American residents in a tract rises by 10 percentage points. An increase in the percentage of Hispanic residents has a larger effect on growth rates: An increase of 10 percentage points in the number of Hispanic residents increases growth rates by 0.0876 percentage points. Interestingly, Census tracts bordering Lake Michigan and tracts with high median incomes do not have higher appreciation rates than other tracts. Another striking result is that census tracts with high poverty rates and a lot of vacant housing in 1990 have high appreciation rates: Growth rates rise by 0.1052 percentage points when the percentage of households that are in poverty rises by 10 percentage points, and they rise by 0.1781 percentage points when there is a similar increase in the amount of vacant housing in the census tract. Census tracts with older housing do not have lower growth rates, but increasing the amount of housing that is owner occupied by 10 percentage points adds 0.1053 percentage points to appreciation rates. 7 TABLE 2 House price growth rate regressions, census tracts Descriptive statistics Mean Constant Distance from city center Census tract borders lake African-American Hispanic High-school dropout Completed college Median income ($10,000) Poverty Vacant Owner-occupied Median house age (years) 6.370 0.048 0.424 0.193 0.226 0.116 2.475 0.249 0.105 0.366 44.121 Standard 3.091 0.214 0.444 0.263 0.103 0.142 1.133 0.200 0.083 0.244 9.651 Regression Coefficient 7.389* –0.401* –0.118 0.392* 0.876* 0.536 1.289* –0.005 1.052* 1.781* 1.053* –0.014* Standard error 0.329 0.014 0.152 0.117 0.175 0.413 0.422 0.060 0.284 0.465 0.256 0.003 Notes: The dependent variable is the estimated average growth rate in house sales prices between 1990 and 1996. The mean of the dependent variable is 5.620 and the standard deviation is 1.574. There are 851 observations. The R2 for the regression is 0.855. An asterisk indicates statistical significance at the 5 percent level. Figure 6 and table 2 suggest that a process of gentrification is underway in the City of Chicago. Census tracts closer to the city center that had a higher percentage of vacant housing, high poverty rates, and high percentages of African-American or Hispanic households experienced higher appreciation in house prices than other locations in the 1990s. These census tracts are the same ones that the 2000 Census shows have had significant increases in population, and much of this population increase is accounted for by higherincome, white households. These areas are served by public transportation and are close to the center of Chicagos central business district. New, expensive housing is being built for young professionals who formerly were moving to the suburbs. Conclusion This article presents strong evidence of the return of centralization to the City of Chicago. Growth in suburban employment caused Chicagos central business 8 district to decline in importance steadily until the 1980s. By 1990, the city center was enjoying renewed employment growth. Partly due to this growth, high-priced housing returned to locations near the city center. Although house prices increased slowly in census tracts near the city limits, prices rose very rapidly near the city center. By the end of the 1990s, the traditional negative house-price gradient had been restored. House values are estimated to decline by more than 8 percent with each mile of distance from the city center. It is too early to judge whether this trend will continue. The majority of the Chicago metropolitan areas jobs are now in the suburbs. Furthermore, the city continues to suffer from poor schools and other social problems. But employment growth in the central business area, the presence of numerous million-dollar homes, the destruction or conversion of housing projects near the city center, and the growing importance of households with two central-city workers suggest strongly that the inner city is enjoying a resurgence. 2Q/2002, Economic Perspectives APPENDIX: ESTIMATION PROCEDURES FOR CONSTANT-QUALITY PRICE INDEXES Let Vit represent the sales price of house i at time t, and let yit = log(Vit). The following regression equation is the basis for the hedonic house price index: 1) yit = dt + Xitb + uit. In equation 1, Xit is a vector of housing characteristics, uit is an error term, and dt and b are parameters to be estimated. In our application, Xit includes the natural logarithms of lot size and building area, the age of the house, a dummy variable indicating that the house has more than one story, and distance from the city center. The estimates of dt are the coefficients for a series of dummy variables indicating the quarter during which a house is sold. The estimated coefficient will be biased if unobserved housing characteristics are correlated with the error term. The repeat sales estimator avoids this bias by analyzing differences in sales prices of houses that sell at least twice during the sample period. If the coefficients for the housing characteristics do not change over time, the estimating equation for the repeat sales estimator is 2) yit yis = dt ds + uit uis. In equation 2, s represents the date of a houses earlier sale. The base coefficient, d0, is normalized to zero because it is not identified. McMillen and Dombrow (2001) generalize the standard repeat sales estimator by using a smooth continuous function g(T) to represent the time trend in house prices, where T represents the date of sale. The estimator is written as: 3) yit yis = g(Ti) g(Tis) + uit uis. In equation 3, Ti is the day of sale for house i and Tis is its previous sale date. Following Gallant (1981, 1982), McMillen and Dombrow (2001) use a Fourier expansion to model g(Ti) and g(Tis). The first step in the Fourier expansion is to transform the time variable to lie between 0 and 2p. The transformed variables are zi = 2pTi / max(T) and zis = 2pTis/max(T). The Fourier expansions are g(Ti) = a0 + a1zi + a2zi2 + Sq(lqsin(qzi) + gqcos(qzi)) and g(Tis) = a0s + a1szis + a2szis2 + Sq(lqssin(qzis) + gqscos(qzis)), where q = 1, , Q. The restriction that g(Ti) and g(Tis) are the same underlying function is imposed by setting a1 = a1s, l1 =ÿl1s, and so on. These restrictions imply: 4) yit yis = a1(zi zis) + a2(zi2 zis2) + Sq[lq(sin(qzi) sin(qzis)) +ÿgq(cos(qzi) cos(qzis))] + uit uis. Federal Reserve Bank of Chicago By convention, the price index is normalized to zero in the first period. Imposing a similar constraint on the Fourier index implies g(0) = 0, which implies a0 + g1 + + gQ = 0. The estimated price index can then be constructed from ordinary least squares (OLS) estimates of equation 4 as a1z + a2z2 + Sq(lqsin(qz) + gq(cos(qz) 1)), where z is a set of target dates. The standard repeat sales estimator and McMillen and Dombrows extension rely on an assumption that a single regression is adequate for an entire city. Nonparametric estimation allows for local geographic variation in house price appreciation rates. The nonparametric estimator used here was proposed by Cleveland and Devlin (1988), and is referred to as locally weighted regression (LWR). In constructing an LWR estimate for a given location, more weight is placed on nearby house sales than on distant sales. Let di be the distance between observation i and the target location for the price index. LWR uses a window of nearby observations to estimate the regression: The nearest b observations are given weights that decline with distance, whereas more distant observations receive no weight. In the empirical application, I set b equal to 10 percent of the total sample size. Let d(b) represent the distance of the most distant observation receiving weight in estimation. Following common practice, I use the tri-cube function: 3 ¨ ¥ d ´3· 5) wi ©1 ¦ i µ ¸ I di b d (b) , ©ª § d (b) ¶ ¸¹ where I() is an indicator function that equals one when the condition is true and zero otherwise. The weights fall smoothly from a maximum of one at the target location to zero at distance d(b). The LWR estimate at the target location is simply the predicted value from the weighted least squares regression. Letting yi represent the dependent variable and xi the vector of explanatory variables, the LWR prediction is: ¤ 1 ¤ ¥ n ´ ¥ n ´ wi xi xi 'µ ¦ wi xi yi µ . § i 1 ¶ § i 1 ¶ 6) yˆi xi ' ¦ The target site can be any arbitrary location. Each site will have a unique set of coefficient estimates, which implies a complete price index for the repeat sales estimator. The estimator varies smoothly over space, so estimated price indexes will be similar for nearby sites. However, estimates can differ significantly across more distant locations. 9 The centers of 851 census tracts within the city limits of Chicago are the target points for estimation, leading to 851 separate weighted least squares regressions. I construct price index estimates for each census tract for each day between 1983 and 1998. To summarize the estimated price indexes, I calculate the estimated index for January 1, 1990, and January 1, 1996, and solve for the implied yearly growth rate in prices. The former date corresponds to the 1990 Census, while the latter date is chosen to reduce the potential sensitivity of the estimates to small numbers of observations at the end of the sample period. NOTES 1 The six counties in Illinois are Cook, DuPage, Lake, Kane, McHenry, and Will. 2 The situation is worse in the suburbs, which do not collect information on lot size. 3 Examples of the hedonic approach include Bryan and Colwell (1982), Kiel and Zabel (1997), Mark and Goldberg (1984), Palmquist (1980), and Thibodeau (1989). The repeat sales house price index was first proposed by Bailey, Muth, and Nourse (1963). Applications include Abraham and Schauman (1991), Case, Pollakowski, and Wachter (1997), Case and Quigley (1991), Case and Shiller (1987, 1989), Clapp and Giaccotto (1998), Follain and Calhoun (1997), Gatzlaff and Haurin (1997), Geltner and Goetzmann (2000), Goetzmann and Spiegel (1997), Hill, Knight, and Sirmans (1997), Kiel and Zabel (1997), McMillen and Dombrow (2001), and Stephens et al. (1995). 4 REFERENCES Abraham, J. M., and W. Schauman, 1991, New evidence on house prices from Fannie Mae repeat sales, AREUEA Journal, Vol. 19, pp. 333352. Bailey, M. J., R. F. Muth, and H. O. Nourse, 1963, A regression method for real estate price index construction, Journal of the American Statistical Association, Vol. 58, pp. 933942. Bryan, T. B., and P. F. Colwell, 1982, Housing price indexes, in Research in Real Estate, C. F. Sirmans (ed.), Greenwich, CT: Jai Press. Case, B., H. O. Pollakowski, and S. M. Wachter, 1997, Frequency of transaction and house price modeling, Journal of Real Estate Finance and Economics, Vol. 14, pp. 173187. Case, B., and J. Quigley, 1991, The dynamics of real estate prices, Review of Economics and Statistics, Vol. 73, pp. 5058. Case, K. E., and R. J. Shiller, 1989, The efficiency of the market for single-family homes, American Economic Review, Vol. 79, pp. 125137. , 1987, Prices of single-family homes since 1970: New indexes for four cities, New England Economic Review, pp. 4556. Clapp, J. M. and C. Giaccotto, 1998, Residential hedonic models: A rational expectations approach to age effects, Journal of Urban Economics, Vol. 44, pp. 415437. 10 Clapp, J. M., C. Giaccotto, and D. Tirtiroglu, 1991, Housing price indices: Based on all transactions compared to repeat subsamples, AREUEA Journal, Vol. 19, pp. 270285. Cleveland, W. S., and S. J. Devlin, 1988, Locally weighted regression: An approach to regression analysis by local fitting, Journal of the American Statistical Association, Vol. 83, pp. 596610. Follain, J. R., and C. A. Calhoun, 1997, Constructing indices of the price of multifamily properties using the 1991 Residential Finance Survey, Journal of Real Estate Finance and Economics, Vol. 14, pp. 235255. Gallant, A. R., 1982, Unbiased determination of production technologies, Journal of Econometrics, Vol. 20, pp. 285323. , 1981, On the bias in flexible functional forms and an essentially unbiased form: The Fourier flexible form, Journal of Econometrics, Vol. 15, pp. 211245. Gatzlaff, D. H., and D. R. Haurin, 1997, Sample selection bias and repeat-sales index estimates, Journal of Real Estate Finance and Economics, Vol. 14, pp. 3350. Geltner, D. and W. Goetzmann, 2000, Two decades of commercial property returns: A repeated-measures regression-based version of the NCREIF Index, Journal of Real Estate Finance and Economics, Vol. 21, pp. 521. 2Q/2002, Economic Perspectives Goetzmann, W. N., and M. Spiegel, 1997, A spatial model of housing returns and neighborhood substitutability, Journal of Real Estate Finance and Economics, Vol. 14, pp. 1131. Hill, R. C., J. R. Knight, and C. F. Sirmans, 1997, Estimating capital asset price indexes, Review of Economics and Statistics, Vol. 79, pp. 226233. Kiel, K. A., and J. E. Zabel, 1997, Evaluating the usefulness of the American Housing Survey for creating housing price indices, Journal of Real Estate Finance and Economics, Vol. 14, pp. 189202. Mark, J. H., and M. A. Goldberg, 1984, Alternative housing price indices: An evaluation, AREUEA Journal, Vol. 12, pp. 3049. Federal Reserve Bank of Chicago McMillen, D. P., and J. Dombrow, 2001, A flexible Fourier approach to repeat sales price indexes, Real Estate Economics, Vol. 29, pp. 207226. Palmquist, R. B., 1980, Alternative techniques for developing real estate price indexes, Review of Economics and Statistics, Vol. 66, pp. 394404. Stephens, W., Y. Li, V. Lekkas, J. Abraham, C. Calhoun, and T. Kimner, 1995, Conventional mortgage home price index, Journal of Housing Research, Vol. 6, pp. 389418. Thibodeau, T. G., 1989, Housing price indexes from the 197483 SMSA Annual Housing Surveys, AREUEA Journal, Vol. 17, pp. 110117. 11 Location trends of large company headquarters during the 1990s Thomas Klier and William Testa Metropolitan areas highly value the presence of company headquarters, and local governments tend to actively pursue and attract them. The keen competition among Chicago, Dallas-Ft. Worth, and Denver in April and May 2001 in the wake of Boeings announcement that it would relocate its headquarters from Seattle highlighted the perceived benefits, including prestige, that the presence of a well-known company can confer on a metropolitan area. Of course, there are also tangible benefits. Headquarters employ a sizable and highly skilled white-collar work force and generate local demand for numerous specialized business services such as accounting and legal. In addition, headquarters often play a major role in corporate giving, as well as what are generally referred to as corporate citizen activities (Schwartz, 1997). It is not unusual to find that the landscape of a town has been defined by the presence of one or more corporate headquarters. For example, Columbus, Indiana, is dominated by public buildings designed by noted architects, courtesy of Cummins Engine and other local donors. Similarly, Eli Lilly, headquartered in Indianapolis, supports numerous local charities and public programs through the Lilly Endowment. In this article, we provide information on recent locational trends for company headquarters, which will be helpful to policymakers as they design development efforts and expenditures. We document changes in the spatial distribution of corporate headquarters of large U.S. domiciled corporations during the most recent decade. In order to perform this analysis, we use a comprehensive set of data on publicly traded companiesspecifically companies employing more than 2,500 people worldwide. We allocate headquarters to the 50 most populous metropolitan areas for 1990 and 2000 and examine the spatial changes that have taken place across 1) individual metro areas, 2) U.S. Census regions, and 3) the distribution of metro areas with respect to their population size. To identify and 12 disentangle spatial changes, we further examine the sources and nature of headquarters growth across metropolitan areas using both simple data displays and multiple regression analysis. The regression analysis allows us to distinguish among competing factors in their influence on the location of headquarters. Because policymakers are interested in attracting footloose headquarters, and perhaps nurturing small local companies as they grow to become large ones, we also document the extent and nature of headquarters turnover or churn for three sample citiesNew York, Chicago, and San Franciscobetween 1990 and 2000. We find a high degree of turnover and migration of headquarters, but an even higher degree of headquarters growth that has come about as small local companies have grown large. This result implies that policies to assist the growth of local indigenous firms of smaller size may be more beneficial than policies aimed at recruitment of footloose companies. Policymakers and site selection professionals will also be interested in the evidence we provide as to where headquarters are now emerging. Several broad spatial shifts in headquarters location have been observed prior to the 1990s. One of the persistent characteristics of the U.S. economy has been the concentrated location of large company headquarters in a relatively small number of large metropolitan areas. That is not surprising if one considers the nature of headquarters operations. Headquarters employ highly skilled professionals and they demand ready access to highlevel business services, such as legal, financial, and advertisingall of which tend to be found in large Thomas Klier is a senior economist and William Testa is a vice president and senior economist at the Federal Reserve Bank of Chicago. The authors would like to thank Dan McMillen for helpful comments and Woong Lim and Jeff Rasmussen for research assistance. 2Q/2002, Economic Perspectives metropolitan areas. Furthermore, since headquarters facilities must control and administer an often far-flung organization, ready access to state-of-the art communications infrastructure, as well as personal transportationthat is, air transportation and connectionsare a necessity in todays economy. As a result of these demands, a relatively small number of metro areas enjoy a comparative advantage in hosting headquarters. Our findings on headquarters location are generally consistent with those of earlier studies. Large metropolitan areas continue to have a comparative advantage in hosting headquarters of large companies. In fact, our analysis reveals no change in the overall share of large company headquarters domiciled in the 50 largest U.S. metropolitan areas between 1990 and 2000. However, there have been significant shifts within this distribution of metropolitan areas. Among the 50 largest metropolitan areas, those with population between 1 million and 2 million experienced the largest growth in population in the 1990s and developed concentrations of large company headquarters. In contrast, New York, the largest metropolitan area, continued its long-term trend of slowly losing dominance in terms of headquarters count. More generally, we find no evidence that the very largest metropolitan areas increased their share of corporate headquarters during the decade. Indeed, the share of headquarters domiciled in the five largest metropolitan areas fell from 36 percent in 1990 to 33 percent in 2000. This shrinkage at the top of the distribution is something of a surprise, because the rapid globalization trends during the 1990s were predicted to give rise to an increased concentration, that is, a few global headquarters cities. The reasoning goes that, as trade, transportation, and communications barriers fall, as they did in the 1990s, the potential market size of large companies grows. At the same time, the complexity of the corporate control functions for these companies increases. As a result, headquarters will increasingly locate in a small number of cities having abundant and specialized business and financial services or in cities with very intense concentrations of such industries. In these places, the firm administering a national or international market can stay abreast of innovation and otherwise acquire the information, ideas, and assistance it needs to succeed. Furthermore, headquarters will find it advantageous to locate near others of their ilk, again supporting the trend toward concentration in a small number of services-intensive metro areas. To some degree, this tendency was borne out in our multiple regression analysis; those metropolitan areas containing high concentrations of financial services activity were favored with greater headquarters gains Federal Reserve Bank of Chicago over the decade of the 1990s. However, our finding that the most populous cities continue to lose share may also mean that the technological advances and falling costs of travel and communication have improved the ability of headquarters located in smaller cities to gather information and services and to administer their farflung global markets and operations. Another reason that large cities have not done better is that population and associated markets have been shifting to mid-tier cities, especially in the South and West. Headquarters locations often follow shifting markets; indeed, we find that a regression variable reflecting market growthspecifically, population growthtends to correlate with headquarters growth. A variable indicating that the metropolitan area is located in the South census region is also significantly related to headquarters growth. While the West gained population as well, it did not gain headquarters to the same extent as the South. Apparently, in addition to the beneficial effects of local market growth, several prominent urban areas in the South have matured as commercial centers. In particular, Atlanta, Houston, Nashville, and Southeast Florida laid claim to much of the regions increase in corporate headquarters. We also find that, since regions tend to specialize in certain industries, headquarters concentration has tended to grow along with metro areas and their specialized industries. Large headquarters often emerge in the cities and regions in which successful new companies or industries grow. This is especially so for young industries and companies that rely heavily on research and development (R&D) and new technologies, for which close communication between the central office, lab, and production operations is essential. For example, we would expect the emergence of high-technology industries in Silicon Valley to have been accompanied by the growth of large corporate headquarters in the San Francisco Bay area, and this has in fact been the case. This metropolitan area did remarkably well in increasing its tally of corporate headquarters during the 1990s, garnering most of the growth of companies associated with the so-called new economy. In fact, just under half of the increase in headquarters there during the decade resulted from the growth of existing companies.1 More generally, we find that the shift in the geographic distribution of high-tech industry headquarters over the decade is unlike the overall trend displayed for all industries. That is, high-tech headquarters are becoming more concentrated in large metropolitan areas rather than dispersing toward the smaller and medium-sized cities. Financial companiesespecially bankshave also bucked the general trend by shifting toward larger 13 metropolitan areas. In this instance, profound deregulation has encouraged firm consolidation and market expansion. In response, the now-larger companies have chosen to locate their headquarters in larger metropolitan areas. Overall, then, our findings for the 1990s suggest that the largest urban areas continue to be highly preferred as headquarters locations. However, we identify a changing trend in the distribution of large headquarters across metropolitan areas. This trend implies that the second tier of metropolitan areas may begin to enjoy greater success in the competition for headquarters. The evidence shows that corporate headquarters are dispersing to mid-sized metropolitan areas and following shifting population and markets, especially toward the South. We also find that, for all metro areas, policies that emphasize the nurturing and growth of local companies rather than, or in addition to, recruitment of firms from outside the area may be beneficial. Our research indicates that company headquarters do not migrate so much as they grow and decline. Literature review The growth and locational patterns of large corporate headquarters have been a subject of research since the latter half of the twentieth century (see Lichtenberg, 1960, Evans, 1973, and Quante, 1976, for a synopsis of earlier work). Studies have examined various periods and drawn on a variety of data sources. Generally, the work utilizing large data sets tends to be cross-sectional, whereas studies tracking the distribution of headquarters over time tend to rely on Fortune 500 data. Horst and Koropeckyi (2000) and Holloway and Wheeler (1991) base their time-series analysis on data for Fortune 500 companies. Holloway and Wheeler (1991) conduct their empirical analysis for the 1980s using annual data for that decade. Horst and Koropeckyi (2000) utilize the same data from 1975 through 1999 (in five-year intervals). Shilton and Stanley (1999) utilize data for all publicly traded companies, regardless of company size, and Davis (2000) draws on data from the Survey of Auxiliary Establishments (U.S. Bureau of the Census). A common finding in all these papers is the high degree of concentration among headquarters. For example, Shilton and Stanley (1999) report that 40 percent of their sample is located in only 20 U.S. counties. They explain this stylized fact by the comparative advantage of cities to support headquarters