The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
E C O N O M I C 1 9 8 7 Can Services Be a Source o f Export-Led Growth? Evidence From the Fourth District. Major structural changes in the U.S. labor market during the past few years have resulted in a dramatic increase in the number of service jobs— and in speculation about whether service exports can sustain a regional economy. Economist Erica L. Groshen examines this question in the context of economic activity in the four largest metro politan statistical areas (MSAs) in the Fourth Federal Reserve District. 2 R E V I E W Q U A R T E R Economic Review 3 is published quar terly by the Research Department of the Federal Reserve Bank of Cleve land. Copies of the issues listed here are available through our Public Information Department, 216/579-2047. Editor: William G. Murmann Assistant Editor: Robin Ratliff Design: Michael Galka Typesetting: Liz Hanna Opinions stated in Economic Review are those of the authors and not "1 / T Identifying Amenity and Productivity JL U Cities Using Wage and Rent Differen tials. Households and firms value sitespecific characteristics of cities because of either their amenity value or their productiv ity advantage. The authors utilize the rela tionship between regional wage and rent dif ferentials to classify a sample of 38 metropolitan areas on the basis of both high and low amenity and high and low productivity. O / I FSLIC Forbearances to Stockholders L* and the Value o f Savings and Loan Shares. Policies of forbearance to stock holders of insolvent firms by federal deposit guarantors represent a wealth transfer from federal deposit insurers, and ultimately from federal taxpayers, to the stockholders of the insured institutions. Economist James B. Thomson investigates the value of forbear ances to the stockholders of stock-chartered thrift institutions that are insured by the Fed eral Savings and Loan Insurance Corporation (FSLIC). Proxies for FSLIC forbearance policy are shown to be related to the stock-market values of these institutions. necessarily those of the Federal Reserve Bank of Cleveland or of the Board of Governors of the Federal Reserve System . Material m ay be reprinted provided that the source is credited. Please send copies of reprinted materials to the editor. IS S N 0013-0281 Can Services Be a Source of Export-Led Growth? Evidence From the Fourth District by Erica L. Groshen Erica L . Groshen is an economist at the Federal Reserve Bank of Cleve land. The author would like to thank Patricia Beeson, Randall Eberts, Lorie Jackson, and Owen Flumpage for helpful comments, and Ralph Day for expert programming. Introduction The U.S. labor market is currently undergoing dra matic structural change as service jobs rapidly re place manufacturing jobs. In I960, manufacturing jobs clearly dominated the labor force, claiming 42 percent of total employment, compared with 11 percent for the service sector. Today, servicesector jobs (not including trade or transportation) claim 23 percent of employment, roughly the same percentage as manufacturing jobs. The change in employment com position within cities in the Fourth Federal Reserve District is even more pronounced. As this trend continues both locally and nationally, it is important to know whether services can sustain an economy in the same way manufacturing has done. More specifically, can the service sector pull new dollars into the local economy by exporting services? Interest in the exportability of serv ices stems from the widely held view that the vigor of regional and national economies is linked to the health of their export sectors. Trade among regions of a country plays much the same role in regional health as does international trade in the growth of national economies. When viewed within this export-base model of regional growth, the relative decline of manufacturing employment raises several issues related to the prospects and process of future regional growth. Is the export base vanishing, reducing the poten tial for further regional growth? Are there other sectors that could be transformed into part of a regional export base? This paper explores the exportabil ity of services in order to address these questions. First, the service sector and exportation methods are described, particularly for those service indus tries most likely to be exported directly. Employ ment in service industries, particularly business services, is growing faster than employment in most other sectors of the economy, and faster in the Fourth District than in the U.S. Two possible explanations for this growth suggest that trade in services may increase: services may be exported directly to consumers out of the region, or they may be exported indirectly, embodied in exported manufactured goods. Differences in consumption of services among cities are not part of the export base, while direct exports are. Service-sector export activity can be measured indirectly by estimating the variation across the U.S. in the relative concentration of service employment in local economies. Under various assumptions discussed below, large varia tions in the location quotients of a service activity across cities can be indicative of trade across areas. This technique allows identification of highly traded service industries and offers evi dence of strengths and weaknesses in individual service industries in the four largest MSAs (met ropolitan statistical areas) in the Fourth District: Cincinnati, Cleveland, Columbus, and Pittsburgh. I. What Are Services and How and Why Are They Exported? Kendrick (1986) states that “...the distinguishing characteristic of service-producing versus goodsproducing industries is that service outputs are in tangible and cannot be stored.” Although this defi nition encompasses many more economic activities than those usually classified as services, it captures the essence of what services have in common. In the discussion that follows, the term “services” refers specifically to the aggregate of lodging places, personal services, business serv ices, health services and hospitals, repair services, recreational services, legal services, educational services and schools, engineering services, accounting services, and social services. These comprise standard industrial classification (SIC) codes 70 through 89. Some discussions of the service sector include many or all of the other industries that are commonly considered to comprise the “service-producing” sector: communication; utili ties; finance, insurance, and real estate; wholesale and retail trade; and administration. This paper takes a narrower definition of services for two reasons. First, most of the growth in the serviceproducing sector of late has been in the narrower class of services, particularly in business services, which seem particularly amenable to export activ ity. Second, the data that were available for this study cover only this portion of the service sector. This paper concentrates on profes sional and business services (also called the “producer” services), which together account for more than a third of employment in services. The professional services include legal, accounting, engineering, and educational services. Business services include services normally rendered to places of business rather than to final consumers, and comprise the following: advertising; services to buildings; computer and data processing serv ices; management, consulting, and public rela tions services; equipment rental and leasing; credit reporting and collection agencies; direct mail advertising services; blueprinting and photo