operations. In fact, Horst and Koropeckyi (2000) report a strengthening of that effect during the 1990s as evidenced by a substantial drop in Fortune 500 headquarters located in non-metropolitan counties. In addition, the 14 advantage of certain cities in hosting headquarters seems to depend little on the historical and perhaps serendipitous presence of individual companies. For example, despite Bostons ongoing strength as a domicile of Fortune 500 companies headquarters, only two of the 15 present in 1999 had been there since 1975 (Horst and Koropeckyi, 2000). What exactly are the competitive advantages of large cities? The central function of corporate headquarters is the acquiring and dissemination of information. The demand side of the profit equation requires that corporate headquarters stay abreast of emerging developments in their markets. Meanwhile, the competitive supply or cost element of the profit equation suggests that firms must adapt new production technologies and management strategies. In turn, both of these categories of activities will often require dissemination of information and administration to a wideranging geography of operations. Thus, major airports represent a critical infrastructure for corporate headquarters, along with major highways, and telecommunications (Dow Jones, Inc., 1977). Air connections allow headquarters personnel to travel to direct their own operations both domestic and international, as well as to interact with others in their industry at conventions and trade shows (Boyle, 1990). Significantly, a major airport also brings meetings, conventions, suppliers, and customers into the home city. Several other features of the headquarters as a learning operation also imply a need for the large scale of a metropolitan area. The learning curve of technology is often shortened by proximity to other similar firms, as firms learn of new ideas through interaction. For example, Walcott (2001) documents the location of both health and bio-tech firms in proximity to Eli Lilly in Indianapolis (and in other production centers and emerging markets) as contributing to the companys successful acquisition of information. Accordingly, the clustering of firms can reflect a competitive advantage (Porter, 2000; Glasmeier, 1988). Professionals and highly skilled personnel are also more easily recruited and retained in cluster locations (Dow Jones, 1977). This follows as job mobility and advancement are enhanced by the information and career advancement opportunities that proximity to a host of firms and jobs provide to both the primary worker and, often, to the spouse (Ady, 1986). The persistent concentration of headquarters in certain individual cities that contain important business service sectors, such as New York and Chicago, also points to the ready access to purchased services as enabling factors for the concentration of headquarters. Concentrations of business service firms, such 2Q/2002, Economic Perspectives as media, law, accounting, and consulting, in large cities may enable firms to achieve cost and price advantages by shopping among a host of nearby business service providers. Possibly these services are purchased by headquarters and subsequently delivered to branch operations throughout the organization (see Ono, 2001). So too, the purchase of business services can be part of the organizations learning functions. Companies also learn and acquire services effectively from sources outside of their own industry. Lichtenberg (1960) observed the following 40 years ago: Like producers of unstandardized products, the central office executives produce answers to unstandardized problems, problems that change frequently, radically, and unpredictably. These problems are solved quickly only by consultation with a succession of experts. But most central offices would find it inefficient if not impossible to staff themselves internally with all of the specialized personnel and services that they must call on from time to time to solve their problems. Nor is it convenient to transport the experts to their plants or maintain effective contact by telephone or letter. All of these considerations dictate a concentration of central offices in a tight cluster near each other and near their suppliers. In recent years, however, we have seen a loosening of the location ties of business services industry and corporate headquarters. In particular, the phenomenon of outsourcing, along with advances in communication and air travel, may be facilitating a shift of large corporate headquarters away from the very large metropolitan areas that once dominated. Sassen (2001a) observes that many of the largest cities worldwide particularly London, New York, and Chicagohave been losing numbers of headquarters of the worlds largest companies for over three decades, even while business service industries there continue to grow.2 She hypothesizes that the outsourcing of complex service functions by global headquarters operations has been accelerating, and that this has liberated corporate headquarters to locate in any number of places that may be strategic for administration or control of the companys establishments. Drucker (1989) once advised firms to sell the mail room, while Sassen now claims that they are selling both the mail room and the board room. Hence, the locational concentration of complex business services rather than headquarters themselves has become the key feature by which to identify dominant global cities.3 It is not only outsourcing of business services that may be liberating corporate headquarters from large cities. Technological changes are inexorably lowering the costs of communications and travel to corporate Federal Reserve Bank of Chicago headquarters themselves. While globalization and technological changes are expanding potential markets for companies and increasing the complexity of management operations, they are also enabling cheaper and more effective communication across the world and across the spectrum of a companys facilities. The need for face-to-face communication to efficiently solve the most complex problems and the most delicate negotiations may never be eliminated by electronic communication (Quante, 1976). However, the use of remote communications is certainly accelerating (Townsend, 2001). As a result, administration from smaller and more remote locations may be easier than before. For now, the tensions between firm complexity/scope and better communications technology may be partly offsetting each other in terms of their effects on headquarters location and city size. Still, headquarters concentrations may be shifting toward metro areas that do not rank at the top of the size distribution. Horst and Koropeckyi (2000) and Holloway and Wheeler (1991) analyze the change over time in the concentration of headquarters location across metropolitan areas. Both studies find evidence of redistribution among the headquarters cities away from New York to mostly mid-size metropolitan areas. In 1955, the first year the Fortune 500 list was compiled, the New York metro area was home to 31 percent of all company headquarters on the list, the vast majority of which were located right in the city (28 percent of all Fortune 500 headquarters). While the metro area share of national headquarters remained stable until the early 1970s, the city began to lose headquarters to its surrounding areas in the mid-1960s. For the last 30 years, the share of headquarters domiciled in the New York metro area has been steadily declining. By 1999, it had fallen to 10 percent of Fortune 500 companies (see Quante, 1976, and Horst and Koropeckyi, 2000). Of course, the location of company headquarters has also been affected by the varying fortunes of industries and lines of business over time. As Holloway and Wheeler (1991) clearly establish, shifts in headquarters dominance by city size are related less to relocations of existing headquarters than to the growth of local companies that become large enough to be included in the Fortune 500 list. This implies that the indigenous growth of stellar companies and emerging industry clusters are an important explanatory factor in the shifting of headquarters concentration.4 Of course, this effect is symmetric with respect to industry decline. However, as an added wrinkle, a continued concentration of corporate headquarters has been observed to lag behind the decline of its overall industry in a 15 region (Rees, 1978). For example, corporate headquarters of large manufacturing companies tended to remain in large Northeast and Midwest cities long after their production capacity had migrated south and west. In sum, previous studies have documented a strong central tendency for headquarters to locate in large urban areas. However, the distribution of headquarters among regions and along the size hierarchy of urban places has been less stable, and the underlying reasons more elusive. Accordingly, the data must tell their own story for the 1990s. Data In order to document recent location patterns of large company headquarters, we analyze Compustat data on publicly traded companies for the years 1990 and 2000. The data represent a panel of all public companies whose shares are traded in the U.S., with the exception of American Depository Receipts (ADRs), closed-end mutual fund and index shares, and preFinancial Accounting Standards Board (FASB) companies.5 Active companies are either publicly traded or are required to file with the Securities and Exchange Commission. Similar to the previous literature, this article focuses on the headquarters of large companies. We define a company to be large if its total employment worldwide is at least 2,500.6 The data do not identify information on employment located at the headquarters site itself. However, data from the Census of Enterprise Statistics (U.S. Department of Commerce, 1992) are somewhat helpful in identifying employment at so-called auxiliaries, which are defined as separate establishments of multiestablishment companies that perform administration, management, research, and other supporting functions. These data report the average employment at auxiliary establishments to be 68, while companies with auxiliaries averaged 1,555 domestic employees overall. Most, but not all, of these auxiliaries are headquarters. Since the companies in our data set are only modestly larger in total employment size, their average headquarters size is also likely to be modestly larger. A recent survey by Aksoy and Marshall (1992) of 20 major international firms domiciled in the United Kingdom, employing as many as 150,000, reported only two head offices with more than 300 employees. (Furthermore, headquarters employment for these large U.K. companies declined appreciably during the 1980s and early 1990s.) In this article we aggregate headquarter locations by metropolitan areas. In particular, we use the most extensive definitions of metropolitan areas available, the so-called consolidated metropolitan statistical 16 area (CMSA).7 Thus, our results are not affected by relocations of headquarters from a central city to a suburban location within the same metropolitan area. We believe that these metropolitan areas largely share common locational attributes that are considered in the headquarters siting decision. Some of the important attributes include hub airports, access to business service firms, and a common skilled labor pool. Using our company-wide employment cutoff of 2,500 employees results in 1,397 metropolitan-area based records for 1990 and 1,805 records for 2000, about 22 percent of all records in the database.8 Hence, our sample is considerably larger than the Fortune 500, yet it includes essentially all the 2000 Fortune 500 companies. Geography of headquarters The distribution of large company headquarters across U.S. metropolitan areas is highly concentrated. In 1990, only 47 percent of the 276 metropolitan areas were home to at least one large company headquarters facility; in 2000, the figure was 52 percent. Even among headquarters-occupied metropolitan areas, the distribution of headquarters is highly skewed. However, the list of metro areas that are home to most company headquarters hardly changed during the 1990s. Both at the beginning and at the end of the last decade, the 50 most populous metropolitan areas were home to 87 percent of all large company headquarters (see table 1).9 There was considerable variation in the growth of headquarters during the decade among the largest metropolitan areas. Ten of the largest 50 metropolitan areas showed no net gain of headquarters. On the other hand, the ten fastest growing metropolitan areas experienced a net increase in headquarters of at least 100 percent (see table 2). It turns out that among the 50 largest metropolitan areas, those with population between 1 million and 2 million (ranked 2350 in table 2) experienced the largest growth in both population and large company headquarters during the last decade (see table 1). In contrast New York, the largest metropolitan area, continued its long-term trend of slowly losing dominance in terms of headquarters count. Despite this erosion, even at the end of the 1990s New York was home to more than twice as many headquarters of large companies than the runner-up metropolitan area, Chicago. Figure 1 shows the distribution of headquarters and population among metropolitan areas by quartiles (defined by population) in the year 2000.10 Notice the remarkable concentration of headquartersin absolute terms as well as relative to the concentration of populationin quartile 1, the 69 most populous metropolitan areas. The top quartile (labeled quartile 4 in 2Q/2002, Economic Perspectives TABLE 1 Population and headquarters across metro areas Percent of population 1990 2000 Top 5 metro areas Top 5 excl. New York Rank 6 to 22 Rank 23 to 50 Top 50 Remainder All Percent of headquarters 1990 2000 % Change, 1990–2000 Population Headquarters 28 18 28 15 71 28 27 18 29 16 72 28 36 20 36 15 87 14 33 19 38 16 87 13 11 12 16 18 14 13 19 29 35 45 30 23 100 100 100 100 14 29 Sources: Compustat, Census Bureau, and authors’ calculations. the figure) of metropolitan areas contain 78.6 percent of population and 92.1 percent of the large publicly traded company headquarters. This corroborates for the decade of the 1990s the agglomerative pull of large metropolitan areas found in previous studies. An alternative, more comprehensive, way to characterize the geographic distribution of headquarters location across metropolitan areas is by means of a Lorenz curve. A Lorenz curve graphs cumulative frequency distributions. It shows the degree to which a distribution is concentrated by the distance between the actual distribution and the 45 degree line, which represents an egalitarian distribution. Figure 2 shows the concentration of headquarters among the 50 most populous metro areas. It graphs the cumulative distribution of headquarters on one axis versus the cumulative distribution of metropolitan areas on the other axis. In that distribution, each metro area is treated as an equally weighted entity. The shape of the plotted line reveals the degree of concentration in the distribution of headquarters. For example, if each of the largest 50 metropolitan areas contained the same number of corporate headquarters, the graph line would be identical to the 45 degree line. In contrast, to the extent that some metropolitan areas host disproportionate numbers of headquarters, the graph curve will be bowed out toward the southeast, away from the 45 degree line. Figure 2 shows these curves for both 1990 and 2000 to illustrate changes in the concentration of headquarters within the largest 50 metropolitan areas. The various panels show curves for all headquarters and headquarters classified by selected major industry group (we chose a few prominent industries). For the year 2000, we find that the degree of concentration of headquarters among the largest metropolitan areas is quite similar across the various sectoral breakdowns, with about 60 percent of headquarters residing in the largest ten metro areas, as measured Federal Reserve Bank of Chicago by the number of headquarters. One notable exception to that general finding is the high-tech manufacturing sector, which is significantly more concentrated (about 80 percent of headquarters are found in the ten largest, by headquarters, metropolitan areas). Over the past 25 years, high-tech industries, such as computing and telecommunications equipment and software, have grown rapidly and displayed an acute tendency to concentrate heavily in a few metro areas, such as San Jose, RaleighDurham, Austin, and Boston. Young industries characterized by a high degree of innovation and competition appear to be loath to spatially separate their headquarters activity from their R&D or their production plants (Malecki, 1980). A comparison of Lorenz curves for headquarters for 1990 and 2000 also illustrates that corporate headquarters have become more ubiquitous across medium-sized metropolitan areasspreading to less headquarters-intensive areas. This trend is consistent across major industry groups with two exceptions. High-tech manufacturing shifts outward along part of its distributionwith the more headquarters-intensive MSAs gaining share of high-tech activity from 1990 to 2000. The same can be said for the finance, insurance, and real estate (FIRE) sector, only to a more pronounced degree. Upon further investigation, the increase in concentration of headquarters in that sector can be explained by an increase in the concentration of headquarters in the banking sector. This presumably is a response to regulatory changeslargely looseningbeginning in the early 1980s and continuing through the 1990s. So called deregulation has encouraged banks to grow in size which has, in turn, shifted the distribution at the top of the industry even further toward larger banks. Regulatory changes have allowed banks to enter new product lines, which has acted to increase their size and, in some instances, to merge with other, nonbanking, financial firms. 17 TABLE 2 Top 50 metro areas, by 2000 population Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 Metro area New York–Northern New Jersey–Long Island, NY–NJ–CT–PA CMSA Los Angeles–Riverside–Orange County, CA CMSA Chicago–Gary–Kenosha, IL–IN–WI CMSA Washington–Baltimore, DC–MD–VA–WV CMSA San Francisco–Oakland–San Jose, CA CMSA Philadelphia–Wilmington–Atlantic City, PA–NJ–DE–MD CMSA Boston–Worcester–Lawrence, MA–NH–ME–CT CMSA Detroit–Ann Arbor–Flint, MI CMSA Dallas–Fort Worth, TX CMSA Houston–Galveston–Brazoria, TX CMSA Atlanta, GA MSA Miami–Fort Lauderdale, FL CMSA Seattle–Tacoma–Bremerton, WA CMSA Phoenix–Mesa, AZ MSA Minneapolis–St. Paul, MN–WI MSA Cleveland–Akron, OH CMSA San Diego, CA MSA St. Louis, MO–IL MSA Denver–Boulder–Greeley, CO CMSA Tampa–St. Petersburg–Clearwater, FL MSA Pittsburgh, PA MSA Portland–Salem, OR–WA CMSA Cincinnati–Hamilton, OH–KY–IN CMSA Sacramento–Yolo, CA CMSA Kansas City, MO–KS MSA Milwaukee–Racine, WI CMSA Orlando, FL MSA Indianapolis, IN MSA San Antonio, TX MSA Norfolk–Virginia Beach–Newport News, VA–NC MSA Las Vegas, NV–AZ MSA Columbus, OH MSA Charlotte–Gastonia–Rock Hill, NC–SC MSA New Orleans, LA MSA Salt Lake City–Ogden, UT MSA Greensboro–Winston–Salem–High Point, NC MSA Austin–San Marcos, TX MSA Nashville, TN MSA Providence–Fall River–Warwick, RI–MA MSA Raleigh–Durham–Chapel Hill, NC MSA Hartford, CT MSA Buffalo–Niagara Falls, NY MSA Memphis, TN–AR–MS MSA West Palm Beach–Boca Raton, FL MSA Jacksonville, FL MSA Rochester, NY MSA Grand Rapids–Muskegon–Holland, MI MSA Oklahoma City, OK MSA Louisville, KY–IN MSA Richmond–Petersburg, VA MSA Population (000s) HQs Net change HQ number Net change HQ % 21,200 16,374 9,158 7,608 7,039 239 85 109 66 91 16 4 13 22 39 7.2 4.9 13.5 50.0 75.0 6,188 5,819 5,456 5,222 4,670 4,112 3,876 3,555 3,252 2,969 2,946 2,814 2,604 2,582 2,396 2,359 2,265 1,979 1,797 1,776 1,690 1,645 1,607 1,592 1,570 1,563 1,540 1,499 1,338 1,334 1,252 1,250 1,231 1,189 1,188 1,183 1,170 1,136 1,131 1,100 1,098 1,089 1,083 1,026 997 70 66 34 76 70 53 31 19 23 50 35 18 39 27 20 21 13 23 2 19 26 9 11 9 6 13 21 14 7 5 16 2 25 6 4 12 6 10 13 7 6 9 6 10 21 15 11 1 18 29 25 16 –1 12 12 –4 8 12 12 9 0 –1 5 1 1 5 7 –1 4 2 5 7 3 –1 –2 9 1 16 2 2 –2 0 2 11 2 0 5 2 4 6 27.3 20.0 3.0 31.0 70.7 89.3 106.7 –5.0 109.1 31.6 –10.3 80.0 44.4 80.0 81.8 0.0 –7.1 27.8 100.0 5.6 23.8 350.0 –8.3 80.0 50.0 62.5 50.0 27.3 –12.5 –28.6 128.6 100.0 177.8 50.0 100.0 –14.3 0.0 25.0 550.0 40.0 0.0 125.0 50.0 66.7 40.0 Note: HQ indicates headquarters. Sources: See table 1. 18 2Q/2002, Economic Perspectives FIGURE 1 FIGURE 2 Headquarters and population by MSA quartiles, 2000 Distribution of headquarters among 50 largest metro areas percent A. All 100 cumulative frequency of headquarters 80 100 Headquarters Population 80 60 40 60 20 40 1990 2000 0 Quartile 1 Quartile 2 Quartile 3 Quartile 4 20 Sources: Compustat, Census Bureau, and authors’ calculations. 0 Presumably, the tendency of larger organizations to prefer headquarters locations in larger metropolitan areas has thus brought about the shift observed in figure 2, panel C. In addition, deregulation has loosened restrictions that had been placed on banks to serve markets across state lines, or within states, across county lines, and other boundaries. This has facilitated geographic consolidation of markets in the banking sector, often through a merger.11 For example, the merger between Banc One of Columbus and NBDFirst Chicago in 1998 resulted in a headquarters choice of Chicago. These industry-specific events produced a headquarters location trend in the 1990s that was the opposite of that of most industries in which midsized metropolitan areas were the relative gainers. Mid-sized metropolitan areas were the gainers not only because of headquarters choices, but also because they grew faster in population size. They emerged as sizable markets so that their companies and headquarters grew along with them. Nonetheless, the growing prominence of mid-sized metropolitan areas does not account for the entire shift of headquarters toward these places. Figure 3 illustrates the distribution for headquarters across all industries, as well as for population for the largest 50 metro areas in 1990 and 2000. We can see that headquarters are more concentrated among metro areas than population. This is true for both 1990 and 2000. However, during the 1990s the relative difference between the distribution of headquarters and population narrowed. This is demonstrated in panel B of figure 3, which plots the vertical distance between both distributions at both points in time. While the contour of that distance has not changed much, it narrowed across the entire range of the distribution Federal Reserve Bank of Chicago 0 20 40 60 80 100 cumulative frequency of MSAs B. High-tech manufacturing cumulative frequency of headquarters 100 80 60 40 20 2000 0 0 20 40 60 1990 80 100 cumulative frequency of MSAs C. Finance, insurance, and real estate cumulative frequency of headquarters 100 80 60 40 1990 20 2000 0 0 20 40 60 80 100 cumulative frequency of MSAs Sources: Compustat, Census Bureau, and authors’ calculations. 19 during the decade. In addition, from panel A of figure 3 we can tell that that movement was driven in large part by a redistribution of headquarters as opposed to a redistribution of population. Different growth and reorganizational experiences across industries also become important in understanding the regional shifts in headquarters that have taken place. In examining the shifts among the four major regions as defined by the U.S. Bureau of the Census,12 we find that at the beginning of the decade, both the Northeast and the Midwest regions were the most headquarters-intensive among the four. That is not surprising as the industry structure of the Northeast and Midwest reflects their rich manufacturing history. Even though manufacturing plants spread beyond their regions boundaries long ago, many of the countrys headquarters of industrial companies continued to be located there in 1990 (see Rees, 1978). As these industries companies decline in size and importance or are acquired by overseas companies, these headquarters are evaporating. So too, with a lag, headquarters sometimes follow their operating manufacturing plants to FIGURE 3 Distribution of headquarters and population A. All large headquarters percent 100 1990 2000 1990 2000 80 headquarters headquarters population population 60 40 20 0 0 20 40 60 80 100 cumulative frequency of MSAs B. Vertical distance between the two distributions percent 9 6 1990 2000 3 0 -3 0 20 40 60 80 100 cumulative frequency of MSAs Sources: Compustat, Census Bureau, and authors’ calculations. 20 Sun Belt locales.13 As a result, both the Northeast and Midwest regionsbut especially the Northeastcontinued to shed such headquarters during the 1990s. Figure 4 illustrates the U.S. geography of all large company headquarter locations in the year 2000.14 Figure 5 clearly shows the 1990s to be the decade of the South. While leading the country in population share at the beginning of the decade, it represented just over 25 percent of all large company headquarters. But during the 1990s the number of headquarters domiciled in the South grew much faster than its population share. In fact, at the end of the decade that regions share of headquarters had virtually pulled even to its share of population. Apparently, in addition to the beneficial effects of local market growth, several prominent urban areas in the South have matured as commercial centers. In particular, Atlanta, Houston, Nashville, and Southeast Florida laid claim to much of the regions increase in corporate headquarters (see figure 6). In contrast, the West continued to grow its population at a faster rate than its headquarters. Hence, it remains the least headquarters-intensive region on a per capita basis, despite the tremendous growth in high-tech manufacturing in the 1990s (see figure 7). High-tech manufacturingdefined at the 3-digit SIC level as pharmaceuticals, computers and office equipment, communication equipment, electrical components, and aircraft and partsbehaved very differently from the rest of manufacturing during the 1990s.15 The West experienced the strongest growth in hightech manufacturing headquarters, leaving it with the highest share at the end of the decade, ahead of the Northeast. The Midwest, on the other hand, experienced an almost commensurate drop in its share. Underlying that phenomenon is the well-known growth of the high-tech sector during the 1990s, a large part of which occurred in and around Silicon Valley. The rest of manufacturing experienced little change in its regional distribution; the Midwest regions share remained essentially unchanged, whereas the Northeast lost share and the South gained share. The role of regional industry specialization can be seen in examining the individual components of growth and decline for a few representative metropolitan areas. (see table 3). Table 3, panel A starts in 2000 and looks at the history of large headquarters over the previous ten years. We distinguish the following categories: 1) survivor in same metropolitan area with same company name and as large company; 2) indigenous company that grew during decade above 2,500 employees; 3) company is the result of merger involving companies listed separately in 1990merged 2Q/2002, Economic Perspectives FIGURE 4 Where the headquarters are, 2000 Note: Figure includes boundaries of four census regions: West, Midwest, South, and Northeast. Sources: Compustat, Census Bureau, and authors’ calculations. entity in MSA as listed; 4) company relocated; 5) company newly established, and 6) other. Panel A shows interesting differences and similarities across the three metropolitan areas. First, the incidence of companies relocating across metropolitan areas, while big news in the business press, does not affect the distribution of headquarters in a noticeable way. For all three metropolitan areas, between 7 percent and 10 percent of the headquarters active in 2000 had moved since 1990.16 On the other hand, we can see strong differences in the degree of churn across these three metropolitan areas. San Francisco, the center of the Internet and high-tech expansion of the last decade, finds itself with 57 percent of its large headquarters in 2000 either having been started during the decade (26 percent)17 or growing above the large company threshold (31 percent). FIGURE 5 Distribution of headquarters and population by region B. 2000 A. 1990 percent percent 50 50 40 Headquarters Population 40 30 30 20 20 10 10 0 Headquarters Population 0 Midwest Northeast South West Midwest Northeast South West Sources: Compustat, Census Bureau, and authors’ calculations. Federal Reserve Bank of Chicago 21 Model FIGURE 6 Population and headquarters growth in 1990s, 50 largest MSAs HQ growth rate 600 560 West Palm Beach 520 West South Northeast Midwest 480 440 400 360 Orlando 320 280 240 200 Nashville 160 120 Atlanta 80 Houston 40 0 -40 -10 0 10 20 30 40 50 60 70 population growth rate Sources: Compustat, Census Bureau, and authors’ calculations. 80 To more rigorously test the relationship between the factors discussed above and the change of headquarters at the MSA level, we use multiple regression analysis. Below, we briefly explain the variables and present the results. The dependent variable in our model is the percentage change in the number of headquarters in a metropolitan area. In order to minimize the effect of a small base at the start of the decade, we use only the 50 most populous metropolitan areas (see table 2).18 The descriptive data presented earlier suggest a number of influences on the change in the concentration of headquarters during the last decade. The high degree of concentration of headquarters among a relatively small number of metro areas suggests the existence of a scale effect in hosting headquarter operations. That effect is measured in our model by the level of population. While the coeffi90 cient for this variable should reflect the scale effect, we estimate the model only for the largest metro areas, so it should also pick up the redistribution from the largest to the medium-sized metro areas. Hence, the expected sign is ambiguous. We also include a variable measuring the percent change in population during the decade. This variable Neither New York nor Chicago approaches these numbers. By the same token, the latter two are characterized by larger survival rates of large company headquarters. Table 3, panel B traces the 1990 headquarters to the year 2000. The table disFIGURE 7 tinguishes the following categories: Non-high-tech vs. high-tech manufacturing headquarters 1) survivor in same metropolitan area by region with same company name and as large percent company; 2) indigenous company whose 50 employment fell below 2,500 over the 1990 non-high-tech manufacturing 2000 non-high-tech manufacturing decade; 3) company is the result of merg1990 high-tech manufacturing er involving companies listed separately 40 2000 high-tech manufacturing in 1990merged entity in MSA as listed; 4) company is result of merger in30 volving companies listed separately in 1990merged entity in different MSA; 20 5) company relocated to different MSA; 6) company went out of business; and 7) other. Again, similarities dominate. 10 About half of the 1990 headquarters survived in the same metro area. With 0 the exception of New York, we find Midwest Northeast South West relocation of companies to be a rather Sources: Compustat, Census Bureau, and authors’ calculations. rare occurrence, involving between 5 percent and 8 percent of the companies. 22 2Q/2002, Economic Perspectives sector to proxy for the degree to which a metro area specializes in the provision of Churn rate of headquarters business services. We expect a positive sign for two reasons. First, much of the A. Looking back from 2000 activity in FIRE industries is of the type Categories Chicago New York San Francisco purchased and outsourced by headquarters. (- - - - - - - - - - - - - - - percent - - - - - - - - - - - - - - -) Purportedly owing to the forces of gloSurvivor 49 41 30 balization, headquarters are increasingly Growth 12 10 31 seeking to locate where such services are Merged or acquired 12 18 5 accessible. Second, headquarters of FIRE Moved in 6 10 5 New 17 20 26 industries, especially banking, have been Other 4 2 2 rapidly consolidating, forming companies of large size, and perhaps doing so in B. Looking forward from 1990 metropolitan areas that already specialize Categories Chicago New York San Francisco in such activities. We also control for the Survivor 55 44 52 regional composition of headquarters No longer large 3 6 2 growth by a binary variable that measures Merged/acquired stayed 8 20 4 if the MSA is located in the South, as deMerged/acquired left 18 7 27 fined by the census region. Moved out 5 14 8 Out of business 8 8 4 The regression results point to the Other 1 2 4 effect of the change in population as well as the provision of business services in Note: Total may not add to 100 due to rounding Source: Compustat. influencing headquarters growth at the metro area level (see table 4). Consistently, these two variables are statistically should capture the shifting of markets away from the significant in the three model variations we estimattraditional centers of commerce and population and ed. We find that headquarters growth is elastic with show a positive sign. We might also see such a rerespect to growth in population: An increase in the sponse to growing population because the universe growth of population by 1 percent is associated with of large companies is increasingly composed of sera 2 percent increase in the growth of headquarters. vice rather than manufacturing companies.19 In addiA 1 percent increase in the earnings share in the FIRE tion, service companies tend to be more regional than national or international in TABLE 4 market scope. However, various past Regression results studies argue that headquarters need not follow markets. That is because enhanced Variables Model 1 Model 2 communication technology may allow Intercept –0.08 –0.72 control and oversight functions to be (0.62) (0.65) conducted from afar. Level of population (millions) –0.061 –0.038 Two variables control for the sectoral –0.04 –0.04 composition of the metropolitan areas. Change in population 2.14 2.09 First, we measure the share of manufac(0.96) (0.92) turing earnings in all nonfarm earnings Manufacturing share –0.69 0.83 (1989 data) in each metropolitan area. (1.79) (1.83) We expect a negative sign insofar as the FIRE share 8.95 9.45 (5.05) (4.82) Northeast and Midwest have been losing their dominance in manufacturing proSouth — 0.63 (0.27) duction to other regions. However, as documented by Rees (1978) and others, R-squared 0.21 0.30 headquarters tend to remain behind, or Adjusted R-squared 0.14 0.22 follow regional demand shifts only with Notes: Standard errors are in parentheses. Numbers in bold are long lags. Second, we compute a compastatistically significant. FIRE is finance, insurance, and real estate. rable share for employment in the FIRE TABLE 3 Federal Reserve Bank of Chicago 23 sector corresponds to a 9.5 percentage point increase in the growth rate of headquarters. Finally, if a metro area is located in the South, we observe headquarters growth that is about 0.6 percent higher than in metro areas located in the rest of the country. Conclusion Headquarters of large companies continue to be desired and actively pursued by states and regions. Our findings for the 1990s provide further evidence to support the historical trend that the largest urban areas are highly preferred as headquarters locations. The momentum of this locational preference apparently continued throughout the decade. However, the evidence does point to some changes in the distribution of large headquarters among sizable metropolitan areas. First, the very largest metropolitan areas witnessed a drain of headquarters to the middle tier of cities during the 1990s. New York City had been experiencing an erosion for several decades, but the trend is more pervasive than that. Apparently, secondtier cities have improving chances of success in the competition for large company headquarters. This tendency for gains among the second tier may surprise some analysts of globalization, who have predicted that the larger and more complex companies that result from globalization would flock to the very largest metropolitan areas in search of the most extensive communications, talent, ideas, and transportation. Further investigation is needed to understand the nature of the shifting distribution of headquarters by size of metropolitan area. Significant shifts are also taking place among regions and among metropolitan areas. Among large multi-state regions, the South was a big gainer in the 1990s. To some extent, this reflects the shifting of markets and population growth to the South. Yet, the West also gained population but did not experience headquarters gains to the same extent. Apparently, in addition to market growth, the maturing of key urban areas in the South is contributing to the regions attractiveness. Among both metropolitan areas and regions, the performance of indigenous industries and individual companies is also key. Our research clearly shows that company headquarters do not migrate so much as they grow and decline. NOTES 1 Of 91 headquarters in the San Francisco Bay area at the end of 2000, 28 represent public companies that grew during the decade and 20 represent companies that went public during the decade. 2 See Sassen, 2001a, p. 109. 3 See, for example, Scott, 2001, p. 82. 10 It is essentially unchanged from 1990. Federal Reserve Bank of Chicago (2000) and DeYoung et al. (2002). 11 The four census regions are defined as follows: West: Alaska, Arizona, California, Colorado, Hawaii, Idaho, Montana, Nevada, New Mexico, Oregon, Utah, Washington, and Wyoming; Midwest: Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, and Wisconsin; Northeast: Connecticut, Massachusetts, Maine, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, and Vermont; and South: Delaware, District of Columbia, Maryland, Virginia, West Virginia, North Carolina, South Carolina, Georgia, Florida, Alabama, Mississippi, Arkansas, Tennessee, Kentucky, Louisiana, Texas, and Oklahoma; also see figure 4. 12 4 Microsoft may be one prominent example where a dominant company chose a non-standard indigenous location. In contrast, Gateway Computers move from North Sioux City, SD, to San Diego, CA, in 1999 attests to the countervailing pull that urban economies can exert on large companies. 5 Compustat created pre-FASB company records upon introduction of FASB rule 94 regarding the accounting of financial service subsidiaries to show consistency between current and historical data. Our results are robust to lowering the cutoff for large companies to 2,000 employees. 6 For example, the Chicago CMSA encompasses the primary metropolitan statistical areas of Chicago, IL, Gary, IN, Kankakee, IL, and Kenosha, WI. 7 In 1990, there are 61 (4.2 percent of all large company records) large company headquarters located outside metropolitan areas; in 2000 there are 66 (3.5 percent). 8 Horst and Koropeckyi (2000) note that a metro area must have an employment base of at least 750,000 to be considered large enough to develop a strong agglomeration of support services (p. 26). 9 24 This continues a trend that has been documented for the 1960s and 1970s; see Rees, 1978. 13 14 The figure shows a dot for each headquarters location in the database, regardless of location in metro area. This map represents 1,871 headquarters. We use the Organization for Economic Cooperation and Development definition of high-tech industries, which is based on R&D intensity (see National Science Board, 2000). 15 16 Holloway and Wheeler (1991) identified the dynamics for all the records in their data set. Their finding on the importance of moves very closely matches ours: 10 percent of all additions of headquarters in the top 55 metropolitan areas were due to relocation. 2Q/2002, Economic Perspectives That term is somewhat misleading as start-up is measured relative to the universe of the database; in other words, a private company that was taken public would be classified as a start-up. In fact, 20 of 24 new companies in the San Francisco metro area were initial public offerings. 17 Holloway and Wheeler (1991) estimate a model for 55 metro areas. In order to be included in that set, a metro area had to be host to at least one Fortune 500 headquarters both in 1980 and 1987. Their 18 dependent variable is a measure of the change in corporate dominance, which is measured by the change in the proportion of total Fortune 500 assets held within a metro area. From 1990 to 2000, the share of service sector companies in our database increased from 9.6 percent to 17.2 percent, while manufacturing companies fell from 43.1 percent to 37.2 percent. 19 REFERENCES Ady, Robert M., 1986, Criteria used for facility location selection, in Financing Economic Development in the 1980s, Norman Walzer and David L. Chicone (eds.), New York: Praeger. Glasmeier, Amy, 1988, Factors governing the development of high tech industry agglomerations: A tale of three cities, Regional Studies, Vol. 22, No. 2, pp. 287301. Asu, Aksoy, and Neill Marshall, 1992, The changing corporate head office and its spatial implications, Regional Studies, Vol. 26, No. 2, pp. 149162. Goodwin, William, 1965, The management center in the United States, Geographical Review, Vol. 55, No 1, pp. 116. Boyle, M. Ross, 1990, Corporate headquarters: An elusive economic development target, Economic Development Commentary, Vol. 13, No. 4, pp. 2030. Holloway, Steven R., and James O. Wheeler, 1991, Corporate headquarters relocation and changes in metropolitan corporate dominance, 19801987, Economic Geography, Vol. 67, No. 1, pp. 5474. Compustat, 2000, database. , 1990, database. Davis, J. C., 2000, Headquarters, localization economies, and differentiated service inputs, Brown University, mimeo. Horst, Toni, and Sophia Koropeckyi, 2000, Headquarters effect, Regional Financial Review, February, pp. 1629. Lichtenberg, Robert M., 1960, One-tenth of a nation, Cambridge, MA. Harvard University Press. DeYoung, Robert, William C. Hunter, and Gregory F. Udell, 2002, Whither the community bank?, Chicago Fed Letter, No. 178, June. Malecki, Edward J., 1980, Corporate organization of R&D and the location of technological activities, Regional Studies, Vol. 14, pp. 319334. Dow Jones & Co., Inc., Market Research Department, 1977, Business on the Move, New York. National Science Board, 2000, Science and engineering indicators, National Science Foundation, Arlington, VA. Drucker, Peter, 1989, Sell the mailroom, Wall Street Journal, July 25, p. A16. Evans, Alan W., 1973, The location of the headquarters of industrial companies, Urban Studies, Vol. 10, pp. 387395. Federal Reserve Bank of Chicago, 2000, The Changing Financial Industry Structure and Regulation: Bridging States, Countries, and Industries, proceedings of the 36th annual Conference on Bank Structure and Competition. Federal Reserve Bank of Chicago Ono, Yukako, 2001, Outsourcing business services and the role of central administrative offices, Federal Reserve Bank of Chicago, mimeo. Porter, Michael, 2000, Location, competition, and economic development: Local clusters in the global economy, Economic Development Quarterly, Vol. 14, No. 1, pp. 1534. Quante, Wolfgang, 1976, The Exodus of Corporate Headquarters from New York City, New York: Praeger. 25 Rees, John, 1978, Manufacturing headquarters in a post-industrial urban context, Economic Geography, Vol. 54, No. 4, pp. 337354. Scott, Allen J. (ed.), 2001, Global cities in the twentyfirst century, in Global City-Regions Trends Theory, Policy, Oxford: Oxford University Press, pp. 5977. Sassen, Saskia, 2001a, The Global City, second edition, New York: Princeton University Press. Shilton, L., and C. Stanley, 1999, Spatial patterns of headquarters, Journal of Real Estate Research, Vol. 17, pp. 341364. , 2001b, Global cities and global cityregions: A comparison, in Global City-Regions Trends Theory, Policy, Allen J. Scott (ed.), Oxford: Oxford University Press, p. 82. Townsend, Anthony M., 2001, The Internet and the rise of the new network cities, 19691999, Environment and Planning B, Vol. 28, No. 1, pp. 3958. Saxenian, Annalee, 1994, Regional Advantage: Culture and Competition in Silicon Valley and Route 128, Cambridge, MA: Harvard University Press. U.S. Department of Commerce, Bureau of the Census, 2001, 1992 Enterprise Statistics, Washington DC: U.S. Government Printing Office. Schwartz, Joel, 1997, Corporate philanthropy today: From A. P. Smith to Adam Smith, National Commission on Philanthropy and Civic Renewal, Washington, DC, working paper. Walcott, Susan, 2001, Growing global: Learning locations in the life sciences, Growth and Change, Vol. 32, No. 4, Fall, pp. 511532. 26 2Q/2002, Economic Perspectives Post-resolution treatment of depositors at failed banks: Implications for the severity of banking crises, systemic risk, and too big to fail George G. Kaufman and Steven A. Seelig Introduction and summary Bank failures are widely viewed in all countries as more damaging to the economy than failures of other types of firms of similar size for a number of reasons. The failures may produce losses to depositors and other creditors, break long-standing bankcustomer loan relationships, disrupt the payments system, and spill over in domino fashion to other banks, financial institutions and markets, and even to the macroeconomy (Kaufman, 1996). Thus, bank failures are viewed as more likely to involve contagion or systemic risk than are failures of other firms. The risk of such actual or perceived damage is often a popular justification for explicit or implicit government-provided or -sponsored safety nets, including explicit deposit insurance and implicit government guarantees, such as too big to fail (TBTF), that may protect de jure uninsured depositors and possibly other bank stakeholders against some or all of the loss.1 But even with such guarantees, bank failures still invoke widespread fear. In part, this reflects a concern that protected and/or unprotected depositors may not receive full and immediate access to their claims on the insolvent banks at the time that the institutions are legally declared insolvent and placed in receivership.2 That is, they may suffer post-resolution losses in addition to any loss at the time of resolution. Unprotected depositors may be required to wait until the proceeds from the sale of the banks assets are received. Protected depositors may also not be paid in full immediately if the insurance agency has no authority or procedures for advancing payment before receipt of the sales proceeds, or if there is insufficient time to collect and process the necessary data on who are the insured depositors and how much is insured for each depositor. If depositors are not paid the full value of their claims immediately, some or all of the deposits are effectively temporarily frozen. In the absence Federal Reserve Bank of Chicago of an efficient secondary market for frozen deposits, both protected and unprotected depositors will experience losses in liquidity. Protected depositors will also experience present value losses if they are paid the par value of their claim after the date of resolution without interest. At the same time, the ability of the bank to conduct its normal lending business is greatly reduced. It is effectively partially or totally physically, as well as legally, closed. Indeed, a European bank analyst recently observed that The issue is not so much the fear of a domino effect where the failure of a large bank would create the failure of many smaller ones; strict analysis of counterparty exposures has reduced substantially the risk of a domino effect. The fear is rather that the need to close a bank for several months to value its illiquid assets would freeze a large part of deposits and savings, causing a significant negative effect on national consumption (Dermine, 1996, p. 680). That is, both the great fear of bank failures and the magnitude of any damage that such failures impose on other sectors of the economy are triggered as much if not more by losses in liquidity by both insured and George G. Kaufman is the John Smith, Jr., Professor of Finance at Loyola University Chicago and a consultant to the Federal Reserve Bank of Chicago. Steven A. Seelig is a financial sector advisor at the International Monetary Fund (IMF). The authors are indebted to George Benston (Emory University), Maximilian Hall (Loughborough University), Edward Kane (Boston College), Daniel Nolle (Office of the Comptroller of the Currency), Yuri Kawakami (IMF), and participants at the annual meeting of the Financial Management Association and at seminars at the IMF, Concordia University (Montreal), York University, Dalhousie University, and the University of the South for helpful comments on earlier drafts. The views expressed in this article are those of the authors and not necessarily those of the IMF. 27 uninsured deposits as by credit losses in the value of uninsured deposits.3, 4 The potential magnitude of losses to depositors and other stakeholders in bank failures is likely to affect both the supply of and demand for government guarantees to protect some or all bank stakeholders and to influence the resolution options available to a deposit insurer. The larger the potential losses in bank resolutions are perceived to be, the greater the demand for government guarantees by depositors and other stakeholders is likely to be and the more willing governments are likely to be to bow to such political pressures and supply the guarantees. Likewise, the larger the potential losses, the greater the probability that the accounts will be partially or totally frozen, the greater the potential harm to the macroeconomy, and the more likely the government will supply the guarantees to minimize the potential damage. Thus, the way depositors are treated at insolvent institutions in terms of the magnitude of the losses they may incur and their access to the value of their deposit claims has important public policy implications. It follows that the probability and magnitude of government guarantees may be reduced by reducing the perceived losses to depositors and other stakeholders in resolving insolvent banks. This article examines both the sources and implications of potential depositor losses in bank resolutions. In particular, we examine post-resolution depositor losses due to delays in paying both protected and unprotected depositors at failed banks the full current value of their claims in a timely fashion after a bank is officially declared insolvent and resolved. For de facto insured depositors, the value of their claims is the par value of the eligible deposits at the time of resolution less any explicit deductible or loss-sharing amount. For de facto uninsured depositors, the value of their claims is the present value of the estimated eventual pro-rata recovery value of the banks assets, which is likely to be less than the par value. Although losses in value to depositors in bank failures at the time of resolution have been frequently analyzed, this article contributes to the literature by