copying services; commercial photography, art, and graphics; stenographic and duplicating serv ices; personnel supply agencies; and commercial research and development. As is evident from this list, these activities often require a face-to-face meeting, or at least telephone contact, between supplier and consumer. In many cases they are done at the behest or on the premises of the consumer, so that the services are not storable. Although these features imply that services cannot be exported by the same means as manufactured goods (for example, shipping by rail or truck), they do not eliminate the possibility of service-sector exports. There are two ways to export services directly: activities may be transported and sold to persons outside the area, or individuals may travel to the area to purchase services. Sometimes consultants visit their clients; other times clients travel to consultants. Data is transmitted to programmers or to a distant mainframe computer. Construction equipment is transported to leasors. Establishing the possibility that serv ices may be transported addresses only one side of the issue: the necessary condition. The other side of the question is, why would they be traded? The export-base growth model, the simplest explanation for the existence of regional trade, is based on production economies of scale. If largescale production reduces average production cost for some products, the minimum efficient scale (MES) may exceed the needs of the surrounding community. Then, welfare of all the communities will be maximized by specialization and trade among communities. Each community produces a subset of the products with scale economies, and this becomes their export base. The communities use proceeds from exports to import goods that are not produced locally. Regional growth is the result of expansion of the export base. Products with no economies of scale (that is, in wiiich MES is small relative to local demand) are pro duced and consumed locally. The prevalence of interregional and international trade in manufac tured goods is assumed to stem from larger MES in manufacturing than in service production. This export-base growth model is the source of the conventional view of the service-producing sector as one that grows only as a result of a healthy manufacturing sector and that does not generate “new” income for an area. The manufacturing sector, characterized by larger firms, generates income for the area through the sale of goods outside the region or country. Serv ices, on the other hand, are provided by small local companies, and merely recycle within the local economy the income created by the manu facturing sector. This perception of the service sec tor as dependent upon the manufacturing sector, however, has changed recently. To the extent that technological changes increase the MES of service provision, we can expect increases in servicesector trade. And, although they are not storable, services can be exported directly and conse quently have the potential to spur local economic expansion. The question centers on the extent to which services are, and will be, traded relative to the manufacturing sector. 3 II. Changing Industrial Composition o f Employment in the U.S. and Ohio This section begins with a description of recent changes in the composition of employment in the United States and in the Fourth District, focusing particularly on employment growth in the services and their components. The rest of the section considers the reasons to expect growth in servicesector employment and to link growth with trade in services. Industrial Composition o f Employment in Ohio and the U.S.: 1970 and 1986 O h io IVQ United States IVQ IVQ IVQ SOURCE: U.S. Bureau o f Labor Statistics. FI GURE 1 Figure 1 summarizes the changes in industrial composition of the work force in Ohio and in the nation. Overall, Ohio employ ment since 1970 has become more similar to that of the nation as a whole. Ohio entered the 1970s with only 15 percent of its employment in the service industries, compared to 17 percent for the U.S. By 1986, Ohio had almost matched the national figure of 23 percent of the labor force employed in the service industries. Since 1970, the number of jobs in U.S. service industries has doubled. In contrast, the manufacturing indus tries gained no jobs, so manufacturing’s share of total U.S. employment fell from 28 to 19 percent. In the U.S. and in Ohio, the number of service jobs now almost equals or exceeds the number of manufacturing jobs. But the relative growth of services in Ohio and in the major cities of the Fourth District followed a different pattern than in the U.S. At first, the region lagged behind national growth; now it appears to be catching up. Table 1 sum marizes the pattern of growth of employment in the service industries in the U.S., in Ohio, and in the four largest cities of the Fourth District. The relative growth of services in Ohio since 1970 came in two phases. Until 1983, shrinkage of manufacturing employment in Ohio, combined with modest (but below nationalaverage) service growth, led to increases in the service industries’ share of employment. However, since late 1983/early 1984, above-average growth in business services in Ohio has led to growth above the national average for the state’s service industries as a whole. Even though the state’s share of employment in the service industries now nearly matches that of the U.S., Ohio’s 1986 rate of service-industry job creation of 7.1 percent continues to exceed the U.S. rate of 5.6 percent. Because manufacturing was heavily concentrated in Ohio, the decline in manufactur ing employment since 1970 was particularly dra matic here: employment share shrank from 37 to 24 percent. This resulted from the net elimination of almost a quarter of the state’s manufacturing jobs and from the growth of other sectors, partic ularly the service industries. Where is this recent growth taking place? Nationally, the two largest components of services are health services and business services, which together account for more than half of total service-industry employment. Health services sup plied 38 percent of the growth in services until 1982. Since then, it has supplied only 17 percent of service-sector growth and has not increased its share of national employment. In contrast, busi ness services contributed 22 percent of service growth until 1982 and 38 percent of growth since then. Thus, although health services were an impor tant source of service growth through the 1970s and early 1980s, the mid-1980s have seen a rapid expansion of employment in business services. Ohio has consistently kept pace with the growth of health services, maintaining an edge over the U.S. in percent employed in that in dustry. In contrast, throughout the 1970s and early 1980s, the state lagged the U.S. in the level and growth of business service employment, but now exceeds the national pace of expansion. In 1986, the growth of business services in the state was 13-1 percent, compared to 8.5 percent for the U.S. Patterns of growth vary somewhat among MSAs within the Fourth District. Pittsburgh and Cleveland have the largest proportion of employment in the service sector. The strongest similarity among the four MSAs is the widening gap between their expansion in services and that of the U.S. (which includes rural areas) since 1984. Because services tend to be concentrated in urban areas, a city that only matches, instead of exceeds, the national average in service employ ment probably has a relative lack of services. Summary o f Service Employment Growth, 1970-1986 