analyzing the implications of losses in liquidity after resolution, in particular, losses from delayed depositor access through the freezing of insured and/or uninsured accounts, which have not been thoroughly analyzed up to now. Because the magnitude and timing of the losses in both value and liquidity to depositors in bank insolvencies are in some measure under the control of the deposit insurance agency or the government, the article also develops public policy recommendations on how to minimize losses to depositors from all 28 sources, but in particular the losses to depositors from delayed access to their funds after resolution. On the one hand, as noted, if this loss could be reduced, it could contribute to reducing both the demand for and supply of broad government guarantees, including reducing if not eliminating the need for TBTF. In the United States, the Federal Deposit Insurance Corporation (FDIC) currently pursues such a strategy. In many instances, it effectively makes the current value of their permissible claims available to both insured and uninsured depositors one or two business days after a bank is legally failed. Combined with faster resolution after economic insolvency that reduces depositor losses at the time of resolution, this strategy makes it more politically possible to resolve even large insolvent banks with losses to uninsured depositors. The banks are legally closed in terms of effectively terminating the ownership claim of the old shareholders and transferring ownership to new shareholders. Except in infrequent cases of liquidation, when there is no demand for the banking services in the community, the resolved banks are not physically closed. Thus, there is little, if any, interruption in their banking business.5 However, this practice is not followed in most other countries. Rather, in these countries, both insured and uninsured depositors are paid the value of their claims only through time after the resolution of the bank. These delays may at times stretch many months for insured deposits and many years for uninsured deposits. As a result, to reduce the potential adverse economic and political ramifications from such additional losses to depositors, governments in these countries are often reluctant to resolve insolvent banks with losses to uninsured depositors and permit the banks to continue in operation by effectively protecting all depositors and other stakeholders, including senior management. On the other hand, reductions in potential losses and in delays in payment could reduce depositor discipline on banks, thereby increasing the banks fragility and the probability of failure. Thus, either solution appears to have drawbacks as well as advantages; and an intermediate solution in terms of delay time in paying depositors may be preferred in reducing the potential damage from bank failures and maximizing aggregate economic welfare. This article models the tradeoffs between increased market discipline and increased probability of government bailout as the time delay by the insurance agency in paying depositors the full value of their claims is varied to solve for the optimal depositor access delay time. First, we identify and analyze the sources of potential losses to depositors in bank failures. Then, we 2Q/2002, Economic Perspectives discuss the implications of delayed depositor access at insolvent banks in terms of the effects on depositor discipline, on the one hand, and depositor pressure to protect all deposits, on the other. We consider ways that policymakers can reduce depositor losses from bank failures. Next, we describe the FDICs current procedure to provide depositors with full and immediate access to their claims at the time institutions are declared insolvent and placed in receivership and provide an overview of the history of immediate payment in the U.S. Then, we consider the advantages and disadvantages of full and immediate depositor access. We model the access timing decision graphically to solve for the optimal delay time. We then report on a survey of depositor access practices across countries conducted by the FDIC in spring 2000. Finally, we develop conclusions and best practice recommendations regarding depositor access to funds at resolved insolvent institutions to enhance the safety and efficiency of banking systems. Sources of potential losses to depositors Past analyses have identified five potential sources of economic loss to depositors or the government deposit insurance agency, which stands in the shoes of the de jure insured depositors, from the resolution of insolvent depository institutions: 1. Poor closure ruleEmbedded losses in value from a delay between the time when a bank becomes economically insolvent (that is, where the market value of the assets declines below the market value of the liabilities, which is the present value of the maturity value of the deposits and other debt) and the time it becomes eligible to be declared legally insolvent. 2. Regulatory forbearanceEmbedded losses in value from a delay in the time from when a bank becomes legally eligible to be declared insolvent and the time it is actually resolvedthat is, legally declared insolvent by the regulators or other authorized party (official recognition of the insolvency), a receiver appointed, and the existing owners removed. 3. Insufficient information and processing delay Possible losses from any time necessary after resolution for the deposit insurance agency to determine the identity of qualified protected and unprotected depositors and the qualifying deposits and to pay the depositors. 4. Bad market conditions after resolutionPossible losses (or gains) from any delay in the receivers selling the bank as a whole or in parcels after the bank is declared legally insolvent, either because of operational problems or to wait for a better market. 5. Inefficient receiverLosses from delay in the receivers distributing the proceeds from the Federal Reserve Bank of Chicago sales to the uninsured depositors and the deposit insurance agency. These potential losses occur sequentially. The first two sources of losses occur before the date of resolution because economically insolvent banks are permitted to stay open and operate under their existing owners and managers. The first loss arises from a poor legal closure rule that focuses on book or regulatory values that often overstate bank assets and understate bank liabilities compared with their economic or market values, particularly when a bank approaches insolvency. In the United States, banks (although not bank holding companies), unlike other corporations, are not subject to the jurisdiction of the bankruptcy process and courts. Rather, they are legally closed and a receiver appointed by their chartering or primary federal regulator. The second loss reflects regulatory forbearance from fear of imposing losses and injuring favored stakeholders of the insolvent bank (for example, shareholders, management, other employees, borrowers, or uninsured depositors), injuring other financial institutions, reducing the availability of banking services, or injuring the regulators own reputation as public guardians against bank failures. In addition, until the date of official resolution of the bank, embedded losses from the continued operation of insolvent banks are not booked and accrue only to the deposit insurance agency. Both insured and uninsured depositors can withdraw their maturing funds from these banks at par value, effectively stripping the banks of their best and most liquid assets. Because they are not officially booked, the embedded losses to the insurance agency are generally difficult for much of the public to recognize and easy for regulators to disguise, hide, and deny. Only at and after the date of official recognition of insolvency are the total embedded losses booked and visible to all and a pro-rata share imposed on the remaining unprotected depositors. This encourages regulators to delay closure. As a result, regulators are at times poor agents for their principalshealthy banks and taxpayers. The costs of regulatory forbearance in encouraging moral hazard behavior by the banks and increasing eventual losses to depositors in the U.S. and abroad have been amply documented (Kane, 1989 and 1990; Kane and Yu, 1995; Kaufman, 1995 and 1997a; Barth, 1991; and Gupta and Misra, 1999). The costs of a poor closure rule and forbearance include not only increased credit and market losses, but also increased losses from fraud and asset stripping, which is more likely at insolvent or near-insolvent institutions, and the misallocation of financial resources, leading to misallocations of real resources and reductions in aggregate economic welfare. 29 The final three sources of potential loss occur after the date of official resolution when the institution is placed in receivership. Losses to depositors from delays in receiving reimbursement and liquidating bank assets may be either credit/market losses or liquidity/ present value losses or both. Before insured depositors can be paid, their identities and amount of qualifying deposits must be determined and certified. Before uninsured depositors can be advanced the value of their claims, they also must be identified and certified and the recovery value of the bank assets estimated. The length of these operational delays depends on the state of information (record-keeping) technology in use and represents a potential liquidity or present value loss. The fourth source of loss is a credit loss that arises because of attempts, legitimate or not, by the receiver to avoid fire-sale losses or depressing asset prices by selling quickly into perceived temporarily weak markets, from self-dealing by the receiver, or legal obstacles that prevent the receiver from disposing of assets quickly. The fifth and last source of loss from delays in distributing the funds from the sale of the assets of the bank is primarily a liquidity/present value loss to depositors from operational inefficiencies by the receiver. Implications of post-resolution delayed depositor access to funds Unlike the two sources of losses at the date the institution is legally declared insolvent and placed in receivership, which have been analyzed frequently, the three sources of depositor losses afterwards and the speed with which depositors gain access to their funds have been analyzed only infrequently.6 As noted earlier, at the time of resolution, insured depositors have claims for the par value of their deposits (adjusted for any coinsurance) at the date of resolution and uninsured depositors for the present value of the estimated pro-rata recovery value of their deposits. In the absence of an efficient secondary market, delay in offering depositors full access to their permissible funds decreases the liquidity and, in the absence of interest payments, the present value of the deposit claims and greatly intensifies both public fears and actual costs of bank failures. As noted by the Swedish Central Bank (Riksbank): Freezing a companys assets and suspending its payments from the time the bankruptcy order is issued could have serious implications if applied to banks. A banks liabilities do after all form an active part of its business operations, and its borrowing and interbank funding activities reflect among other things the banks central role in the payment system. Suddenly freezing the repayment of the liabilities at one or more big banks could have immeasurable 30 consequences for the banking system as a whole (Viotti, 2000, p. 55). Moreover, the fear of such inaccessibility to ones account is likely to have important political as well as economic consequences. Affected depositors are more likely to demand full and immediate access to their funds, and regulators and governments are likely to bow to the political pressures and both delay official recognition of insolvency (forbear) and fully protect more if not all depositors (too big to fail) if and when insolvency is finally declared. At the same time, the government itself is likely to view any loss in depositor liquidity as potentially detrimental to the aggregate economy and may be reluctant to permit conditions that would trigger this loss. Thus, it may maintain insolvent institutions in operation and protect all depositors and possibly other creditors in full. This strategy is likely to increase the ultimate cost of the losses to the government. Moreover, such response further reduces market discipline and encourages additional moral hazard behavior by the banks. Reducing potential losses to depositors The adverse effects from bank failure can be reduced by reducing losses from any or all of the above five sources to both depositors and the deposit insurance agency. Indeed, if troubled banks could be resolved before the market value of their equity capital turned negative, losses would be restricted only to shareholders. Depositors would be unharmed. Little, if any, more serious adverse effects would then be felt from bank failures than from the failure of any other firm of comparable size. Failures could be freely permitted to weed out the inefficient or unlucky players. Deposit insurance would effectively be redundant. In the U.S., the Federal Deposit Insurance Corporation Improvement Act (FDICIA) attempts to reduce the first two sources of losses through prompt corrective action (PCA), which both imposes a more efficient closure rule2 percent tangible equity to asset ratio and reduces regulatory discretion to forbear by requiring mandatory sanctions on financially troubled institutions. These include resolution when the discretionary sanctions applied appear to be ineffective as reflected in a continued decline in the banks capital ratio. We describe how the FDIC reduces the third source of lossinsufficient information and processing delayin the next section. The fourth source of loss, bad market conditions, could be reduced by careful monitoring by the appropriate agency of the receivers motivations or justification for delaying selling bank assets. This monitoring would verify 1) that the probabilities are sufficiently 2Q/2002, Economic Perspectives high that relevant asset markets are only temporarily depressed and may be expected to recover shortly; and 2) that the assets can be managed efficiently in the meantime, so that the present value of the projected sales proceeds to depositors and the deposit insurance agency will be higher than without a delay. Recent experience in most countries, including the United States, suggests that delays in asset sales, although often politically popular, rarely produce financial gains (Kane, 1990, and Gupta and Misra, 1999). Thus, it may be desirable to specify timely sales schedules. The fifth source of lossinefficient receivercould be reduced by requiring receivers to distribute their proceeds more quickly as they are received and monitoring and enforcing their compliance with this policy. Procedures for immediate and full payment of depositor claims at resolution If losses are incurred in resolving an insolvency, governments, out of fear of political pressure by depositors for bailouts or of systemic risk, may prefer to provide depositors with immediate and full access to their claims at the time of resolution when the institution is legally placed in receivership. To do so, the deposit insurance agency can accelerate the identification of the depositors and the value of their claims and advance funds to them before it is paid by the receiver or encourage the development of an efficient secondary market in the claims. The U.S. appears to be one of the very few countries that generally does not freeze accounts at failed banks when they are resolved. Except in unusual instances, the FDIC provides all depositors with almost immediate and full access to the value of their claims at resolution, based on losses from poor closure rules and regulatory forbearance, so that there is no loss of either liquidity or present value from post-resolution sources (FDIC, 1998a).7 The FDIC advances the funds. Although it may not receive full and immediate payment for all the assets in the resolution of a failed bank, the FDIC typically advances the pro-rata present value of the estimated recovery value through an advance dividend payment to all depositors at domestic offices of the bank on or about the next business day after its appointment as receiver.8 In addition, for insured and ex-post protected deposits, the FDIC advances the difference between the par value of the account and the present value of the estimated recovery amount, so that these depositors receive the par value of their deposits. The FDIC does not advance uninsured depositors a dividend equal to the estimated recovery amount primarily in cases where it cannot quickly obtain reliable estimates of the recovery value of the assets. Federal Reserve Bank of Chicago Payment of insured deposits is either at the bank that assumed the insured deposits of the resolved banks or, if the insured deposits are not assumed by another bank, at the site of the failed bank operating in receivership.9 Payment of the advance dividend on de facto unprotected deposits at domestic offices, which is generally for less than par value, is at the failed bank, unless these deposits are assumed by another bank at par value.10 However, since 1992, the least cost resolution provisions of FDICIA have made assumptions of uninsured deposits by another bank unlikely, unless there is no or next to no loss to the FDIC in the transaction.11 The FDIC can make funds available quickly because it has the legal authority to advance the funds and it has mostly solved the technical problems that underlie delays in payments after resolution. As noted earlier, to give the FDIC sufficient time to prepare for these payments and transfers, including identifying the owners and total of eligible accounts, banks are generally declared insolvent at the end of business on Thursdays or Fridays, and depositors are given access to their funds on the following Monday. Reliable estimation of recovery values of bank assets, however, often requires longer than a weekend. And examiners and supervisors in the U.S. are typically provided with additional time. Under FDICIAs prompt corrective action, bank examiners and supervisors are effectively required to progressively increase their familiarity with a bank as soon as its financial situation deteriorates to the extent that it becomes classified as undercapitalized, including increasing the frequency of on-site visits. Moreover, when a bank is considered in imminent danger of failing, is declared critically undercapitalized, or is being resolved for other reasons by its primary federal or chartering regulator, the FDIC is notified in advance and prepares for a possible sale of all or part of the bank to other institutions at auction at the highest price (FDIC, 1998c). To do this, it has to prepare detailed financial information on the bank to be provided on a confidential basis to potential bidders prior to the auction and to gather the information needed to make the determination as to which of several resolution alternatives will be least costly to it. Thus, the FDIC typically sends its resolutions staff into the bank some days prior to its being closed to collect the needed information (FDIC, 1998a). The data collected are used to arrive at both market valuations for the assets of the bank and estimates of the number and holdings of insured depositors and other creditor classes. As a result, except in the case of major fraud, the FDIC is generally able to estimate recovery values reasonably accurately before the bank is legally resolved and put in receivership, 31 and the deposits need not be frozen after closure while the magnitude and impact of the payout are being estimated.12 If, after recovery is completed, the proceeds to the FDIC exceed the amount it advanced the uninsured depositors, the depositors are paid the difference up to the par value of their claims plus interest. Any remainder is paid to more junior creditors and eventually to shareholders. If the proceeds fall short of the amount the FDIC advanced to the uninsured depositors, the FDIC bears the loss. Thus, to protect itself, it advances to the uninsured depositors only a conservative estimate of the present value of the recovery value.13 History of immediate and full payments of depositor claims Immediate and full access for all depositors, or even for only ex-post protected depositors, to their permissible funds has not always been the practice of federal deposit insurance agencies in the U.S., has not been the practice of state insurance agencies in the U.S., and is not the current practice of deposit insurance agencies in most other countries. In large measure, the delayed access, particularly for protected depositors, reflects the inability of the insurance agency both to legally advance payment to depositors before receipt from the receiver and to collect and analyze in a timely fashion the necessary information on what balances and which depositors are insured and on estimates of recovery values, as well as the inability to establish paying agents quickly. The information on eligible insured deposits is complex because of, among other things, poor and/or noncomputerized records and depositor ownership of multiple accounts at the same bank. These obstacles provide a physical rather than a policy reason for not providing immediate and full access to both protected and unprotected depositors. Before the establishment of the FDIC in 1934, depositors at failed banks, even in states with state insurance programs, had all or part of their accounts frozen and were generally paid only as the assets were liquidated and funds collected (FDIC, 1998b, and Mason, Anari, and Kolari, 2000).14 The delay in liquidating a failed banks assets and paying the depositors averaged nearly six years (Bennett, 2001). Even when the FDIC was initially established, it did not pay insured depositors immediately. The FDICs Annual Report for 1934 explains that Payments of the insured portion of depositors claims against the banks which closed during 1934 were started promptly after the receiverships began. The interval between the appointment of the receiver and the first payment to 32 insured depositors varied from 2 to 22 days, the average being seven days. Upon notification of suspension, preparations were begun for payment of the insured deposits. Before payment can be made an analysis of the deposit liabilities of the closed bank is necessary. Balances due to depositors in the various classes of deposit accounts carried by the bank must be brought together in one deposit liability register, in order that the net insured deposit of each depositor in each right and capacity may be determined, as required by law. After the period in which the stockholders might enjoin the State authorities from placing banks in liquidation had expired, depositors were paid as rapidly as their claims were presented. (FDIC, 1935, p. 26). Similarly, before the mid-1960s, the former Federal Saving and Loan Insurance Corporation (FSLIC), which insured savings and loan (S&L) associations before the FDIC, often disbursed funds to insured depositors at failed S&Ls only slowly through time; and before the early 1980s, the FDIC did not advance payments to unprotected depositors (FDIC, 1998a).15 Likewise, Ohio, Maryland, and Rhode Island, states that experienced widespread failures of perceived state insured thrift institutions in the 1980s, generally reimbursed insured depositors at these institutions in full, but only slowly over a number of years, so that depositors suffered significant present value losses and liquidity costs (Kane, 1992; Pulkkinen and Rosengren, 1993; and Todd, 1994). Contrary to current FDIC practice, the insured depositors in these states were effectively insured in future or nominal values only, not in present values. Full and immediate depositor access does not exist in most other countries.16 For example, the Canadian Deposit Insurance Corporation provided depositors of the failed Confederation Trust Company in 1994 with access to the insured portion of their deposits 52 days after the bank was declared legally failed, although faster advance payments were made in cases of critical need (Canada Deposit Insurance Corporation, 1995). Article 10 of the Directive of the European Union (EU) dealing with deposit-guarantee schemes, which became effective on July 1, 1995, requires that each member countrys national insurance agency pay insured depositors within three months of the date on which the competent authorities make the determination that the bank is unable to repay its deposits in full and deposits become unavailable to the depositors. But, this period may be extended for three three-month periods to a maximum of 12 months if necessary in exceptional circumstance. 2Q/2002, Economic Perspectives These delay schedules appear to have been imposed to limit the maximum delay due to obtaining and processing the relevant deposit data and to encourage faster payment, rather than to prolong delay in order to increase market discipline. No harmonizing directive applies to the treatment of uninsured depositors and other creditors in the EU. This is left to the laws of the individual countries. The competent authority that can declare an institution insolvent and the authoritys powers are also determined by each country. In general, private receivers are appointed to sell or liquidate the bank. The unprotected claimants are paid the recovered values as they are collected and distributed by the receiver.17 In most instances, this process is not fully completed for many years, so that depositors do not have access to the full recovery value of their claims for an equal number of years. Advantages and disadvantages of immediate and full payment of depositor claims Immediate and full payment of insured and uninsured depositor claims has both advantages and disadvantages. The major advantage, particularly for uninsured depositors, is that it may forestall political pressure by depositors on their governments to delay resolving insolvent banks and to make all depositors completely whole when they do. Moreover, by not requiring banks to be effectively physically as well as legally closed, speedy payments also reduce the potential damage to the macroeconomy and reduce the need for the government to provide guarantees. Thus, TBTF appears alive and well in most countries outside the U.S., which generally do not provide for such speedy payments. Indeed, before the enactment of deposit insurance in the U.S. in 1933, Senator Carter Glass, the influential chairman of the Senate Banking Committee at the time, had proposed more rapid payment to depositors at failed banks as a superior alternative to insurance (Bradley, 2000; Kennedy, 1973; and Willis and Chapman, 1934). In describing the Glass proposal, Willis and Chapman (1934, pp. 6567) write: It was a fact that the receiverships were in the habit of extending anywhere from a few months to as long as twenty-one years. Recognizing that in bank failures the source of difficulty and losses is not primarily found in lack of assets, but ... that the resources of depositors are tied up and rendered unavailable for long periods ... liquidation power and not guaranty was demanded ... insuring an almost immediate settlement within a short time upon the basis of the estimated worth of the [failed] banks Federal Reserve Bank of Chicago assets. ... This plan was considered by the [Banking] Committee entirely adequate to the protection of the bank depositor against most of the evils to which he had been subject, while leaving him still with a measure of individual responsibility for the protection of his claims through the selection of a well-qualified bank. The plan called for the establishment of a federal government liquidating corporation that would estimate a banks recovery value immediately upon its failure, quickly sell the bank as a whole or in parts, and quickly pay the proceeds to the receiver for speedy disbursement to the depositors. But this plan was found too difficult to implement at the time, primarily because it required accurately estimating the market value of the failed banks assets quickly. However, the advantages of such a scheme had also been seen by others, particularly during the banking crisis of the early 1930s, when nearly 10,000 banks, or some 40 percent of the total number of banks, failed. For example, in 1931, the Federal Reserve Bank of New York attempted to have depositors at failed banks receive the recovery value of their claims faster by requesting healthy member banks to buy the assets of failed banks and advance the proceeds to them for immediate distribution (Bradley, 2000, and Friedman and Schwartz, 1963). This proposal does not appear to have been successful. In 1933, the New York State Banking Department entered into agreements with several large New York City banks to partially assume the deposits of failed banks and be reimbursed from the liquidation of a corresponding amount of assets. At the same time, the Reconstruction Finance Corporation began to loan funds to closed banks to make quick partial payment to depositors (Kaufman, 2002a). But providing immediate depositor access to the full value of their permissible funds may also have important disadvantages; in short, it may be a doubleedged sword. It may reduce market discipline on the banks. Knowing that they face a delay, and at times a very lengthy delay, in gaining access to the full value of their claims after resolution, both insured and uninsured depositors have a greater incentive to monitor the financial health of their banks and to discipline them when necessary by charging higher interest rates commensurate with the greater perceived risk or transferring their deposits (running) to perceived safer banks.18 Immediate payment would reduce this incentive. In addition, under full and immediate access as practiced by the FDIC, any unexpected losses from delays in asset sales and distribution of the sales proceeds will accrue to the deposit insurer rather than to the unprotected depositors. This would further reduce the incentive for 33 unprotected depositors to monitor their banks. We model the tradeoff between the advantages and disadvantages of full and immediate access in the next section to examine the implications more carefully and to identify the optimal time delay in providing depositors with full access. Modeling the access delay decision expected loss from bailout pressure by the maximum amount. In figure 1, where the two schedules are shown as crossing, this is shown as Q. If instead the additional market discipline schedule lies above the bailout pressure schedule at all points from the date of resolution, the optimal delay time is infinite. If the bailout pressure schedule lies above the additional market discipline schedule at all points, the optimal delay time is the date of resolution. This would imply that accounts should not be frozen at all and depositors should be given immediate access to the value of their claims. If an inability to advance payment or technical problems prevent the government from providing depositors with access at the optimal time, the government is likely to bail out all stakeholders and keep the bank in operation. This reinforces the importance of both resolving institutions as quickly as possible with no or minimum loss and developing faster procedures for certifying protected deposits and estimating recovery values. It follows that by providing depositors with immediate and full access to their claims, as described earlier, the U.S. implicitly assumes that additional potential losses from bailout pressures immediately exceed potential gains from additional market discipline. As discussed above, the primary basis for reducing the cost of failure to depositors by advancing them funds immediately after a bank failure is to minimize the economic disruption that can result from the loss of liquidity associated with freezing deposits. However, there is a clear tradeoff with market discipline. On the one hand, the greater the perceived loss that insured or uninsured depositors may potentially suffer, the greater their incentive to monitor their banks condition and discipline the bank for taking excessive risks, either by withdrawing funds or by requiring higher interest rates to compensate for the increased risk. On the other hand, the greater the expected loss in either value or liquidity, the greater the public pressure will be for government protection of most if not all stakeholders. This is likely to increase the cost of resolution to the government. Given this tradeoff, it is The FDIC survey of depositor access possible to solve for the optimal time for the distribupractices across countries tion of payments on depositor claims on a failed bank. We can model this tradeoff graphically. Because the govIn February 2000, the FDIC surveyed 78 deposit ernment can affect, if not set, the delay time, includinsurers in 64 countries outside the U.S. on aspects of ing the time necessary to process the relevant deposit their deposit insurance systems. The countries chosen data and estimate the recovery values, it effectively were those that had explicit deposit insurance schemes serves as a policy tool. in place. Thirty-seven surveys were returned, providing Our model is shown in figure 1. The time delay in the insurance agency proFIGURE 1 viding depositors with full access to the Effects of additional market discipline and bailout pressure value of their claims after resolution or (as functions of depositor payment delay time) the length of time accounts are frozen Absolute value (payment delay) is measured on the horiof additional zontal axis. The reduction in expected expected loss or gain Expected loss (or gain) from additional market disloss from cipline and the increase in expected loss bailout pressure from intensified bailout pressure are measured on the vertical scale. These are shown in absolute terms. In the absence Expected gain of an efficient secondary market for from market discipline depositor receivership claims, both the reduction and increase in expected loss • from additional market discipline and • bailout pressure, respectively, may be • · expected to increase with the delay time. • The optimal delay time occurs when the • Payment reduction in expected loss from additional delay Q 0 time market discipline exceeds the increase in 34 2Q/2002, Economic Perspectives insight into the deposit insurance practices of 34 countries.19 While the surveys covered a wide range of deposit insurance practices, this article examines only that portion of the survey relating to the availability of funds to depositors after a bank has been declared insolvent and differences in the treatment of insured and uninsured depositors.20 When examining fund availability practices, one must recognize the difference between policy intent and practice. A deposit insurer may wish to pay quickly, but not have the legal, technical, or informational capacity to do so. Conversely, the authorities may believe in instilling market discipline by imposing costs on depositors through delayed access to funds, but may not have the political resolve to carry out such a policy. Consequently, we analyzed only the 30 responding countries that had actually experienced bank failures since 1980. Of these, three (Bahrain, Jamaica, and Sweden) did not specify a time frame within which they had paid depositors, since the failures occurred prior to the creation of a deposit insurance scheme. Insured deposits As table 1 shows, only three countries (Japan, Italy and Peru) provided immediate payment of insured deposits. Japan has protected all depositors in those banks that it has declared insolvent to date and used resolution techniques that provided for immediate access to funds. In Italy, the Interbank Deposit Protection Fund also provided insured depositors with immediate access to their insured deposits. Peruvian depositors have had access to some but not all of their insured deposits in some failures the day after failure, for example, in the most recent failure in November 1999. But in other failures, the depositors have had to wait as long as eight months for even the initial payment. According to the Peruvian Deposit Insurance Fund, the factors that determine the speed with which insured depositors get access to their funds are the potential systemic effects that would be triggered by the failure of a specific bank and the quality of information given to the insurer by the liquidation agency. Five other countries gave insured depositors access to their funds within one month of the failure, and the majority of all respondents followed the EU guidelines and gave insured depositors access within no more than three months. The Isle of Man Financial Supervision Commission was still in the process of attempting to pay off insured depositors more than six months after the failure of a bank in 1999. Three other countries, Poland, the Czech Republic, and Greece, reported that they were able to make funds available to insured depositors within six months. It is interesting to note that Federal Reserve Bank of Chicago almost all of the respondents provided insured depositors with all their funds at one time. Only the deposit insurers in Italy, Austria, Latvia, and Peru paid in installments. The responses from Peru and the experience of the Isle of Man suggest that much of the reason for the delay in paying insured depositors may not be a conscious policy of promoting insured depositor discipline. Rather, it reflects the technical difficulties associated with paying off a bank quickly. Uninsured deposits The survey results presented in table 2 clearly indicate that the practice of advancing funds to uninsured depositors is largely unique to the United States. Twenty-three of the respondents indicated that uninsured depositors cannot be fully protected at failed banks in their countries, and only three deposit insurers (Canada, Japan, and Slovakia) indicated that they had the power to advance funds to cover uninsured depositors. The timing of availability of funds to uninsured depositors is typically dependent on the type of resolution. Japan and Tanzania are notable examples of countries that have used resolution techniques to protect all depositors. In other countries, such as Italy and Brazil, uninsured depositors have immediate access to their deposits if a resolution results in the transfer of these deposits to another financial institution. In most countries, unprotected depositors have to wait for the liquidation process to yield sufficient cash for payments to be made to them. The practices surrounding the liquidation of assets and payment of claims follow the national practices for bankruptcy, with discretion vested in the courts or the liquidator, receiver, or administrator for the failed bank estate. In all cases where the uninsured depositors were dependent on a liquidation process for their proceeds, they received access to their funds in installments. A review of the comments received from the respondents suggests that, while most deposit insurers do not have the discretion to protect uninsured depositors in liquidations or to advance funds from their deposit insurance funds to uninsured depositors, they can use resolution strategies that protect uninsured depositors. This suggests that these countries will probably resort to keeping insolvent banks in operation through nationalization in whole or in part and/or extending blanket guarantees to depositors. Conclusions and recommendations This article identifies and analyses five potential sources of loss to depositors in bank failures, two that are recognized at the time an insolvent bank is resolved and placed in receivership and three that 35 TABLE 1 Funds availability, insured deposits Country Regulation/ Immediate laws payment At least 1 insolvent bank since 1980 Austria (1) Bahraina Belgium Brazil Canada Czech Republic France Germany (1) Greece Hungary Isle of Man Italy (1) Italy (2) Jamaicaa Japan Latvia Lithuania Netherlands Nigeria Peru Poland Romania Slovakia Spain Swedena Tanzania Trinidad and Tobago Turkey Uganda United Kingdom No insolvent banks since 1980 Austria (2) El Salvador Germany (2) Mexico Oman Portugal Taiwan Yes No Yes No No Yes Yes No Yes No No Yes Yes Yes No No Yes Yes No Yes Yes Yes Yes Yes Yes No Yes No Yes Within 7 days Within 1 month Within 3 months Within 6 months >6 months Yes Payment Installments Yes Yes Yes All at one time All at one time All at one time All at one time All at one time All at one time All at one time All at one time All at one time Installments Installments Yes Yes Installments All at one time All at one time Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Installments All at one time All at one time All at one time All at one time All All All All All at at at at at one one one one one time time time time time Yes Yes Yes Yes Yes Yes No a Denotes countries whose failures occurred prior to the establishment of the current deposit insurance scheme. Note: For countries with two deposit insurance funds, the number in parentheses following the country name indicates which fund dealt/did not deal with bank failure. For example, in the case of Austria, deposit insurance fund 1 has dealt with an insolvent bank since 1980, while deposit insurance fund 2 has not dealt with any bank failures in that period. Source: Federal Deposit Insurance Corporation. occur afterwards. The three sources of post-resolution losses arise from delayed payment of depositor claims which may lead to losses in credit value and/or liquidity. The loss of liquidity through the effective freezing of some or all of the deposits by the deposit insurance agency, pending reliable data on what deposits and depositors are protected and/or the receipt of the proceeds from the sale of bank assets, has two conflicting effects. On the one hand, fear of delayed payment increases monitoring and discipline by depositors. On the other hand, fear of delayed payment 36 increases pressure from depositors for protection and government willingness to supply such protection to reduce the chances of systemic risk. This article analyzes these effects. Countries follow different practices with respect to delaying payment, with different consequences for market discipline and resolution policies. In the U.S., the FDIC currently does not generally freeze deposits at resolved institutions. Rather, it effectively advances the proceeds to depositors at the time of resolution, frequently before it collects them from asset sales in its capacity as 2Q/2002, Economic Perspectives TABLE 2 Funds availability, uninsured deposits Regulation/ laws Uninsured can be fully protected Yes Yes No Yes Yes Yes No No No No Yes No No Yes No Yes No Yes No No No No Yes Italy (2) Jamaicaa Japan Latvia Lithuania Netherlands No Yes Yes No Yes No No Yes No No Yes Nigeria No Yes Peru Poland Romania Slovakia Spain Swedena Tanzania Yes Yes Yes Yes Yes No No Yes Yes No Yes Yes No Yes Trinidad and Tobago Turkey Yes No Uganda United Kingdom Yes Country At least 1 insolvent bank since 1980 Austria (1) Bahraina Belgium Brazil Canada Czech Republic Republic Germany (1) France Greece Hungary Isle of Man Italy (1) Deposit insurer can advance funds Yes Payment schedule Time before accessing 5–6 months Installments Several months Depends on intervention Not permitted No bankruptcy proceedings have finished yet. Installments Installments None No No 2 years Installments Installments Installments Resolution method affects schedule No Yes No Yes Yes Yes No Yes Yes Immediate access if assets and liabilities assigned to another institution; otherwise wait until receiver allocates assets. Yes All deposits protected so far Installments Installments Installments No No Yes No No 12 months Normal bankruptcy laws between receiver and uninsured depositors; if funds available for creditors of their rank, paid out in due course. No provision for depositors of insolvent banks to be paid from Deposit Insurance Fund. 0–1 year Installments Installments No case Approximately 12 months Installments Full compensation; depositors All at one time had access to their deposits within the shortest period. Whenever sufficient funds from Installments realization of assets are available. Since 1980, depositors unable All at one time to access explicitly uninsured deposits. No No Yes Yes Yes Yes Yes No Yes No Yes No No No insolvent banks since 1980 Austria (2) El Salvador Yes Yes No No Germany (2) Mexico Oman Portugal No No Yes Yes Yes No No No Taiwan No No No Handled by liquidators or administrators. No bank failure Bank failures, but no insured deposits system No bank failures All at one time Yes No Installments No Yes Yes No No explicitly uninsured depositors prior to 1999. No order to close a financial institution during the past 15 years. a Denotes countries whose bank failures occurred prior to the establishment of the current deposit insurance scheme. Note and source: see table 1. Federal Reserve Bank of Chicago 37 receiver. Thus, insured depositors receive near immediate payment of the par value of their deposits and uninsured depositors generally receive near immediate payment of the present value of their pro-rata share of the estimated recovery value. This practice may reduce market discipline, but it may reduce bailout pressure even more. If so, given the loss at resolution, insolvent institutions are more likely to be resolved and uninsured depositors not protected. In contrast, most other countries freeze deposits and delay payments to both insured and uninsured depositors, according to a schedule or until the funds are collected from asset sales, both because of the inability to estimate quickly the amount that needs to be paid out and because of restrictions on advancing funds before collection of the sales proceeds. These differences in the treatment of depositors at insolvent institutions have important implications for a countrys bank resolution practices, in particular, for banks considered too big to fail. The smaller the perceived overall loss in bank failures, the easier it is economically and politically to resolve insolvencies with losses to de jure unprotected depositors. In the U.S., if regulatory prompt corrective action is successful in limiting losses (negative net worth) to relatively small amounts, say, to not more than 5 percent of assets at large banks (the loss experienced by the Continental Illinois National Bank in 1984 was about 4 percent) and uninsured depositors have immediate and full access to their funds, then losses to large uninsured depositors would be restricted to a rate that is well within the boundaries that most of these depositors can tolerate without panicking (for example, losses they appear to be willing to bear in commercial paper or other short-term debt investments). Moreover, since enactment of depositor preference, which subordinates deposits at foreign offices and other creditors to domestic deposits and the FDIC, losses at failed banks can be charged to these accounts before domestic depositors. Thus, losses to domestic depositors and the FDIC may be even smaller. As a result, if the losses are small and access to the remaining deposits is immediate, uninsured depositors are less likely to exert political pressure on the government to extend the safety net to them, governments are less likely to be fearful of systemic risk, and too big to fail protection may be avoided. The combination of the FDICs payment practices and the improved closure rule under FDICIA helps to explain why uninsured depositors at almost all recently failed banks in which the FDIC suffered losses have been required to share pro-rata in the losses (Benston 38 and Kaufman, 1998). But, because no large money center bank has failed since FDICIA, the systemic risk exemption under FDICIA has not been invoked, and it is too early to declare TBTF dead in the U.S. Nevertheless, speedy payment to depositors is likely to reduce the need for its use. In contrast, most other countries may find it more difficult to resolve large insolvent banks with losses to depositors, because these losses are not necessarily minimized and uninsured deposits are often frozen until payment is received from private receivers. These countries governments are thus under greater pressure to protect all depositors and are more fearful of igniting systemic risk if they do not. Thus, TBTF appears to be alive and healthy in these countries, and taxpayer losses in bank failures may be expected to be relatively larger. Because cross-country differences in insured depositors access to their funds affects both the intensity of market discipline and the probability of government bailout, cross-country studies of the effectiveness and efficiency of alternative deposit insurance structures that specify the existence of such programs or differentiate between explicit and implicit programs only by a single yes/no (or 1/0) variable, and thus omit reference to access delay, are likely to be incomplete and inaccurate. Our analysis in this article suggests that the best strategy for achieving aggregate bank stability, characterized by efficient exit of inefficient or unlucky banks through failure at no or least cost to the economy, involves resolving these banks before or shortly after their net worth turns negative and providing full and immediate (or near-immediate) access for insured depositors to the par value of their deposits and for uninsured depositors to the present value of their prorata share of the estimated recovery value at resolution. Such a strategy minimizes the potential for systemic risk and permits otherwise TBTF banks to be resolved just like any other insolvent bank. However, the ability to provide full and quick depositor access may be constrained both by lack of legal authority for regulators to advance payment before receiving the funds from asset sales and by technical problems that interfere with this outcome, such as the unavailability of accurate and accessible account data and facilities for speedy analysis of the data and the inability to estimate recovery values accurately and quickly. If this is indeed the optimal policy, policymakers in each country need to develop procedures for reducing the delays caused by these problems. 2Q/2002, Economic Perspectives NOTES 1 Too big to fail in the United States does not imply that the bank has not failed. All resolved banks since shortly after the resolution of the Continental Illinois Bank in 1984 have been legally failed. Rather, a large insolvent bank may be too big not to protect some or all noninsured stakeholders when failed or too big to liquidate quickly and, therefore, may be kept in operation temporarily, protecting all creditors during the delay (Kaufman, 1990 and 2002b). This interpretation was recently reinforced by Federal Reserve Chairman Greenspan (2000), who stated that the issue is that an organization that is very large is not too big to fail, it may be too big to allow to implode quickly. But certainly, none are too big to orderly liquidate ... and presumably, not to protect non-guaranteed deposits from loss. Since the enactment of the Federal Deposit Insurance Corporation Improvement Act (FDICIA) in 1991, TBTF may more accurately be termed the systemic risk exemption. Periodic restricted depositor access to accounts is common in many countries, for example, in Argentina during the recent currency crisis, and was so historically in the United States during a general banking crisis to reduce conversion into specie or foreign currency, even if the banks may be solvent, for example, in the U.S. during the banking panics of 1893 and 1907. 