Total % of 19861 1986 Average Annual G row th Rate 1970-79 1980-84 1985 1986 United States Total Manufacturing Services Health Business 100,167 19,186 23,072 6,586 4,809 100.0 19.2 23.0 6.6 4.8 2.7 1.0 4.5 5.6 6.3 1.1 -1.1 3-8 3.8 7.0 4.8 -0.2 6.1 3.5 10.6 4.2 0.2 5.6 4.4 8.5 Ohio Total Manufacturing Services Health Business 4,475 1,109 999 344 192 100.0 24.8 22.3 7.7 -0.6 -2.8 2.2 5.2 0.0 6.4 4.3 1.6 -0.2 4.0 5.5 n.a. 3.5 5.8 3.9 11.8 5.0 -0.7 7.1 5.1 13-1 Cincinnati Total Manufacturing Services Health Business 651 148 155 49 35 100.0 22.7 23.8 7.5 5.4 2.0 0.2 4.7 n.a. n.a. 0.0 -2.7 2.9 n.a. n.a. 6.5 3.5 7.2 1.7 12.5 6.2 0.1 8.5 4.6 14.0 Cleveland Total Manufacturing Services Health Business 881 206 224 72 50 100.0 23.4 25.4 8.2 5.7 0.8 -0.8 3.2 n.a. n.a. -1.2 -4.3 0.9 n.a. n.a. 4.0 -1.7 6.5 5.0 13.9 4.3 -0.7 7.0 6.1 8.9 Columbus Total Manufacturing Services Health Business 630 106 146 38 32 100.0 16.8 23.2 6.0 5.1 2.6 0.2 5.5 n.a. n.a. 0.7 -1.8 4.1 n.a. n.a. 7.3 1.3 7.7 6.7 -0.1 6.3 7.9 9.3 5.1 15.3 Pittsburgh Total Manufacturing Services Health Business 842 129 253 80 n.a. 100.0 15.3 30.0 9.5 n.a. n.a. n.a. n.a. n.a. n.a. -1.9 -8.1 3.6 3.5 n.a. 3.5 -4.8 4.7 4.1 n.a. 3-3 -6.7 4.5 2.7 n.a. 1. In thousands, n.a.: not available. SOURCE: Bureau o f Labor Statistics Employment and Earnings Reports. In the four largest MSAs in the Dis trict, business service growth has risen sharply since 1984. In 1983, the proportion of people employed in business services in Cincinnati, Cleveland, and Columbus almost equaled the national average. As of the end of 1986, all three cities had 18 to 20 percent more employees than the national average in this sector. This recent growth in the service industries (particularly in business services) has been dramatic, and service employment growth can be expected to continue to exceed manufac turing growth for four reasons: increasing afflu ence, increased standardization, lower delivery costs, and technological changes that raise the relative cost of small-scale internal provision. In general, employment growth in manufacturing has been limited by rapid produc tivity increases, not by decline in demand for its output. Measured in terms of its share of gross national product (GNP), manufacturing has not shrunk. This has meant greater affluence for con sumers, who have spent an increasing portion of 6 their wealth on services due to a high income elasticity of demand for services. Consumers who are already affluent tend to spend disproportion ately more of further increases in income on the purchase of services, rather than on agricultural products or manufactured goods. Beeson and Bryan (1986) argue that just as increasing produc tivity in nonmanufactured goods (for example, agriculture) in the early twentieth century was associated with a shift toward manufacturing in consumption and employment, so an increase in manufacturing productivity now leads to the serv ice boom. The growth of services is a sign of our increased affluence. Certainly, the growth of health and personal services fits the pattern of increased affluence, but how does this explain the expan sion of business services? Affluence may have shifted consumption toward final products whose component industries tend to use business serv ices most. For instance, an increasingly litigious society needs more legal photocopy shops for its attorneys. Or, increased demand for differentiated or luxury goods will raise demand for advertising services, because luxuries and differentiated products are advertised more heavily than are essen tials or standardized goods. While this explana tion predicts growth in the service sector, it does not predict increases in service-sector trade. The other three explanations for growth (standardization, falling delivery costs, and technological change) have implications for trade because they suggest that the production of services is now increasingly subject to economies of scale. That is, larger size may now enhance efficiency in service provision. First, management and other tech nologies have become specialized and routinized to the extent that there are new economies of scale in consolidating them across company lines. Stanback, et al. (1981) suggest that increased spe cialization found in large firms leads to routinization of functions. Once routine, these functions can be separated from other functions of manage ment. Firms with consistent demand for the serv ice may still provide it internally. Firms with intermittent demand will purchase the service as needed from vendors who specialize in its rou tine provision. To accelerate the process, informa tion, transportation, and communication have become less costly, reducing the necessity for essential components of management to be located near the scene of production, either geo graphically or within the same firm. Furthermore, technological changes may have raised the relative cost of providing intermittent or small amounts of services inter nally. Business services provide a way to purchase some portion of the services of an indivisible technology7, or to meet peak loads (for example, due to seasonal, cyclical, or unanticipated de mand growth). The complexity or the larger effi cient scale of new products used by businesses could make it more economical to contract out for services rather than provide them internally, especially for small establishments. Examples are the use of external computer time-sharing, data processing, and photocopying services, as well as the use of temporary personnel. Taken together, these three points suggest that net costs to separation of compo nents of management and production processes have fallen; that is, scale economies have risen in business service provision. This may explain why between 1975 and 1984 (according to County Business Pattern Data), large (more than 100 employees) establishments’ share of employment in business services rose from 44 to 49 percent. In addition, the Census of Services notes that the percentage of business service establishments that were part of firms with three or more estab lishments rose from 4.7 in 1972 to 11.4 in 1982. An increase in the MES of service provision makes services more similar to manufacturing. In particular, it makes trade more likely. Business service growth comprises two elements: growth of employment because of increased production of services, and increased outsourcing of services formerly provided inter nally (that is, transfer of employment to service firms). During 1975 to 1984, while the size of business service establishments grew, large manufacturing establishments decreased their employment share from 74 to 63 percent, and the average size of manufacturing establishments shrank from 60 to 55 employees. The decline in the average size of manufacturing establishments should increase demand for business services; of course, outsourcing of business services may also be a source of the decline. Growth resulting from increased demand for business services could be due to the increased affluence previously mentioned, or to increased productivity of services (assuming a highly price-sensitive demand). Unfortunately, the intangible nature of services makes it difficult to measure productivity of these industries. Serv ice output (work performed) is difficult to distin guish from input (person-hours). Thus, usual attempts to measure productivity changes have detected only small or negative improvements in the service sector. However, the purchasers of services provide a clue to the direction of changes in productivity. If management acts to maximize profits, their