2 For example, in November 2000, Nicaragua resolved its second bank in 100 days and guaranteed deposits of less than 20,000 cordobas (about $1,500) at the second bank. But only 10,000 cordobas would be paid within five days; the rest would be paid as the banks assets were soldAngry customers gathered outside the closed branches of Bancafe yesterday shouting thieves and vampires, (Financial Times Limited, 2000). 3 4 In addition to losses in liquidity, depositors in many countries also fear partial or complete expropriation of deposits at failed institutions by the government beyond the pro-rata share of any losses. In many countries, banks have not always been very secure depositories for funds and, indeed, have often been perceived as less secure than mattresses. Berger and Udell (2002) have recently speculated that loan relationships are more with the loan officer than with the bank. 5 Speedy payment for insured depositors at failed banks is listed by Garcia (1999) as one of her 15 best practices for a deposit insurance system, but there is no further analysis of this practice nor any discussion of payment of noninsured deposits. Hall (2001) reports on payment practices by European Union countries for insured deposits only, but with no further analysis. 6 Nevertheless, casual evidence suggests that at least some depositors, including fully insured depositors, are still concerned that they may find their deposits at failed banks temporarily frozen. 7 Because the FDIC is generally appointed receiver, it can better estimate losses from delayed sales and need not be concerned with delayed distributions. 8 In those instances where no bank acquires the insured deposits and there are a large number of depositors, the FDIC will either arrange for another bank to act as its deposit transfer agent or it will mail checks to depositors for the insured amounts. 9 Under the Depositor Preference Act of 1993, unsecured depositors at foreign offices of U.S. banks and other creditors, such as fed funds sellers, have claims junior to those of domestic depositors and, unless the too big to fail provision of FDICIA is 10 Federal Reserve Bank of Chicago invoked, will be paid the recovery value of their claims only as the banks assets are sold and all senior claimants have already been paid (Kaufman, 1997b). 11 Before FDICIA, the FDIC generally protected all depositors, including de jure uninsured depositors, particularly at larger banks, through merger (purchase and assumption) with another bank that assumed all deposits at par and received a payment from the FDIC (Benston and Kaufman, 1998, and FDIC, 1998a). In addition to speedy payment of depositor claims, the FDIC also attempts to resolve insolvencies with minimum disruption to either bank customers or financial markets. As noted, unless there is no demand for banking services in the community served or the bank is so severely impaired that there is little or no redeeming financial value, insolvent banks are sold or merged and open for business the next business day after resolution. If additional time is necessary to find a buyer, the FDIC can charter a bridge bank to temporarily continue the business in a new entity. Thus, liquidations with serious disruptions in banking services are rare and likely only for relatively small banks. This practice also reduces pressures for government support of insolvent institutions and is likely to reduce losses to depositors from delayed resolution. 12 Because the FDIC pays the full par amount of insured deposits, incorrect estimates of the recovery values affect only the final allocation of its costs, not the total cost of these payouts. However, the FDIC would suffer a loss if it overestimated the recovery value and transferred the uninsured deposits to an assuming bank that offered a premium that was larger than the estimated loss rate at the time but, ex post, was smaller than the loss rate that was actually realized and reported. In retrospect, it would have been cheaper to the FDIC if it had paid off the uninsured deposits. 13 Note holders at failed national banks were paid the par value of their notes immediately by the U.S. Treasury (FDIC, 1998b). In addition, during bank panics, accounts at all banks in the affected area were frequently partially frozen to limit conversions into specie or currency. For example, Kelly and O Grada (2000, p.1113) note that on October 12[, 1857, New York] savings banks invoked a rarely imposed clause in their articles of agreement limiting withdrawals on demand to 10 percent of the outstanding balance. As noted earlier, a similar constraint was recently imposed on banks in Argentina. 14 The concept of advancing payment to uninsured depositors appears to have been developed by the FDIC in the early 1980s as part of its proposal for modified payoff resolutions, in which an existing or newly chartered bank would assume all the insured deposits of a failed bank in full and all the uninsured deposits partially in an amount equal to the estimated recovery value as reflected in the advanced dividend (FDIC, 1983, pp. III 45 and FDIC, 1997, p. 250). The policy may have been modeled on a number of earlier actual or proposed plans, which we discuss later in the article. Advance dividends were paid in 13 resolutions in 1983 and 1984 and again starting in 1992. The dividend was generally funded by a loan from the FDIC corporate account to the FDIC receiver account (FDIC, 1998a, and FDIC, 1997). 15 As is discussed later, only three (Italy, Japan, and Peru) of the 25 countries other than the U.S. that responded to a survey by the FDIC and that had experienced at least one bank failure since 1980 reported paying even their insured depositors immediately. 16 39 Only three countries in the FDIC survey (Canada, Japan, and Slovakia) report having authority to advance funds to uninsured depositors at failed banks, but few countries responded to this question. 17 A recent study of depositor behavior in Argentina, Chile, and Mexico in the early 1990s found that insured as well as uninsured depositors disciplined riskier banks both by charging higher deposit rates and by withdrawing deposits (Peria and Schmukler, 2001). Among other possible reasons the authors note for this unexpected behavior by insured depositors is potential delays in receiving payment. Likewise, Demirgüç-Kunt and Huizinga (1999) report finding evidence of market discipline in a large number of countries that have government provided safety nets, but do not list delayed payments as one of the possible reasons. 18 19 Austria, Germany, and Italy have more than one deposit insurer. 20 Other results from this survey are discussed in Bennett (2001). REFERENCES Barth, James R., 1991, The Great Savings and Loan Debacle, Washington, D C: American Enterprise Institute. Bennett, Rosalind L., 2001, Failure resolution and asset liquidation: Results of an international survey of depositors, FDIC Banking Review, Vol. 14, No.1, pp. 128. Benston, George J., and George G. Kaufman, 1998, Deposit insurance reform in the FDIC Improvement Act: The experience to date, Economic Perspectives, Federal Reserve Bank of Chicago, Second Quarter, pp. 220. ington, DC. , 1998c, Resolutions Handbook, Wash- , 1997, History of the Eighties: Lessons for the Future, Washington, DC. , 1983, Deposit Insurance in a Changing Environment, Washington, DC. ton, D.C. , 1935, Annual Report, 1934, Washing- Financial Times Limited, 2000, Managua faces crisis with collapse of another bank, Financial Times, November 21. Berger, Allen N., and Gregory F. Udell, 2002, Small business credit availability and relationship lending, Economic Journal, Vol. 112, No. 477, February, pp. 255277. Freidman, Milton, and Anna J. Schwartz, 1963, A Monetary History of the United States, 18571960, Princeton, NJ: Princeton University Press. Bradley, Christine M., 2000, A historical perspective on deposit insurance coverage, FDIC Banking Review, Vol. 13, No. 2, pp. 125. Garcia, Gillian G. H., 1999, Deposit insurance: A survey of actual and best practices, International Monetary Fund, Washington, DC, working paper, No. 99-54, April. Canada Deposit Insurance Corporation, 1995, Annual Report, 19941995, Ottawa. Dermine, Jean, 1996, Comment, Swiss Journal of Economics and Statistics, December, pp. 679682. Demirgüç-Kunt, Asli, and Harry Huizinga, 1999, Market discipline and financial safety net design, World Bank, Washington, DC, working paper, July. Federal Deposit Insurance Corporation, 1998a, Managing the Crisis: The FDIC and RTC Experience, Washington, DC. , 1998b, A Brief History of Deposit Insurance in the United States, Washington, DC, September. 40 Greenspan, Alan, 2000, Question and answer session, The Changing Financial Industry Structure and Regulation: Bridging States, Countries, and Industries, Proceedings of the Conference on Bank Structure and Competition, Chicago: Federal Reserve Bank of Chicago, pp. 914. Gupta, Atul, and Lalatendu Misra, 1999, Failure and failure resolution in the U.S. thrift and banking institutes, Financial Management, Winter, pp. 87105. Hall, Maximilian J. B., 2001, How good are EU deposit insurance schemes in a bubble environment?, in Asset Price Bubbles: Implications for Monetary and Regulatory Policies, George G. Kaufman (ed.), New York: JAI/Elsevier Press, pp. 145193. 2Q/2002, Economic Perspectives Kane, Edward J., 1992, How incentive-incompatible deposit insurance plans fail, in Research in Financial Services, Vol. 4, George G. Kaufman (ed.), Greenwich, CT: JAI Press, pp. 5192. Kelly, Morgan, and Cormac O Grada, 2000, Market contagion: Evidence from the panics of 1854 and 1857, American Economic Review, December, pp. 11101124. , 1990, Principal agent problems in S&L salvage, Journal of Finance, July, pp. 755764. Kennedy, Susan E., 1973, The Banking Crisis of 1933, Lexington, KY: University of Kentucky Press. , 1989, The S&L Insurance Mess: How Did It Happen?, Washington DC: Urban Institute Press. Mason, Joseph, Ali Anari, and James Kolari, 2000, The speed of bank liquidation and the propagation of the U.S. Great Depression, The Changing Financial Industry Structure and Regulation, Proceedings of a Conference on Bank Structure and Competition, Chicago: Federal Reserve Bank of Chicago, pp. 320345. Kane, Edward J., and Min-Teh Yu, 1995, Measuring the true profile of taxpayer losses in the S&L insurance mess, Journal of Banking and Finance, November, pp. 14591478. Kaufman, George, G., 2002a, Reducing depositor illiquidity at failed banks, Loyola University Chicago, working paper, February. , 2002b, Too big to fail in banking: What remains, Quarterly Review of Economics and Finance, forthcoming. , 1997a, Preventing banking crises in the future: Lessons from past mistakes, Independent Review, Summer, pp. 5577. Peria, Maria Soledad Martinez, and Sergio L. Schmukler, 2001, Do depositors punish banks for bad behavior? Market discipline, deposit insurance and banking crises, Journal of Finance, June, pp. 10291051. Pulkkinen, Thomas E., and Eric Rosengren, 1993, Lessons from the Rhode Island banking crisis, New England Economic Review, May/June, pp. 312. , 1997b, The new depositor preference act, Managerial Finance, Vol. 23, No. 11, pp. 5663. Todd, Walker F., 1994, Lessons from the collapse of three state-chartered private deposit insurance funds, Economic Commentary, Federal Reserve Bank of Cleveland, May 1. , 1996, Bank failures, systemic risk, and bank regulation, Cato Journal, Spring/Summer, pp. 1745. Viotti, Staffan, 2000, Dealing with banking crises Proposal for a new regulatory framework, Sveriges Riksbank Economic Review, No. 3, pp. 4663. , 1995, The U.S. banking debacle of the 1980s, The Financier, May, pp. 926. Willis, H. Parker, and John M. Chapman, 1934, The Banking Situation, New York: Columbia University Press. , 1990, Are some banks too large to fail? Myth and reality, Contemporary Policy Issues, October, pp. 114. Federal Reserve Bank of Chicago 41 Following the yellow brick road: How the United States adopted the gold standard François R. Velde Introduction and summary In 1900 L. Frank Baum published a childrens tale, The Wonderful Wizard of Oz. In it, a little girl from the Midwest plains is transported by a tornado to the Land of Oz and accidentally kills the Wicked Witch of the East, setting the Munchkins free. Yearning to return home, she takes the witchs silver shoes1 and follows the Yellow Brick Road to the Emerald City, in search of the Wizard who will help her. She and the companions she meets on her way ultimately discover that the wizard is a sham, and that the silver shoes alone could have returned her to Aunt Em. Littlefield (1964) and Rockoff (1990) have decoded Baums tale as an allegory on the monetary politics of late nineteenth century America. The silver shoes are the silver standard, the witch of the East represents the monied interest of the East Coast, the scarecrow and the tin man are the farmers and workers of the Midwest, while the cowardly lion is their unsuccessful champion, William Jennings Bryan. The yellow brick road is the gold standard, whose fallacy is exposed by Dorothys triumphant return home borne by the silver shoes. William Jennings Bryan, as nominee of the Democratic Party in the presidential election of 1896, campaigned on a platform to reverse the so-called crime of 1873. The phrase referred to the change in the United States monetary system from bimetallism, in which gold and silver are used concurrently, to the gold standard. Bryan lost, and in 1900 a law was passed firmly committing the United States to the gold standard. The bimetallic controversy soon died away. The United States had taken the yellow brick road. In this article, I recount the historical background to the bimetallic controversy, replacing it in its international context. Bimetallism, which until 1873 had been the system in a number of other countries, disappeared abruptly. I use a model to understand how bimetallism could have been viable in the first place, 42 why it disappeared so suddenly, and whether the United States could have taken another road. Definitions I begin with some definitions. A commodity money system is a monetary system in which a commodity (usually a metal) is also money; that is, the objects that serve as medium of exchange are made of that commodity. The essential feature of such a system is that the commodity be easily turned into money and back. This requires: 1) unrestricted minting, in the sense that the public mint always be ready to convert any desired amount of metal into coin; and 2) unrestricted melting and exporting, allowing money to be converted into the commodity, or into other goods at world prices. A commodity money system based on gold or silver is also described as a gold or silver standard. In such a standard, the medium of exchange may not be limited strictly to coins, but may include notes (privately or publicly issued), as long as the notes are convertible on demand and at sight into coin. The double standard is one where both gold and silver are money. This is also called a bimetallic standard, or bimetallism. The characteristics of bimetallism include: 1. Concurrent use of gold and silver as money, 2. Free minting and melting of both metals, and 3. A constant exchange rate between gold and silver coins. Condition 1 usually means that both gold and silver coins are unlimited legal tender. Any limitation François R. Velde is a senior economist in the Economic Research Department at the Federal Reserve Bank of Chicago. 2Q/2002, Economic Perspectives on the size of debts that can be paid in coins of either metal is therefore a departure from condition 1. All three characteristics should be present to have a proper bimetallic system. For example, conditions 1 and 2 alone define a regime where one metal is the standard (say, silver) and the price of the gold coin is not fixed, but varies according to the market. The fluctuating coin is called trade money. Conditions 1 and 3 alone, but with only one metal freely minted, result in a limping standard. It is similar to a single standard based on the metal freely minted, except that some portion of the money stock is made up of the other metal. The government regulates the size of that portion. A number of countries moved from a bimetallic to a limping standard, as we shall see later. Bimetallism used coins of silver and coins of gold with a fixed exchange rate between the two. For example, the gold eagle and the silver dollar were expected to circulate at a rate of 10:1 (10 silver dollars for one gold eagle), and the values $10 and $1 were inscribed on the coins themselves. Moreover, anyone could take any amount of silver to the mint in exchange for $1 coins and any amount of gold in exchange for $10 coins.2 Let X be the amount of gold in a gold eagle, and Y the amount of silver in a silver dollar. If both coins circulate, then, as money, X ounces of gold are worth 10Y ounces of silver. The ratio (10Y/X) is called the (goldsilver) legal ratio. In the United States, after 1834, an eagle contained 232 grains3 of pure gold, while a dollar contained 371.25 grains of pure silver, so the legal ratio was 16. The relative price of gold and silver as metals in the market is called the market ratio. The history We are so used to our current system of fiat money, where the only thing that matters about a coin or a note is the number inscribed on it, that it takes a slight effort to think of money in earlier times.4 In the Middle Ages, various objects were used as money: round disks made of gold, silver, and copper (or silver alloyed with copper). These disks had designs on them by which one could determine where they came from and how much metal they contained. But they did not have any numbers or other quantitative indication of value. And, in fact, the exchange rates between these objects were not necessarily constant. One should think of such a system as one in which various goods are simultaneously used as means of exchange, because some are more suited to certain transactions than others. For a long time, governments endeavored to stabilize the relation between the various monetary objects, Federal Reserve Bank of Chicago with limited success. The way this was done was by assigning a legal tender or face value to each coin. The government assigns a number Ni to coin i, such that coin i is legal tender for a debt of Ni units of account. Often, Nj = 1 for some particular coin j, and that coin was by definition the unit of account. For a long time, governments had difficulties enforcing these laws, and market rates between two coins often diverged far from the ratio of their legal tender values. Nevertheless, governments kept trying, and by the eighteenth century it was commonly seen as a desirable goal to achieve stability in the relative price between the objects that served as money, so that it might not matter which ones were used in payment of an obligation. By 1800, enough stability had been achieved that denominations could be inscribed on the coins with increasing frequency. But the stability had not necessarily been extended to the whole range of coins, including silver and gold. In practice, a variety of monetary systems existed in Europe by the middle of the nineteenth century: ■ Gold in Great Britain, Portugal, and some colonies; ■ Silver in Central and Eastern Europe and the East (India, China), with gold as trade money in some countries (Netherlands, Germany); and ■ Bimetallism in France, Latin America (with a 15.5 legal ratio), and U.S. (with a 16 legal ratio). The controversy Bimetallism became controversial in the midnineteenth century and remained so until around 1900. I first describe the nature of the controversy and then sketch its history. The controversy around bimetallism ultimately stems from the fact that it is a system that appears to defy economic logic. One of the textbook functions of money is to provide a unit of account and a standard of deferred payment. Accounts are kept in dollars, and debt contracts promise payment of a known quantity of dollars. Thus, money serves as numeraire. The gold standard is a straightforward example, because a dollar is simply defined to be X ounces of gold. In reality, then, it is gold that is used as a numeraire. This poses no particular problem in economic theory. In equilibrium, prices are determined as a vector, or list of numbers, that sets the sum of excess demands for each good to zero (or clears markets). Since nothing changes when all units are consistently changed by a given number, the price vector is indeterminate up to a constant rescaling. Any good (gold, say) can be chosen to be the unit of measurement of value, by setting the price of X ounces conventionally to one. All 43 prices in the economy are thus expressed in ounces of gold or gold dollars.5 But bimetallism is something else: It defines the dollar to be X ounces of gold or Y ounces of silver. As money, the two metals have a fixed relative price, the legal ratio (16 in the United States), whatever the market prices of gold and silver might be. This appears to defy basic economic theory, because it amounts to choosing two goods as numeraire, but prices are indeterminate only up to a single rescaling. In other words, it amounts to fixing by government fiat the relative price between two commodities. Thus, the very existence of bimetallism was at the heart of the controversy. Some argued that, as a monetary system, it was an impossibility and could never be implemented or maintained over any period of time. Rather, they argued, bimetallism would necessarily revert to a single standard, gold or silver, depending on which metal was cheaper on the market. Consider a country with a legal ratio of 16, like the United States. Suppose the market ratio was p, in grains of silver per grain of gold. A legal obligation of $100 could be extinguished by tendering $100 in gold (2,320 grains of gold) or $100 in silver (37,125 grains of silver). Suppose the debtor had $100 in gold in hand, would he tender it? The alternative would be to melt the gold, sell it on the market in exchange for 2,320 p grains of silver, and have the mint turn the silver into $2,320 p/371.25 = $100 (p/16), and tender $100 in silver; the net profit being 100(p 16)/16. If p is greater than 16, it would be better to use silver than gold. In other words, whenever the market price is above the legal ratio, bimetallism would be de facto a silver standard. Should it fall below the legal ratio, the country would suddenly switch to the gold standard. In either case, the cheaper metal (compared with the legal ratio) would replace the other, a mechanism described as Greshams Law in action. Only when the market price happens to coincide exactly with the legal ratio would both gold and silver be used concurrently. We could take into account minting and melting costs: This would determine a narrow band around the legal ratio, within which the market price would be compatible with bimetallism. But as soon as the market price wanders out of the band, bimetallism would collapse to a single standard. At best, according to this argument, bimetallism works occasionally, so that any virtues ascribed to it would be operational only a small part of the time. The rest of the time, the costs of alternating between one standard and the other (minting and melting costs incurred by society as a whole) make bimetallism wasteful and inefficient. It would be better to settle on a single standard on a permanent basis. 