purchase of a service (as opposed to internal provision) indicates that they consider the purchase to be the least-cost alterna tive. It follows, then, that increasing demand may be the strongest available evidence of increasing productivity in business services. The strength of service-sector em ployment growth in the Fourth District may be due to catch-up growth of locally consumed serv ices or to establishment of service exports to non local consumers. Although growth is encouraging for the region in either case, only growth due to exports adds to the economic base of the region. III. The Regional Economic Base and Measurement of Service Export Activity Three explanations for service-sector growth sug gest that many services may be increasingly ex portable from one location to another. Thus, a region (or city or country) could become a service exporter as part of its economic base. Export activ ity may be direct (sales of services across boun daries) or indirect (sales of goods containing embodied services across boundaries). Because of the regional specialization of economic activity, and because business services are purchased pri marily as intermediate goods, differences in local production of business services are related to dif ferences in the regional concentration of other industries, as well as to direct export activity. If a region’s export industries are intensive users of services, that region indirectly exports services. But these are exports consistent with the old view of services as a secondary, sup porting sector, rather than as an independent part of the economic base. Services that are primarily exported indirectly do not attract “new” dollars into the region directly, unless they attract other producers by making the region more competitive. Tliis is one reason for the concern that services are not a viable part of an economic export base for a region (see, for example, Cohen and Zysman [1987] and Perna [1987]). For policy purposes, identification of potential indirect exports may be less relevant than identification of direct exports. Alternatively, expansion in the mar ket for business services outside the firm implies the potential both for nonlocal provision of serv ices, and for incentives for the formation of direct service-exporting companies. In this case, services are clearly part of the economic base of the area. Services may be traded across city, regional, or national boundaries. In 1981, the United States was a net exporter of business serv ices. The growing importance of international trade in services has been recognized by the Con ference Board, which recently issued a report emphasizing the need for lower import restric tions for services among our trade partners (Basche [1986]). Keil and Mack (1986) suggest that a useful measure of the export potential of an industry is the extent to which employment share (relative to the national average) varies among cities. Local employment share divided by national share is called the location quotient. In their framework, if location quotients for an industry vary strongly among cities, the cities with larger shares are probably exporting that industry to cities (or, perhaps, countries) with smaller quotients. Cities use the proceeds to purchase the products of industries in which they have low location quotients. If export activity is heavy in an industry, one would expect to see many cities with very little employment in the industry, and others with heavy concentrations of jobs in the industry. If little export activity occurs, all cities will have about the same percentage of employ ment in the industry. Thus, a service industry with a high variance of location quotients across areas is likely to be an industry with trade activity and, therefore, export potential. O f course, the prod ucts of such an industry are also likely to be imported (and, thus, to be a source of dollar out flow) for many cities. Exports per capita (based on size of the city’s labor force) of industry in city are defined as quantity produced minus local consumption ( all in per capita terms: i (X'j) (1 ) Csp , v„ y„-c;„. j (Q,j), Qi} Keil and Mack measure by em ployment in industry as a proportion of total employment in city divided by the industry aver age employment share across the nation’s cities. This is defined as the location quotient for indus try in city (/.. ). Two assumptions are made. First, labor productivity is constant across cities and industries (that is, for all Second, all consumption patterns are con stant across cities (that is, = for all so: (2) = Under these assumptions, we can take the sample variance of each side for each industry as follows: (3) = If industry is characterized by a high trade volume, some cities thus will have high imports ( ^ < < 0 ) and others will have high exports (X ;/» 0 ) . Therefore, the variance of the for industry7 across cities will be high. On the other hand, if little of an industry’s product is traded, all will be of similar size, so their var iance across cities will be small for the industry. If the two assumptions of identical productivity and consumption patterns across regions hold, variation in the relative size of the labor force in industries across cities is directly j i i,j). i j Qjj/L,j = Q/L, C{j Ct, Xtj L'j . (Q/L) . Cr i s)(X tp Xtj s (Q/L)2 i Xtj s i j), 7 Business Service Industry Consumption in 1981 All Industries Business Services Intermediate demand Final demand Personal consumption Inventory and investment Net exports Government purchases Total 52.4 32.8 8.9 0.6 10.1 100.0% 82.0% 18.0 8.6 0.0 1.5 7.8 100.0% Industry Share of Final Demand Share of of all Industries Total Output1 Share of Business Services Intermediate Output2 Relative Use of Business Services3 0.6 11.8 7.3 7.4 15.3 15.6 3-0 1.6 2.8 15.1 3.9 11.3 1.9 1.9 3.8 5.1 9.8 0.4 19.3 16.7 3.6 1.6 4.2 11.0 3.8 8.8 1.8 15.3 14.0 0.3 2.1* 0.7 0.6 0.8 0.9 0.2 2.0* 1.8* 0.5 1.2 1.5 0.9 1.2 1.2 1.2 100.0 100.0 Disposition of Total Receipts____________________ Oil, mining, agriculture, ordnance and forestry Construction Nondurable manufacturing Durable manufacturing Transportation Communication Utilities Wholesale and retail trade Finance and insurance Real estate Hotels, personal and repair services Business services Eating and drinking places Automobile and recreation services Health and professional services Government enterprises Total 47.6% 1.3 5.3 2.5 2.0 5.1 0.5 10.3 2.9 1.5 0.7 21.9 7.0 3.8 1.5 8.1 2.2 2.3 5.9 0.6 100.0 1. Total measured output by enterprises in each industry7, including d ou ble counting due to use o f output as intermediate g o o d s by other enterprises. 2. Total consum ption o f business services as an intermediate g o o d by each industry. 3. This is the num ber in the third colum n divided by the number in the second colum n. Numbers over 1.0 indicate greater-than-average use o f business services as an intermediate g o o d ; numbers b e lo w 1.0 indicate less-than-average use o f business services. ‘ Industry with relative use greater than 1.7. SOURCE: Planting, Mark A , “Input-Output Accounts o f the U.S. Econom y, 1981,” Survey’ o f Current Business, vol. 67, no. 1, January 1987. TABLE 2 related to variation in trade activity among cities for that industry. If the two assumptions do not hold, the variations in the location quotients may not be detecting trade activity. The remainder of this section and the next section explore the plausibility of these assumptions. Differences in the relative cost of local factors of production can lead to differences in location quotients among areas. Because of cost-minimizing substitution among inputs by service providers, such differences cause variation in the labor input even if quantity produced does not vary. Conversely, if labor is more productive in one industry than in another, variation in employment will understate the relative value of exports in the more productive industry. (How ever, to the extent that the focus of the exercise is to identify employment creation by export activ ity, this bias is appropriate.) A related problem arises if there are systematic biases in the way labor input is measured. For example, industries with more variation in their use of part-time labor may appear to have more export activity by this measure than one where the full-time/part-time ratio is consistent among most employers. It is difficult to account for regional differences in service-sector productivity, because service-sector output is not available. National level estimates use income accounts, which are not available for regions. As a first attempt to check the plausibility of the equal productivity assump tion, the results that follow (based on employ ment) were compared to calculations based on variation in receipts. Differences were negligible. IV. Differences in Consumption of Services Among Areas Differences in regional concentration of services may be related to consumption patterns of cities’ residents and businesses, rather than to direct export activity. Most variations in personal serv ices location quotients are probably due to varia tion in city residents’ consumption. For example, regional differences in the taste for hairdressing or in climate could generate nonexport-based variations in location quotients. For this reason, the analysis below excludes personal services. Controlling for variation due to indirect exports (differences in consumption by cities’ businesses) is more problematic. The technique applied below does not distinguish between direct and indirect export activity. How ever, if the users of services are regionally dispersed, indirect exports are likely to be a smaller portion of total exports than if the users tend to be concentrated geographically. Table 2 presents national consump tion patterns for business services in 1981, the latest year available. The upper panel compares the disposition of total output, as measured by receipts, of all domestic industries to that of bus iness services. On average, about half (47.6 per cent) of the output of U.S. firms is purchased by other firms as an input to their own production. The other half is produced for final demand, pri marily personal consumption (about 33 percent of the total) and government purchases (10.1 percent of the total). Investment and inventory changes consume another 8.9 percent, while net exports were less than 1 percent of total output. The pattern for business services is markedly different. Intermediate demand con sumes 82.0 percent of total output, with the remainder fairly evenly split between personal consumption and the government. In short, the demand for business services is indeed primarily a derived demand from that for other industries. Thus, variation in the level of personal consump tion is not likely to be a significant source of vari ation in the provision of business services. It is also interesting to note that the U.S. balance of trade in business services, although small (only 1.5 percent of business service output), is better than the average for U.S. industries. The second panel of table 2 indi cates which industries are the largest consumers of business services. Two factors are important in the level of consumption: the relative size of the consuming industry and its relative use of busi ness services. The first column of the table com pares industries by their share of output consumed as final demand. The second column shows each industry’s share of total output, which includes products sold to other firms as intermediate goods. The third column lists the share of the total output of business services consumed by each industry. Four industries emerge as heavy (double-digit) users of business services: whole sale and retail trade; construction; durable manu facturing; and nondurable manufacturing. The trade industries alone use almost 22 percent of the output of business services, not only because the sector is large, but also because the industries have a high relative use of these services. Con struction also combines the influences of large industry and heavy use. In contrast, manufactur ing firms are below-average users of business serv ices. However, because of the size of the sector, they consume 24 percent of the output of busi ness service firms. O f the four largest consumers of business services, the top two (construction and trade) are very regionally dispersed, one is some what dispersed (nondurable manufacturing), and one is fairly concentrated (durable manufactur ing). Construction and wholesale and retail trade have two features in common: seasonal demand and small establishment sizes. Their prominence is consistent with the hypothesis that business services provide smoothing and scale economies to their customers. Thus, by this cursory analysis, the evidence is somewhat mixed. Some of the regional variation in service employment no doubt derives from regional variation in manufac turing consumption. Nevertheless, much of the regional concentration of services is probably due to direct exports, and therefore is a viable part of an economic export base. V. Signs of Service Export Activity by Industry Which service industries are characterized by the most trade activity? Unfortunately, statistics on the service sector are not plentiful. However, in 1982 the U.S. Department of Commerce conducted an economic census of the service sector. From that snapshot of services in the U.S. and in Fourth Dis trict cities, we can get some indication of our serv ice industry strengths and weaknesses. Because much of the most interesting growth in services took place after 1982, the conclusions we can reach about current strengths and export activity from these data are at best limited. The data allow identification of the baseline distribution of industries. However, it is not clear whether sub sequent growth took place in those industries the region lacked or in those industries showing rela tive strength. Table 3 ranks three- and four-digit hotel and producer service industries by the standard deviation of employment location quo tients across MSAs in 1982. The data used are employment totals in taxable (that is, for-profit) 9 Service Industry Export Activity Among MSAs in 1982 Service Industry 10 Percent of Total Employment1 Export Activity Group2 0.06 0.02 1.26 0.06 0.03 0.02 0.61 0.04 0.08 0.05 0.03 0.53 0.03 0.09 0.23 0.55 0.19 0.14 0.55 0.86 0.08 0.47 0.25 0.83 0.81 High High M-High M-High Moderate Moderate Moderate Moderate Moderate Moderate Moderate Moderate Moderate Moderate Moderate Moderate Moderate Moderate M-Low M-Low M-Low Low Low Low Low Low Low Research and development laboratories Schools and educational services, not elsewhere classified Hotels, motels, and lodging places Direct mail advertising services Surveying services Interior designing Engineering services Testing laboratories and facilities Commercial photography, art and graphics Correspondence and vocational schools Stenographic and reproduction services Commercial sports and recreation Blueprinting and photocopying Photographic finishing labs Advertising services Computer and data processing services Equipment rental and leasing services Architectural services Management and public