44 The bimetallic camp argued that the system, far from degenerating into an alternation between standards, could successfully maintain gold and silver in concurrent circulation at the legal exchange rate. How was this possible? A model Velde and Weber (2000) present a simple model that formalizes the intuition underlying the bimetallists arguments. Clearly, the key to the argument is that the existence of bimetallism somehow influences the market price. If the market price is completely independent of the monetary system, and is left free to vary far from the legal ratio, then the reasoning we have sketched above applies, and, depending on the market prices relation to the legal ratio, gold or silver either disappears or circulates at a premium. Either way, bimetallism cannot survive. To give bimetallism a chance, then, we must allow for the market price to be determined within the model, as well as being exposed to demand or supply shocks. This requires specifying explicit supply and demand for gold and silver aside from their monetary uses. One way to do so is to make consumers care about the total stock of gold and silver in nonmonetary uses, which well call jewelry. Lets begin with the case of a single metal used as money, say, gold. The price of gold relative to other goods is a function of the total stock of gold jewelry. When gold coins are melted, this stock increases, and the value of gold falls. When new coins are minted, the stock of jewelry decreases and the value of gold goes up. A certain amount of gold has to be in the form of coins, that is, cash balances, in order to provide liquidity services and serve as medium of exchange. How is the appropriate stock of coined gold versus uncoined gold determined? Let m be the stock of gold coins (in ounces) and let p be the price level, in ounces of gold per consumption good. The total real value of the cash balances, m/p, depends only on the volume of transactions Y, not on the particular metal used as medium of exchange; in the classic quantity theory equation (setting velocity to 1 for simplicity), m/p = Y. Imagine now that all existing gold is in nonmonetary use: m/p = 0, which is not enough. At the other extreme, imagine that all the gold is in the form of coins, so that none is left for nonmonetary usesthen the price of gold would be very high, and the price level (the inverse of the price of gold) would be very low; so m/p would be very high, perhaps infinite. In between these two extremes, there is some value of m that will make m/p = Y. The key is that m and p are affected at the same 2Q/2002, Economic Perspectives time by a single variable, the split between money and jewelry; and the equation m/p = Y only has one unknown, which is that split. With a single standard, then, the price level and the money stock are determined, given the volume of transactions. What happens with two metals? Things become more complicated. On one hand, we have gold and silver jewelry, and the relative value of gold and silver are each decreasing functions of the stocks of jewelry. On the other hand, cash balances can take the form of either gold coins (m1) or silver coins (m2), with silver coins valued at a certain ratio in terms of gold coins (e). That ratio must itself be equal to the ratio of relative prices of gold and silver, as explained earlier. Prices can be expressed in ounces of gold per good, noted p as before, or in ounces of silver per good, p/e. We have an equation of the form (m1 + e m2)/p = Y, but we now have two variables affecting the equation: the split between gold coin and gold jewelry and the split between silver coin and silver jewelry. Two unknowns in one equation mean that there are many possible solutions. (Box 1 presents the model in more detail.) In other words, many different goldsilver ratios are possible. Start from a given ratio, with corresponding quantities of gold and silver jewelry. If one wanted a higher goldsilver ratio, with gold more valuable relative to silver, one could reduce the stock of gold jewelry and drive up the price of gold; then gold coins would be minted, and silver coins would have to be melted to make room for the gold coins, driving down the price of silver. One could do so until the relative price of gold to silver was pushed up to the new ratio. This suggests that there is a whole range of possible goldsilver ratios, with corresponding quantities of coined silver and gold: the higher the ratio, the more silver there is in the money stock. It also suggests that there is ample room for a government, or a large enough group of governments at any rate, to settle on a particular ratio between gold and silver, and that there are no fundamental forces that would push away from that arbitrarily chosen ratio. The relative price of gold and silver is indeterminate, within the range. The model says more than this. Suppose there is a large disturbance to the supply of gold, say, a large increase in gold supplies. How is the monetary equilibrium modified? Part of the new supply of gold can be turned into jewelry, which would tend to cheapen gold and move us away from the existing ratio. But part of the new supply can also be minted; as a result, some silver coins would have to be melted down to make room for the new gold coins. The melted silver would increase the stock of silver jewelry, and cheapen silver. If minting of gold takes place at the right Federal Reserve Bank of Chicago pace, the melting of silver can exactly compensate for the increase in gold jewelry so as to keep the ratio e exactly constant. Of course, there are limits to this process. In particular, the goldsilver ratio can be stabilized around an arbitrary value only so long as there are stocks of gold and silver coins to act as buffers against shocks to gold and silver supplies. Suppose a particularly large discovery of gold takes place. Part of it will have to be minted, and that may completely displace silver from the monetary circulation. If it does, no more silver circulates as coin, and no further increases in silver jewelry can offset the cheapening of gold. Bimetallism turns into a gold standard, and the goldsilver ratio falls. To restore bimetallism requires changing the ratio to a new value more compatible with the existing gold stocks (in our example, reducing the ratio). Thus, for any given worldwide stocks of gold and silver, there exists an upper bound, as well as a lower bound, for values of the ratio compatible with effective bimetallism. Given the stocks, a relatively high ratio requires putting more gold into coins to drive up the relative price of gold jewelry and putting less silver into coins. Too high a ratio cannot be sustained because it would require taking all silver out of coinage, making the system effectively a gold standard. Similarly, too low a ratio leads to a silver standard. This band of possible ratios moves around with changes in world stocks. For example, if the stock of silver increases, it makes it possible to sustain higher ratios. As the relative quantities of gold and silver change over time, so do the bands that constrain the feasible ratios, and we would expect to see the ratio of prices broadly follow the ratio of stocks over long periods. The history (continued) Figure 1 suggests that this was so. The figure plots, in ratio, an estimate of gold and silver stocks since the discovery of the New World. It dips at first, showing that relatively more gold than silver flowed in from the New World. Then, from about 1530, it rose steadily as vast quantities of silver began to come out of the mines in Peru. The ratio of stocks stabilizes in the late seventeenth century, as flows of Brazilian gold increase. We can see that the market ratio followed these movements, as European countries sought to maintain concurrent use of both coins. After 1820, the market ratio is remarkably stable, up to 1873. By contrast, something happens to the ratio of stocks around 1850. Bimetallism became controversial around 1850. The date is not a coincidence. In 1849, it was discovered that the Sierra Nevada Mountains of California 45 BOX 1 The model Time is infinite and discrete. There are three types of goods in the model: a nonstorable general consumption good c, and nondepreciating stocks of gold and silver metal, Q1 and Q2 (in ounces). I treat gold and silver symmetrically. In each period, there is a given amount of consumption good and given increases (or decreases) in the stocks of gold and silver. Total quantities of all goods are thus exogenous. The quantities that are determined within the model are the share of gold and silver stocks in monetary and nonmonetary uses. Gold and silver can each be in either of two forms: coined or uncoined. For simplicity, all gold coins are of the same size and weigh b1 ounces each; likewise with silver coins, each weighing b2 ounces. Let mi (i = 1,2) be the number of existing coins and di the quantity of either metal in uncoined form. We then have an adding-up condition: Qi = bi mi + di bi oz per coin i. I assume that it is costless to convert metal from one form to the other. Converting from coined to uncoined is melting, and converting from uncoined to coined is minting. A key feature of a commodity money standard is that both operations be unimpeded. A representative households preferences are defined over the consumption good and over the stocks of uncoined metal. That is, the household derives direct utility from the uncoined metal only. Let the total utility derived each period be u(c) + v(d), where d stands for (d1,d2). The household discounts future consumption by a factor b < 1. Metal is coined because money is needed for purchases of the consumption good; in other words, there is a cash-in-advance constraint. Both coins are perfect substitutes in the constraint at an endogenous ratio or exchange rate e (in gold coins per silver coin). If p is the price of the consumption good denominated in gold coins, then the constraint is: 1) pc = m1 + em2. The household maximizes utility subject to the cash-in-advance constraint and a budget constraint. The first-order conditions for the households problem include two equations that determine the optimal holding of uncoined metal. Consider the marginal gold coin held by the household. One could spend the coin and consume 1/p more units of consumption good today, bringing a marginal utility u¢(c)/p. The alternative is to melt the coin and hold b1 more ounces of uncoined metal, which would bring a marginal utility of b1v1(d) today (where v1(d) is the derivative of v with respect to its first argument d1); and then, in the next period, convert the metal back to coin and consume 1/p more, bringing a marginal 46 utility b u¢(c)/p (discounted because it takes place in the future). For a silver coin, the tradeoff is the same, except that a silver coin buys e/p units of good. At the optimum, the alternatives should bring the same utility, so that: 2) u a (c ) b1v1 (d ) p 3) e u a (c ) p u a (c ) b2v2 (d ) p e u a (c ) p In equilibrium, the metal stocks that the household chooses to hold, coined and uncoined, must add up to the existing supply: 4) b1m1 d1 Q1 , 5) b2 m2 d 2 Q2 . Equations 1, 2, 3, 4, and 5 are all the equilibrium conditions. The unknowns are e, m1, m2, p, d1, and d2. This leaves one more unknown than we have equations, so we are free to choose e. Formally, there exists a range [ e , e ] of possible ratios, with a different distribution of uncoined metals (d1,d2) for each ratio. At the upper end, there is almost no silver in monetary use, and the world is on the edge of the gold standard. At the lower end, there is no gold coin, and the world is almost on a silver standard. Note that, in any equilibrium, equations 2 and 3 imply that b2 v1 ( d ) , eb1 v2 ( d ) in other words the legal ratio always equals the market ratio. One can reduce the equilibrium conditions to a single equation in the two unknowns d1 and d2: 6) u a( x) x 1 1 C [v1 ( d )(Q1 d1 ) v2 ( d )(Q2 d 2 )]. The right-hand side of equation 6 is the real value of money balances, at market prices. This value is the same in all bimetallic equilibria: No matter what the goldsilver ratio is, the same resources are devoted to monetary transactions. If one changes the legal ratio, say, by increasing it, then silver shifts to nonmonetary uses, driving down the marginal utility of uncoined silver (and hence the relative price of silver). At the same time, gold flows into monetary uses to make up for the lost silver, which drives up the price of gold and maintains real balances constant. This brings the market ratio in line with the legal ratio. 2Q/2002, Economic Perspectives FIGURE 1 Ratio of gold and silver stocks and market ratio ratio of cumulative output, silver/gold market ratio 70 40 60 30 50 40 silver stock/gold stock (left) 20 30 silver/gold market ratio (right) 20 10 1490 1530 ’70 1610 ’50 ’90 1730 ’70 1810 ’50 ’90 10 1930 Source: Velde and Weber (2000). were full of gold, hitherto untouched. Figure 2 shows how large this discovery was, relative to existing stocks, and how the ensuing flow of new gold remained large into the early twentieth century. Returning to figure 1, the market ratio ceases to be stable around 1873. In fact, the value of silver compared with gold collapses and reaches unprecedented levels by 1900. At the same time, major changes take place in the worlds monetary system in rapid succession. In December 1871, newly unified Germany announced that it would switch from the silver standard, predominant in the preexisting German states, to the gold standard. The Scandinavian countries followed in December 1872, as did the Netherlands a few months later. The year 1873 saw the collapse of bimetallism. Germany began implementing its move by retiring existing silver coins, selling them on the world market, and buying gold to coin in replacement. In February, the U.S. suspended the free coinage of silver (see the next section). By the end of the year, the European countries that collectively adhered to bimetallism within the framework of the Latin Monetary Union of 1865 (namely, France, Switzerland, Belgium, Italy, and Greece) had all restricted free minting of silver, and in 1878 they agreed to suspend it indefinitely. The price of silver fell. In 1892, Austria, traditionally a silver Federal Reserve Bank of Chicago country but under an inconvertible paper currency, resumed convertibility; but, as the U.S. did after the greenback, Austria made its currency redeemable in gold, and only gold was freely minted. Russia did the same in 1897. In 1893, India suspended free minting of silver, and adopted a variant of the gold standard in 1899. Latin American countries, traditionally silverbased, increasingly switched to the gold standard. In the Far East, Dutch, English, and French colonies followed suit, as did the Philippines under U.S. control. By 1913, China was the sole major country with free minting of silver. What explains the collapse of a system that had been working for decades? The very large shock to gold supplies in 1850 that is apparent in figure 2 is a clear suspect. The model tells us that a discovery of gold will lead to increased coinage of gold and displacement of silver, leading possibly to the complete replacement of silver. How large of a change in the supply of either metal can be accommodated by a bimetallic system will therefore depend on the shares of the metals in the monetary stock. If very little silver is coined to begin with, it would not take a large increase in gold supply to drive bimetallism to a gold standard. The stability of the market ratio around 15.5, the legal ratio in the European bimetallic countries, suggests that the 47 mechanics of bimetallism were operating as the model predicts, at least initially. Further evidence comes from estimates of the share of gold in the French money stock, shown in figure 3. France, by its size and political importance, was the pivotal bimetallic country in Europe. Figure 3 shows that the share of gold in the French money stock mirrors the movements of the ratio of metals in figure 1. It rises sharply from 1850, then stabilizes in 1865, when silver discoveries in Nevada lead to increased production and coinage of silver, and starts falling slowly thereafter. The model allows us to consider quantitatively whether bimetallism was nearing its breaking point, whether it could have survived longer, and whether the action of Germany alone could have precipitated its downfall. I use estimates of nonmonetary stocks of gold and silver and data on the market ratio between 1873 and 1913 to estimate a model of the demand for gold and silver. I then use this model to predict what the bounds on the ratio were. I do this under three counterfactual assumptions: One is that the monetary system of the world (who was on the gold, silver, or bimetallic standard) remained as it was up to 1871I call this the 1871 system. The second is that Germany alone switches from the silver to the gold blocI call this the 1872 system. Third, I suppose that Germany, Norway, Sweden, the United States, and the Netherlands also switch to goldI call this the 1873 system. Details of the model are in the appendix. The model suggests three points. One is that, in the early 1870s, the world was indeed close to replacing all silver with gold and ending in a gold standard, but that the relative abundance of silver in the 1880s and 1890s would have removed that threat. The second is that Germanys switch to the gold standard actually relieved the immediate pressure on bimetallism: By increasing the monetary demand for gold, Germany was helping to absorb the vast quantities of gold that were threatening the bimetallic standard. The third point is that Germany, by decreasing the monetary demand for silver, was also raising the lower bound on the bimetallic ratio (the lower line in figure 4), since it gave the remaining silver and bimetallic countries a larger mass of silver to absorb into monetary and nonmonetary uses. Figure 4 shows even the move to gold by Norway, Sweden, the Netherlands, and the U.S. was not enough to turn bimetallism into a silver standard, at least immediately. These conclusions make the sudden collapse of bimetallism in 1873 something of a mystery. If bimetallism could continue, and if Germanys choice FIGURE 2 Annual world production as percentage of existing stocks, 180028 annual production/cumulative output (percent) 4 3 gold 2 1 silver 0 1800 ’10 ’20 ’30 ’40 ’50 ’60 ’70 ’80 ’90 1900 ’10 ’20 ’30 Source: Velde and Weber (2000). 48 2Q/2002, Economic Perspectives FIGURE 3 Gold share of total French coin stock percent 100 80 60 40 20 0 1840 ’50 ’60 ’70 ’80 ’90 1900 ’10 Sources: Flandreau (1995) and Sicsic (1989). of monetary regime actually made it easier to do so, why the sudden rush to abandon bimetallism? Bimetallism could have survived long after 1873; it only took enough countries to remain committed to silver, either alone or in a double standard. Conversely, once silver was abandoned by enough countries, its price fell and anyone who stayed on that standard endured a depreciating currency and inflation. The currency depreciates, moreover, not only because its exchange rate falls, but also because the value of the countrys money stock, as metal, is falling: The coins are literally losing their value. The politics of the Latin Monetary Union after 1873 illustrates the problem (Willis, 1901). Founded under the aegis of France in 1865, the union consisted of setting a common bimetallic standard for all member countries and making all coins legal tender throughout the union. As long as the market value of a coin was very close to its face value, be it gold or silver, this was a relatively innocuous provision. With the collapse in the price of silver, free minting of silver was suspended by the member states in 1873. The silver coins remained legal tender everywhere but were now a token coinage. Did the issuing state bear any responsibility to redeem silver coin in gold at its face value? The question was posed when the treaty came up for Federal Reserve Bank of Chicago renewal in 1878, and countries found that a sizable amount of their silver coinage was circulating in other states. Much as some states wished to leave the union, they could not afford to redeem the coins, and were forced to remain. They eventually developed a framework for the redemption of the coins, and the union continued with a limping standard until after World War I. This suggests an explanation for the events of 1873. Once the commitment to bimetallism of a few countries wavered, there was a rush for the door, so to speak. The last one to abandon silver would be left holding the bag, namely, a lot of depreciated silver coins. Germany moved first, and for a few years was able to sell its silver stock at 15.5:1 for gold. When the price of gold started rising, it halted its silver sales, and resigned itself to a limping standard. Other countries like France were able to suspend free minting of silver while their holdings of silver were still relatively low. Indeed, figure 3 shows that France was in fact simply letting itself go to a gold standard, exchanging its silver at 15.5:1, when the growth in silver output of the 1860s, followed by Germanys decision, reversed the trend and made it acquire silver. Should bimetallism ever end, it would be left holding the bag. Faced with that possibility, it may have seemed better to abandon bimetallism. 49 The collapse of 1873 reflects a deep feature of my model of bimetallism. Recall that the model displays a multiplicity of equilibria, represented by the range of possible goldsilver ratios; at the extremities of that range are the gold standard and the silver standard. This multiplicity is a familiar result for fiat currencies in models that only generate demand for one type of currency; with two currencies, there is nothing to pin down the real value of balances held in either form, as long as the rates of return are the same on both. In a commodity money system, a similar effect takes place, except that quantities of gold and silver jewelry have to adjust in order to maintain equal rates of return on both currencies (that is, maintain a fixed price ratio). What is properly an indeterminacy in a fiat money world (nothing determines nominal prices, and real prices and quantities are identical in all equilibria) is a multiplicity in the bimetallic world (some quantities are different across equilibria, but some nominal values are indeterminate). The collapse of 1873 may be seen as a sudden shift from one equilibrium (bimetallism at a 15.5 ratio) to another equilibrium (a gold standard equilibrium). What prompts the sudden shift is the fact that, while monetary functions are carried out just as well by a mixture of gold and silver at a 15.5 ratio, or by gold alone, the relative price of gold and silver can be very different in the two cases. In other worlds, holders of silver are not indifferent at all about which equilibrium prevails. In the 1860s, France was on the verge of ridding itself of all silver, and then saw that it was acquiring silver again: This made it a potential loser should bimetallism end. Rather than run the risk, France abandoned bimetallism, thus precipitating the event it feared. We will see that the interests of holders of silver were also at play in the American segment of our story. The “crime of 1873” In the United States, the end of bimetallism became known, by those who regretted it, as the crime of 1873. Let us briefly review the historical background. The United States had been officially on the bimetallic standard from 1792; coins of $1 and less were made of silver, coins of $5 and more of gold. Initially, the ratio was set at 15:1. In practice, very little was minted in either metal, mostly old Spanish silver continued to circulate (the dollar was in fact the colonial name of the Spanish piece of 8 reals, minted in abundance in Mexico with silver from Peru). In 1834, the ratio was changed to 16:1 by debasing the gold coin. In the late 1840s and early 1850s, the California discoveries resulted in a great amount of gold coins FIGURE 4 Limits on goldsilver ratio implied by model, three counterfactuals silver/gold ratio (oz silver/oz gold) 30 gold standard (1873 system) 25 gold standard (1872 system) 20 gold standard (1871 system) 15.5 ratio 15 10 silver standard (1873 system) silver standard (1872 system) 5 silver standard (1871 system) 0 1873 50 ’78 ’83 ’88 ’93 ’98 1903 ’08 ’13 2Q/2002, Economic Perspectives being minteda new mint had to be set up in San Francisco to handle the flow. At the same time, as the model predicts, silver coinage was melted down. The loss of small coins became particularly acute, prompting Congress to take a first step toward a gold standard in 1853. Until then, fractions of the dollar, ranging in value from 5 cents to 50 cents, contained exactly the right amount of silver in proportion to their face value: A 5 cent coin contained 1/20 as much silver as the dollar, etc. After 1853, the fractions of the dollar contained only 93 percent of the silver that they used to. Moreover, their capacity as legal tender, which had been unlimited, became limited to debts of $5 or less. Finally, the coins were not issued freely in exchange for silver brought to the mint. Instead, the quantities minted were regulated by the Secretary of the Treasury, and the coins were to be sold to the public in exchange for gold coins. This made the smaller denominations partly token: Their face value was 7 percent higher than justified by their content, and they were made on demand by the government. For these coins, the legal ratio (the ratio of the silver contained in $10 of dimes, divided by the gold contained in a gold eagle) was 14.88 instead of 16. For those coins at least, the threat of being melted down was held at bay. But the U.S. remained on a bimetallic standard, because the silver dollar was still minted on demand in unlimited quantities and was unlimited tender. With the Civil War, the U.S. ceased to be on a bimetallic system. Instead, during the greenback era from 1862 to 1879, the government issued an inconvertible paper currency called the greenback. It was legal tender just like coins. Instead of seeing bad money displace good money, gold coins continued to circulate, but at a premium over greenbacks, a premium that varied with the fortunes of war and reached 150 percent in 1864. Once the war ended, the premium fell back under 50 percent and slowly declined over time, as the government kept the quantity of greenbacks under tight control. After some debate, the decision was taken in 1873 to resume convertibility, scheduled for January 1, 1879. Meanwhile, a law was passed in February 1873 to revise and amend minting laws. Of course, no minting had taken place during the years of the greenback era, since the mint would have paid any incoming gold or silver in greenbacks at face value. The act prescribes the minting of gold coins and subsidiary silver coins as before 1862, but does not mention the silver dollar at all. The silver dollar would not be coined on demand anymore.6 Federal Reserve Bank of Chicago This was the crime of 1873. Not much notice was taken at the time, but it became much more controversial later, during the deflation of 187996. The deflation had two sources. One was the fact that the U.S., having expanded its money supply in the form of greenbacks during the Civil War to finance its expenditures, now had to reduce it (or at any rate let it grow more slowly) in order to bring the value of greenbacks up to par. Resumption of convertibility, in fact, required that a dollar in greenback be worth the same as a dollar in gold. After