relations consulting Personnel supply services Credit reporting and collection agencies Detective agencies and protection services Other repair shops and related services Legal services Services to dwellings and other buildings Electrical and electronic repair shops Accounting, audit, and bookkeeping services 0.13 0.47 Standard Deviation of Location Quotients 0.192 0.161 0.106 0.100 0.076 0.075 0.074 0.073 0.071 0.066 0.065 0.063 0.061 0.059 0.056 0.056 0.054 0.052 0.045 0.042 0.038 0.031 0.028 0.028 0.026 0.023 0.018 1. Industry em ploym ent as a percent o f total em ploym ent in all industries in all U.S. MSAs. 2. Export activity group is a som ewhat arbitrary grouping o f the industries on the basis o f standard deviation o f the location quotients in that industry' across all U.S. MSAs (sh ow n in the last tw o colum ns). The follow ing service industries w ere excluded from this analysis: trailering parks and camps; all health services; bow ling alleys and billiards and p o o l establishments; telephone answering services; other serv ices; photographic portrait studios; funeral services and crematories; automotive sendees; autom obile parking; all personal services; reup holstery and furniture repair; other health services; other business services, not elsewhere classified; auto rental and leasing without drivers. SOURCES: U.S. Department o f Com m erce, Census o f Service Industries 1982 (service industry em ploym ent by city); Department o f C om m erce County Business Patterns 1982 (total em ploym ent by city). TABLE 3 establishments by industry and MSA from the 1982 U.S. Department of Commerce Census of Services. Personal services are excluded from the analysis because variations are probably due primarily to variations in demographic characteristics. Health services are excluded because these data omit the tax-exempt sector and hospitals, which are both important employers in health services. The industries are grouped from high evidence of export activity to low export activity, based on the standard deviation of loca tion quotients for the industry. At the top of the list are research and development laboratories and private technical schools, reflecting the inherent exportability of knowledge (although the variation in research and development employment may be due to indirect exports). The large variance in employment by hotels, motels, and other lodging places among MSAs reflects the variation in the extent of tourism and convention activity among cities. On the other hand, at the bottom of the list are legal services and accounting, audit, and bookkeeping services. Most of the output of these two industries is probably consumed locally. For comparison, table 4 presents standard deviations for two-digit manufacturing industries. The higher level of aggregation in these data should tend to reduce variation. Never theless, the standard deviations of location quo tients among MSAs of these industries are, in general, higher than those of the service indus tries. Application of the same export activity groupings used in table 3 puts more than half (10 of 19) of the manufacturing industries in the high or moderately high categories, compared to four of the 27 service industries included. There is, however, considerable overlap in the ranges covered by the standard deviations in services and manufacturing. For example, three service indus tries (R&D labs, schools not elsewhere classified, and hotels) show more evidence of export activ ity to other regions than does primary metals. Thus, while the more traditional view of manufac turing as inherently export industries and services as inherently local industries has some validity, a subset of service industries is at least as geo graphically concentrated as the bulk of manufac turing industries. The pattern of export activity across MSAs in the District shows no strong con sistency, except perhaps for the lack of concentra tion in the most heavily traded industries. Each city has a unique pattern of strengths, which is to be expected if the proximity of cities increases the probability that they trade heavily with one another. Cleveland’s concentration in private, forprofit schools and educational services, not else where classified, is the only entry from the high or moderately high exportability groups. A few industries appear more than once. For instance, personnel supply services is prominent in three of the four cities. Management and public rela tions consulting; detective and protection agen- Manufacturing Industry Export Activity Among MSAs in 1982 Manufacturing Industry (SIC) Percent of Total Employment1 Export Activity Group2 0.12 High High High High High High M-High M-High M-High M-High Moderate Moderate Moderate Moderate Moderate Moderate Moderate M-Low Low Petroleum and coal (29) Textile mill products (22) Leather and leather products (31) Furniture and fixtures (25) Lumber and wood products (24) Instruments and related products (38) Primary metals (33) Miscellaneous manufacturing (39) Paper and allied products (26) Transportation equipment (37) Apparel and other textile products (23) Stone, clay, and glass products (32) Chemical and allied products (28) Rubber and miscellaneous plastics (30) Electrical and electronic equipment (36) Food and kindred products (20) Nonelectrical machinery (35) Fabricated metal products (34) Printing and publishing (27) 0.53 0.15 0.36 0.29 0.75 0.85 0.42 0.56 1.43 1.17 0.49 0.99 0.69 2.31 1.63 2.63 1.87 1.68 Standard Deviation of Location Quotients 0.250 0.237 0.213 0.151 0.145 0.131 0.105 0.101 0.094 0.088 0.082 0.078 0.078 0.061 0.055 0.050 0.049 0.042 0.033 1. Industry em ploym ent as a percent o f total em ploym ent in all U.S. MSAs. 2. Export activity group is relative to the service industry standard deviations, as defined in the text and in table 3SOURCES: U.S. Department o f Com m erce, Census o f Manufactures 1982 (manufacturing industry em ploym ent by city); Department o f C om m erce County Business Patterns 1982 (total em ploym ent by city). TABLE 4 VI. Exportability and the Strengths and Weaknesses of the Service Sector in the District Does the Fourth District export services? This sec tion examines evidence on the extent of concen tration of Fourth District MSAs in export-intensive service industries. Table 5 shows the industries in which the four major MSAs in the District were particularly strong. Also listed are the exportabil ity group of each industry and the estimated number of jobs over the national average. cies; and accounting, audit, and bookkeeping services each appear twice. These are the only cases of repetition. Personnel supply services consists of two main components: employment agencies and temporary help suppliers. From 1982 to 1984, temporary employment was the fastest-growing industry, with more than 50,000 employees in the U.S. (see Carey and Hazelbaker [1986]). Although many of the jobs are low-skill (laborer and cleri cal positions), there are two high-skill sectors of the market: engineering and technical. Fourth District Service Industry Employment Surpluses by MSA in 1982 Export Activity _______ Group1_______ Employment _______ Surplus2 Cincinnati Commercial photography, art and graphics Commercial sports and recreation Photographic finishing labs Management and public relations consulting Personnel supply services Services to dwellings and other buildings Moderate Moderate Moderate M-Low M-Low Low 220 230 270 300 1,550 570 High Moderate M-Low Low Low Low 230 220 1,150 860 550 610 Moderate M-Low M-Low Low Low 400 1,960 1,330 540 250 Moderate Moderate M-Low Low 5,810 370 1,010 840 Cleveland Schools and educational services, not elsewhere classified Stenographic and reproduction services Personnel supply services Detective agencies and protection services Other repair shops and related services Accounting, audit, and bookkeeping services Columbus Architectural services Management and public relations consulting Credit reporting and collection agencies Other repair shops and related services Services to dwellings and other buildings Pittsburgh Engineering services Testing laboratories and facilities Personnel supply services Detective agencies and protection services 1. Export activity group is a grouping o f industries by the standard deviation o f the location quotients in that industry across all U.S. MSAs. See table 1 and text for explanation. 2. Surpluses are rounded to the nearest 10. Only industries with em ploym ent surpluses o f m ore than 200 are included. SOURCES: U.S. Department o f Com m erce, Census o f Service Industries 1982 (service industry em ploym ent by city); Department o f C om m erce County Business Patterns 1982 (total em ploym ent by city). TABLE 5 Temporary work affords workers an opportunity for flexible schedules and exper imentation with positions. It allows employers to adjust to temporary employment needs due to seasonal or cyclical fluctuations, to employee absences, or to demand shifts of dubious per manence. The size of the industry in these Fourth District cities may indicate that local employers were more hesitant about adding permanent per sonnel than were others nationally. On the other hand, it may have signaled the beginning of growth: that is, as an indication of positions soon to be added to permanent staff. If the prominence of personnel supply services results from the high-skill sectors, it may signal that the engineering and technical schools in Cleveland, Columbus, and Pittsburgh produce a concentration of technically skilled people who export some of their services to areas without such schools. The most striking entry among the surpluses is engineering services in Pittsburgh; these exports generate about 5,800 jobs for the city’s economy, over and above the jobs demanded for the local economy. Also of interest is the concentration by some of the cities in industries that are not, in general, characterized by export activity. In particular, Cleveland and Columbus show evidence of concentration in accounting, audit, and bookkeeping services, although this industry ranks the lowest in signs of export activity of all 27 industries analyzed. Per haps this signals the beginning of a trend toward trade in these industries. Patterns of consistency across cities are much stronger in the region’s service employment deficits. Table 6 shows the indus tries in which the four MSAs were apparently net importers. In general, these four large cities import legal, research, hotel, computing, and engineering services. R&D labs and legal services both employ significantly fewer people than the national average in all four major Fourth District MSAs. The following industries appear three times on the lists: engineering services; computer Fourth District Service Industry Employment Deficits by MSA in 1982 Export Activity _______ Group1_______ Employment _______ Deficit2 Cincinnati Research and development laboratories Hotels, motels, and lodging places Engineering services Computer and data processing services Architectural services Legal services Accounting, audit, and bookkeeping services High M-High Moderate Moderate Moderate Low Low 240 680 740 610 280 1,700 410 High M-High Moderate Moderate Moderate M-Low Low Low 220 3,080 490 1,280 300 990 550 990 High Moderate Moderate M-Low Low Low 240 490 480 490 1,840 360 High M-High Moderate Moderate Moderate Moderate M-Low Low Low Low Low 480 3,070 1,210 1,440 440 420 350 200 1,800 1,000 400 Cleveland Research and development laboratories Hotels, motels, and lodging places Engineering services Computer and data processing services Equipment rental and leasing services Management and public relations consulting Legal services Services to dwellings and other buildings Columbus Research and development laboratories Engineering services Commercial sports and recreation Personnel supply services Detective agencies and protection services Legal services Pittsburgh Research and development laboratories Hotels, motels, and lodging places Commercial sports and recreation Computer and data processing services Equipment rental and leasing services Architectural services Management and public relations consulting Other repair shops and related services Legal services Services to dwellings and other buildings Accounting, audit, and bookkeeping services 1. Export activity group is a grouping o f industries by the standard deviation o f the location quotients in that industry across all U.S. MSAs. See table 1 and text for explanation. 2. Deficits are rounded to the nearest 10. Only industries with em ploym ent deficits o f m ore than 200 are included. SOURCES: U.S. Department o f Com m erce, Census o f Service Industries 1982 (service industry em ploym ent by city); Department o f C om merce County Business Patterns 1982 (total em ploym ent by city). TABLE 6 and data processing services; and hotels, motels, and other lodging places. Hotels, in particular, stand out as a major deficit in Cleveland and Pittsburgh. This suggests that these cities “import” conventions and tourism; that is, people leave these cities to vacation or to attend conventions. The lack of local engineering services employment in Cin cinnati, Cleveland, and Columbus may be due to imports of those services from Pittsburgh. The regional deficits in computer and data processing services employment suggest heavy importation of these services or slowness to begin using them (that is, deficient local demand), as of 1982. This deficit is particularly troubling because between 1974 and 1984, employment nationwide in this industry grew by 250 percent. VII. Conclusion The major points of this paper may be summarized as follows: 1. The composition of employ ment in the United States and in the Fourth Dis- 14 trict is shifting toward services. The Fourth Dis trict currently exceeds the nation in the growth of services as a whole and in the fast-growing busi ness services. 2. Increased minimum efficient scale (MES) for the provision of producer serv ices may be a basic reason for their growth. This implies that trade in services may increase, although as of 1982, there was apparently less trade in producer services than in manufacturing. Services, to the extent that they are exported directly to consumers outside a region, are viable members of the regional economic base. 3. In the producer services (in 1982), the four largest cities in the Fourth District each specialized in a different combination of services; only personnel supply services was an industry of concentration for more than two cit ies. The largest concentration was engineering services in Pittsburgh, which generated about 5,800 extra jobs. 4. Fourth District import patterns were more consistent across cities; employment deficits were pronounced in legal, research, hotel, computing, and engineering services for at least three cities out of four. This information is particularly relevant to the Fourth District because of the recent national and regional decline in manufac turing employment. Can we expect the service industries to replace lost manufacturing dollars? If economies of scale rise in the services, interre gional and international trade in services should continue to grow. There is no reason to expect