resumption, however, deflation continued for another 15 years. The second source of deflation, one that affected all countries on the gold standard, was the fact that these economies demand for gold, driven in part by income growth,7 grew faster than the supplies of gold; and the fact that, because of the collapse of bimetallism, the number of countries on the gold standard increased as well. One interest group suffered from the end of bimetallism, namely the silver producers of the western states. But the silver party drew wider support. The plank of a return to bimetallism at 16:1 was seen by many as a remedy to the deflation, which was hurting debtors, particularly farmers in the Midwest. A greenback party had formed to oppose the return to convertibility and the deflation that it required; that party disappeared after 1880, but the agitation then turned to silver. The strength of the political forces aligned in favor of silver was never quite sufficient to reverse the crime of 1873. In practice, free minting of silver never returned. But the silver dollar regained full legal tender status in 1878, and from 1878 to 1893, the government was compelled by Congress to purchase quantities of silver and turn them into money. This, as well as the numerous nearly successful attempts at restoring free coinage of silver, was enough for some to question the United States commitment to the gold standard for 30 years. The monetization of silver took place under two distinct regimes. In the first regime, from 1878 to 1890, the BlandAllison Act of 1878 required the U.S. Treasury to purchase between $2 million and $4 million in silver every month, at market value, and mint it into dollars (actual purchases were between $2 million and $3 million per month). By the end of 1889, there was $438 million in gold and $311 million in silver in circulation in the U.S. As a result, the United States was on a limping standard. Both metals were legal tender, but only one metal was freely minted. Coins of the other metal were becoming token: While the face value of silver dollars remained $1, the value of their intrinsic content, which was close to $1 when the market ratio 51 was close to 16, fell as the market ratio fell, to 80 cents by 1890. The second regime of silver purchases began with the Sherman Silver Purchases Act of July 1890, which followed the shift in the balance of forces in Congress after five western states were admitted to the Union in 1889 and 1890. On the surface, the act seemed to go further toward monetizing silver, since it increased the required monthly purchases to 4.5 million ounces at market prices (about $4.5 million at the time). This represented the whole silver production of the United States and about 40 percent of world silver production. However, Treasury policy actually mitigated the effect of the act in the following way. The amount was specified in ounces and, as the market price of silver fell, so did the amount spent. The purchased silver, rather than being minted into dollars, was to be held by the Treasury as bullion. In payment of the bullion, the Treasury issued notes which were fully legal tender and redeemable on demand into gold or silver at the Treasurys discretion. Had the Treasury systematically redeemed them in silver, the effect would have been the same as simply minting the purchased silver. The Treasury in fact pursued a policy of redemption in gold. In effect, the government was mandated to buy a given amount of some commodity, and issued (gold-backed) notes in payment. The seeds of further trouble, the disturbed years from 1891 to 1897 (Friedman and Schwartz 1963, p. 104) were contained in the act. The mandated purchases of silver were adding a strain on government finances, increasing expenditures by 25 percent at a time when the McKinley Tariff Act reduced revenues. The result was the disappearance of the federal surplus by 1893. The U.S. federal government finished the fiscal year 1890 with a surplus of $105 million and a gold reserve of $190 million. By June 1894, with $134 million in Treasury notes outstanding, the surplus had turned into a $70 million deficit. The act also left the Treasury holding a growing and increasingly worthless stockpile of silver. In July 1890, when the act was passed, silver was worth $1.06 per ounce. By November 1893, it had fallen to 72 cents. Over that period, the Treasury had bought 169 million ounces of silver, at a cost of $156 million, which, as of November 1893, was worth $121 million. Should the Treasury decide to mint its silver, it could turn each ounce into $1.29 of legal tender, making its stockpile worth $218 million. In effect, the government held a large put option on the private sector. What prevented the Treasury from exercising that option, by coining its silver and repaying the outstanding notes with it? Nothing but its own interpretation 52 of the law that the policy of the United States [is] to maintain the two metals on a parity with each other upon the present legal ratio. In other words, the policy of redeeming notes in gold at par could change overnight. Redeeming notes in silver instead of gold would mean an abandonment of the gold standard and a large devaluation. Should the government run out of gold with which to redeem its notes, it might well be led to redeem them with silver. The very prospect led many to present their notes for redemption in exchange for gold. As a result, the governments gold reserve, which was intended to secure the parity of the legal tender notes (the remaining greenbacks of the Civil War), dwindled from $190 million in June 1890 to $65 million in June 1894. The years 189394 bear interesting similarities with modern currency crises: rising deficits, shrinking reserves, capital flight, and speculation against the currency (Grilli, 1990, and Miller, 1996). President Cleveland took office in March 1893, and his administrations commitment to the gold standard seemed open to question when the Treasury secretary was saying that the Treasury would redeem its notes in silver if it was expedient to do so. In June 1893 India suspended free coinage of silver and the price of silver immediately fell. This prompted a major banking crisis, with hundreds of banks failing, and a sharp recession, with industrial production falling by 27 percent between April and September. The Treasury nevertheless continued to redeem its notes in gold. Faced with a dwindling reserve, it tried to sell bonds for gold. The only bonds it had legal authority to issue were coin bonds, which were redeemable in coin, that is, either gold or silver, and Congress refused to authorize gold bonds, arguing that the Treasury ought to use its large silver stockpile. The Treasury therefore had to pay a risk premium on the bonds it was able to sell, because of the risk that they would be paid at maturity in silver; and when a bond issue was announced, notes were presented for redemption to withdraw gold in order to sell it back to the Treasury. This endless chain was repeated several times. The matter came to a head with the election of 1896, in which Republicans promised to return to bimetallism as soon as a worldwide consensus to do so could be arranged, while Democrats argued for a return to bimetallism at a 16:1 ratio, without waiting for the aid or consent of any other nation. William Jennings Bryan, the Democratic nominee, campaigned for bimetallism with a speech known for its peroration: You shall not press down upon the brow of 2Q/2002, Economic Perspectives labor this crown of thorns, you shall not crucify mankind upon a cross of gold.8 He lost the election to the Republican William McKinley. The year 1896 was the high watermark of bimetallism in the U.S., even if it took a few years to formally seal the countrys commitment to gold, partly because of the silver partys continued clout in the Senate9 and partly because McKinleys first term was taken up with tariffs and the Spanish-American War. In March 1900, however, the Gold Standard Act was passed, unambiguously defining the U.S. dollar as 23.22 grains of fine gold. It also enacted that all forms of money issued or coined by the United States shall be maintained at a parity of value with this standard, and it shall be the duty of the Secretary of the Treasury to maintain such parity and maintained the legal tender status of the silver dollars; moreover, the Treasury notes issued since 1890 could now only be repaid by the Treasury in gold. A gold reserve was created, and the Treasury was authorized to borrow in order to maintain that reserve. In the ensuing years, the root cause of the silver agitation disappeared, as deflation turned to inflation in the wake of large gold discoveries in Australia and Alaska and improvements in methods of extraction. The U.S. would remain firmly on the gold standard until 1934. What if? Friedman (1990a, b) revisits the crime of 1873. In his estimation, the crime of 1873, although not a crime, was a mistake. Had the U.S. restored its bimetallic minting policies in 1873, it would have effectively been on a silver standard and, by his calculations, would have enjoyed a steadier price level than it did. I can use my model to evaluate one assumption underlying Friedmans calculations. He assumed that the rest of the world would have pursued the monetary policies it did, and that the U.S. would necessarily have been on a silver standard. That is, the ratio of 16:1 would have been outside of the bounds I defined earlier. Figure 5 shows that, in my model, this would not have been so, at least initially. Indeed, in the 1870s the U.S. would still have been on a gold standard, and, from 1880 to 1903, it would have been effectively bimetallic. During that period, movements of the price level in the U.S., in the gold-standard countries, and in the silver-standard countries would have been the same. As Velde and Weber (2000) show, bimetallism does stabilize the price level relative to either single standard, as long as the shocks affecting the markets for each metal are not perfectly correlated. Federal Reserve Bank of Chicago However, figure 5 also shows that, ultimately, the U.S. would have been forced onto silver. This is partly due to the growth in the number of gold-based countries and their increasing demand for gold as a medium of exchange, the very causes of the deflation experienced by gold-standard countries in that period. Over the course of the 1880s and 1890s, that demand would have progressively drained the U.S. of its gold coinage. But the other factor driving the bounds in figure 5 upward is the progressive abandonment of silver by other countries, notably Austria, Russia, and India. Those countries might perhaps have stayed with silver had the U.S. remained bimetallic and held out the prospect of continued stabilization of the goldsilver ratio. Indeed, one might speculate that, as far back as 1873, a U.S. commitment to bimetallism might have persuaded France to keep its mints open to silver.10 The need for cooperation on the international financial architecture was well understood at the time. While Bryan and his more extreme followers rejected it, moderate supporters of bimetallism in the U.S. insisted that international cooperation was needed to make a return to bimetallism a realistic proposition. But the difficulties of achieving such cooperation after the events of 1873 is illustrated by an international conference that took place in August 1878 in Paris.11 According to the report of the American delegates, the participants for the most part adhered to the notion that silver had a monetary role to play, a change from the 1865 international monetary conference that had endorsed the gold standard. But the European delegates did not believe there was anything that could be done about the fall in the price of silver, whereas the Americans believed that a policy of action could alter it. The Europeans were not a little suspicious of American intentions and abilities, plausibly reading the support for bimetallism as a disguised push for inflation. Nevertheless, a maintained commitment to bimetallism would have altered politics inside the U.S. and the countrys relations with other countries. Whether U.S. adherence to bimetallism could plausibly have convinced other countries, such as India, to stay on silver and whether this would have prolonged bimetallism up to World War I are questions for future research. Conclusion The very fact that bimetallism was abandoned by all countries that adhered to it in a short space of time has been seen, in and of itself, as an indictment of that monetary system. I show that bimetallism was not an absurdity. Rather, economic theory predicts that such a system would have a multiplicity of possible outcomes, 53 FIGURE 5 Limits on goldsilver ratio, U.S. alone on bimetallism silver/gold ratio (oz silver/oz gold) 55 50 45 40 35 30 25 gold standard 20 15 silver standard 10 5 1876 ’81 ’86 ’91 ’96 1901 ’06 ’11 ’16 Note: Assumes U.S. alone is on bimetallism, with a gold–silver ratio of 16:1. ranging from a low to a high goldsilver ratio, corresponding to a silver standard and a gold standard, respectively, with bimetallic regimes for all the intermediate ratios. The parameters determining the range include the number of countries that are willing to use silver or gold indifferently as money. Thus, bimetallism was a viable monetary arrangement that could be maintained for long periods, if enough countries adhered to it. Moreover, the sudden collapse is understandable as a consequence of the very property that made bimetallism viable: Should the number of countries suddenly change, bimetallism might not be feasible at the existing ratio anymore, prompting a switch to either the gold or silver standard, with potential losses for holders of the other metal. Rather than be the last one 54 left with silver, countries rushed for the door in 1873 and adopted the gold standard. Thus, the decision to abandon bimetallism might seem justified a posteriori, but not necessarily a priori. I show that the United States could have plausibly remained on a bimetallic standard after 1873, in spite of what other countries were doing. But other forces were at workgrowth rates in gold-standard countries and flows of new discoveriesthat could have ultimately forced the United States off bimetallism. It would then have had to choose between the yellow brick road and the white brick road, and the speculative attacks that plagued the U.S. dollar in the 1890s would no doubt have accompanied that difficult decision. 2Q/2002, Economic Perspectives APPENDIX: COMPUTING COUNTERFACTUALS I wish to compute two counterfactuals. The first one assumes that the monetary systems of all countries remain unchanged from 1871 to 1913 and determines whether bimetallism could have continued, or whether the gold standard was bound to occur. To answer this question, I compute the values of the goldsilver ratio at which a gold standard and a silver standard become inevitable for each year. The second counterfactual assumes that the crime of 1873 did not take place, and that the United States had remained on a bimetallic standard at 16:1 which, in practice (given what all other countries did), would have meant a silver standard. Would the price level have been more stable as claimed by Friedman (1990a, b)? My strategy is to use historical data to compute or estimate parameters of the model, and then modify certain parameters as dictated by the counterfactual assumptions. Then I compute the values of the endogenous variables (prices and quantities) by solving the models steady state equations for the new parameters. Data I make use of the following annual data, from 1873 to 1913: 1. the average value in December of the gold silver ratio, 2. the total stock of gold and silver in the world at the end of the year, and 3. the stock of gold and silver coin in each country at the end of the year. Series 1 and 2 are described in Velde and Weber (2000). The same paper uses worldwide stocks of gold and silver coin in 1873, taken from Kitchin (League of Nations, 1930) and Drake (1983). With these series, however, the ratio of gold to silver nonmonetary stocks rises by 15 percent from 1873 to 1890, even as silver depreciates by percent relative to gold. This is difficult to reconcile with the kind of preferences for gold and silver that I wanted to use. Kitchin and Drake both estimated monetary stocks as residuals: They estimated how much gold and silver was produced each year, and how much went into industrial uses, the remainder accruing to money stocks. To get another estimate, I added up directly national money stocks for each year. The Annual Reports of the Director of the Mint provide estimates of these stocks for a growing list of countries in 1873, 1878 to 1883, 1892 to 1907, and 1909 to 1913. For a number of countries, better and continuous series can now be found. Thus, for the United States, the United Kingdom, Germany, France, Italy, Spain, Portugal, the Netherlands, and Japan, I have Federal Reserve Bank of Chicago used the same sources as Rolnick and Weber (1997). Furthermore, for India, I have relied on Atkinson (1909) and Keynes (1913). These countries together accounted for about 50 percent to 55 percent of world output in that period (based on Maddison, 1995). They thus represent a large, but not sufficient fraction of the world. I have relied on the Director of the Mints estimates for the remaining countries. As it turns out, the estimates for 1873 are quite close to those of Kitchin and Drake as used in Velde and Weber (2000), but diverge after that date. Figure A1 plots the market ratio against the ratio of estimated stocks. The slope is negative, which is an improvement. Estimation The specification of preferences over stocks of nonmonetary metal that I use is a constant elasticity of substitution: v(d1,d2) = [(ad1) r + d2r ]1r. My specification obviates the need for a time series of world income. In equilibrium, the market ratio is the ratio of marginal utilities: e aS ( d1 S1 ) . d2 I regress the log of the ratio of worldwide nonmonetary stocks of silver to those of gold on the log of the market ratio and a constant: log(e) B log ( d1 ) C. d2 As figure A1 suggests, there is a somewhat anomalous period from 1893 to 1903. In 1893, India discontinued free minting of silver, and at the same time Austria and Russia committed to a gold standard and the American silver purchases came to an end. The resulting fall in the price of silver was not accompanied by an immediate adjustment in quantities (see the horizontal movement in figure A1). I use the sample from 1873 to 1892 and 1904 to 1913 only. By ordinary least squares (OLS), I find a = 0.23 (standard error: 0.022) and bÿ= 2.36 (standard error: 0.068). I then estimate ¨ ¥ S 1´ · S 1/ B 1 and a exp © C ¦ ¸, § S µ¶ and I find ª ¹ rÿ= 3.34, or an elasticity of substitution between gold and silver of 0.23. It is not very satisfactory to exclude 55 the 11 observations from 1893 to 1903. A modified version of the model with adjustment costs would probably better match the data, at the cost of some complexity. FIGURE A1 Goldsilver market ratio plotted against ratio of estimated world nonmonetary stocks log (nonmonetary gold stock/nonmonetary silver stock) -2.95 Counterfactual 1876 1879 The aim is to determine the -3.00 1874 1877 1878 1880 range of possible goldsilver ratios 1875 1873 1884 1883 1885 for which bimetallism was possible 1900 1886 1882 1895 1901 ( [ e , e ] in the notation of box 1). The -3.05 1889 1887 1881 1897 1896 1888 1899 1902 upper end of the range e corresponds 1892 1898 1894 1891 1890 1893 1903 to the point at which as much silver -3.10 1904 as possible is in nonmonetary uses (driving down its value relative to 1905 gold), and the world uses no silver -3.15 1906 as money. In reality, a significant part of the world was under a silver 1909 standard, in which gold could not -3.20 1907 have replaced silver as medium of 1908 1910 1913 exchange, so the limit on silver in 1912 1911 monetary use is not 0, but rather the -3.25 amount necessary to carry out transactions in the silver countries. Like-3.30 wise, the lower end of the range, e , 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 corresponds to the point at which only log (price of gold/price of silver) gold-standard countries use gold. Notes: Gold–silver market ratio is the December average for each year. Estimated worldwide nonmonetary stocks are in logs. The regression line is computed using Let xs (respectively, xg) be the 1873–92 and 1904–13 only (see text). share of world transactions carried out in silver-standard (respectively, gold-standard) countries. Then, at the upper end of the range of ratios, the stocks of gold and Having the actual money stocks and values for the silver in nonmonetary uses are such that parameters of preferences, I compute a series for x for 18731913. I find the series xs and xg by taking the money (Q2 d2)v2(d) = (1 xg)x, 7) (Q1 d1)v1(d) = xg x, stocks of the countries that were on a silver standard in 1871, as share of the world money stocks. I can solve for d1 and d2 in equations 7 and 8 and compute the corwhere x is the world volume of transactions (the left-hand responding ratio of marginal utilities, that is, the gold side in equation 6, box 1). Similarly, at the lower end, silver ratio. The results are shown in figure 4 (p. 50). (Q2 d2)v2(d) = xs x. 8) (Q1 d1)v1(x) = (1 xs)x, 56 2Q/2002, Economic Perspectives NOTES 1 They became ruby slippers in the movie version. Other gold coins were minted as well (double eagles and half eagles). Bryan can be heard delivering his speech on the Web at <www.historicalvoices.org/earliest_voices/bryan.html>. 8 2 3 A troy ounce contains 480 grains. This narrative draws on Flandreau (1996), Redish (2000), Friedman and Schwartz (1963), and Dewey (1922). 4 One could also use any linear combination of goods in fixed proportion, defining the dollar as X ounces of gold and Y ounces of silver. This system, proposed by Alfred Marshall, is called symmetallism. 5 The Revised Statutes of 1874 limited its legal tender to debts of $5 or less. 6 In 1898, the Senate passed a resolution declaring that repayment of the U.S. debt in silver did not constitute a breach of faith. 9 The matter of the differing legal ratios in the two countries (15.5 in Europe and South America, 16 in North America) would necessarily have been addressed. Prior to the Civil War, costs of transportation and information probably restricted the ability of arbitrageurs to narrow the gap between the ratios across the Atlantic. 10 11 The BlandAllison Act of 1878 had required the U.S. president to invite foreign governments to an international conference on restoring bimetallism. Real per capita income grew by 20 percent in the U.S. and 27 percent in the United Kingdom during the deflation of 187996 (Maddison, 1995). 7 REFERENCES Atkinson, Fred J., 1909, Rupee prices in India, 1870 to 1908; with an examination of the causes leading to the present high level of prices, Journal of the Royal Statistical Society, Vol. 72, No. 3, September, pp. 496573. Dewey, Davis R., 1922, Financial History of the United States, New York: Longmans, Green, and Co. Drake, Louis S., 1985, Reconstruction of a bimetallic price level, Explorations in Economic History, Vol. 22, April, pp. 194219. Flandreau, Marc, 1996, The French crime of 1873: An essay on the emergence of the international gold standard, 18701880, Journal of Economic History, Vol. 56, No. 4, December, pp. 862897. , 1995, Coin memories: Estimates of the French metallic currency 18401878, Journal of European Economic History, Vol. 24, No. 2, Fall, pp. 271310. Friedman, Milton, 1990a, The crime of 1873, Journal of Political Economy, Vol. 98, No. 6, December, pp. 11591194. , 1990b, Bimetallism revisited, Journal of Economic Perspectives, Vol. 4, No. 4, Fall, pp. 85104. Federal Reserve Bank of Chicago Friedman, Milton, and Anna J. Schwartz, 1963, A Monetary History of the United States, 18671960, Princeton, NJ: Princeton University Press. Grilli, Vittorio, 1990, Managing exchange rate crises: Evidence from the 1890s, Journal of International Money and Finance, Vol. 9, No. 3, September, pp. 258275. Keynes, John M., 1913, Indian Currency and Finance, London: Macmillan and Co. League of Nations, 1930, Interim Report of the Gold Delegation of the Financial Committee, Geneva, Switzerland, Section II, p. 26. Littlefield, Henry M., 1964, The Wizard of Oz: Parable on populism, American Quarterly, Vol. 16, No. 1, Spring, pp. 4758. Maddison, Angus, 1995, Monitoring the World Economy, 18201992, Paris: Development Center of the Organization for Economic Cooperation and Development. Miller, Victoria, 1996, Exchange rate crises with domestic bank runs: Evidence from the 1890s, Journal of International Money and Finance, Vol. 15, No. 4, August, pp. 637656. 57 Redish, Angela, 2000, Bimetallism: An Economic and Historical Analysis, Cambridge, New York, and Melbourne: Cambridge University Press. Sicsic, Pierre, 1989, Estimation du stock de monnaie métallique en France à la fin du XIXe siècle, Revue Economique, Vol. 40, No. 4, July, pp. 709736. Rockoff, Hugh, 1990, The Wizard of Oz as a monetary allegory, Journal of Political Economy, Vol. 98, No. 4, August, pp. 739760. Velde, François R., and Warren E. Weber, 2000, A model of bimetallism, Journal of Political Economy, Vol. 108, No. 6, December, pp. 12101234. Rolnick, Arthur J., and Warren E. Weber, 1997, Money, inflation, and output under fiat and commodity standards, Journal of Political Economy, Vol. 105, No. 6, December, pp. 13081321. Willis, Henry Parker, 1901, A History of the Latin Monetary Union: A Study of International Monetary Action, Chicago: University of Chicago Press. 58 2Q/2002, Economic Perspectives