dollars drawn into a region by services sales to have a smaller impact on wealth than dollars earned through manufacturing activity (assuming that income earned from service firms is spent similarly to that from manufacturing firms). The recent growth in services in the Fourth District suggests that they may be able to replace some of the lost manufacturing dollars, but it is unclear just how much replacement any region, and the Fourth District in particular, can expect. References Basche, James R. “Eliminating Barriers to Interna tional Trade and Investment in Services,” Research Bulletin No. 200, The Conference Board, New York, 1986. Beeson, Patricia E., and Michael F. Bryan. “The Emerging Service Economy,” > Federal Reserve Bank of Cleveland, June 15, 1986. Economic Com mentary Browne, Lynn E. “Taking In Each Other’s Laundry—The Service Economy,” Federal Reserve Bank of Boston, July/August 1986, pp. 20-31. land Economic Review, New Eng Bryan, Michael F., and Ralph L Day. “Views from the Ohio Manufacturing Index,” Federal Reserve Bank of Cleveland, Quarter 1 1987, pp. 20-30. Economic Review, Carey, Max L., and Kim L. Hazelbaker. “Employ ment Growth in the Temporary Help Indus try,” vol. 109, no. 4 (April 1986), pp. 37-44. Monthly Labor Review, McCrackin, Bobbie H. “Why are Business and Professional Services Growing So Rapidly?” Federal Reserve Bank of Atlanta, August 1985, pp. 14-28. Economic Review, Perna, Nicholas S. “The Shift from Manufacturing to Services: A Concerned View,” Federal Reserve Bank of Boston, January/February 1987, pp. 30-38. Economic Review, New England Planting, Mark A. “Input-Output Accounts of the U.S. Economy, 1981,” vol. 67, no. 1 (January71987), pp. 42-58. ness, Survey of Current Busi Stanback, Thomas Jr., Peter J. Bearse, TheirryJ. Noyelle, and Robert A. Karasek. (Conservation of Human Resources Series; 20), Totowa, New Jersey: Allanheld, Osmun & Co. Publishers, Inc., 1981. New Economy Services: The Urquhan, Michael. “The Employment Shift to Services: Where Did it Come From?” vol. 107, no. 4 (April 1984), pp. 15-22. Labor Review, Monthly 1 5 Cohen, Stephen S., and John Zysman. “The Myth of a Post-Industrial Economy,” February/March 1987, pp. 55-62. Technology> Review, Howe, Wayne J. “The Business Services Industry7 Sets Pace in Employment Growth,” vol. 109, no. 4 (April 1986), pp. 29-36. Monthly Labor Review, Keil, Stanley R., and Richard S. Mack. “Identifying Export Potential in the Service Sector,” April 1986, pp. 2-10. Growth and Change, Kendrick, John W. “Outputs, Inputs, and Produc tivity in the Service Industries,” in Committee on National Statistics, Commission on Behavioral and Social Sciences and Education, National Research Council, Washington, D.C.: National Academy Press, 1986, pp. 60-89. About Service Industries, Statistics Kutscher, Ronald E., and Jerome A Mark. “The Service-Producing Sector: Some Common Per ceptions Reviewed,” vol. 106, no. 4 (April 1983), pp. 21-24. Monthly Labor Review, Kutscher, Ronald E., and Valerie A Personick. “Deindustrialization and the Shift to Services,” vol. 109, no. 6 (June 1986), pp. 3-13. Monthly Labor Review, Identifying Amenity and Productivity Cities Using Wage and Rent Differentials by Patricia E. Beeson and Randall W. Eberts Patricia E. Beeson is an assistant pro fessor of economics at the Univer sity of Pittsburgh and a visiting scholar at the Federal Reserve Bank of Cleveland. Randall W . Eberts is an assistant vice president and economist at the Federal Reserve Bank of Cleveland. Comments and suggestions by Tim Gronberg, Ron Krumm, and Jo e Stone and compu ter assistance by Ralph Day are gratefully acknowledged. 16 Introduction Many studies have explored the existence of nominal wage differentials between regions. The irrefutable conclusion is that wage differentials exist and that they persist over time.1 Such differ entials are difficult to explain within a neoclassi cal framework in which regions and factors are identical and all factors are free to move in response to interregional factor price differentials. In this case, one must resort to explanations based on institutional barriers and other impedi ments to free mobility. The key to understanding how wage differentials (and other factor price differen tials) can persist in the presence of free mobility is to recognize that some factors are inherently immo bile. For instance, each region has geographic and climatic characteristics that are unique to the area. Even for those areas that share common fea tures, the quality and quantity of the site-specific characteristics may differ. Therefore, firms or households will be willing to pay or accept dif ferent levels of wages depending upon the value they place on these attributes. For instance, firms may find that proximity to improved harbors reduces shipping costs and thus reduces production costs. In this case, firms can offer higher wages and still remain competitive with firms in lower-wage regions be cause of the cost advantage of the harbor. Since I Bellante (19 79 ), Johnson (1983), and Eberts and Stone (1986) are examples of numerous studies that have examined interregional wage differentials. land next to the harbor is limited, the influx of firms attracted by the harbor will increase the de mand for both labor and land. Wages and rents will be bid up until the cost advantage of the harbor is completely offset by the increase in fac tor prices. Thus, wages and land rents vary across regions according to the value firms place on the site-specific attributes in each region and their ability to substitute between factors of production. A similar story can be told about households. Households may value the same harbor that firms find attractive, except for differ ent reasons. The harbor that reduced shipping costs for firms may be attractive to households as a place to enjoy water sports. Consequently, as more households move into the area to take advantage of the harbor, the supply of labor increases and the demand for land increases. Thus, wages fall and land rents rise until individ uals are no longer willing to accept proximity to the harbor as compensation for lower wages and higher land rents. The resultant wage differential between an area with a harbor and one without depends upon the relative magnitudes of the demand and supply responses to site characteris tics. If wages are observed to be higher in the harbor area than in the area without a harbor, then the demand response (the firm’s response) dominates the wage determination process. If wages are relatively lower in the harbor area, then the supply response (the household’s response) dominates the process. In both cases, land rents will be higher because both households and firms value the harbor. Land rents would be lower in the harbor area than in an otherwise comparable area if the harbor was detrimental to both parties. Consequently, by observing relative wages and rents, it is possible to identify whether a region’s bundle of site characteristics has a greater net effect on firm location decisions or household location decisions. The purpose of this paper is to identify metropolitan areas according to the extent to which they are dominated by supply and demand responses to their net bundle of site-specific characteristics. To do this, we esti mate hedonic wage and rent equations for a sample of metropolitan areas. From these esti mates, we derive quality-adjusted wage and rent differentials for each area. The metropolitan areas are then classified into four groups based on the relative values of an area’s wage and rent differen tial vis-a-vis the national average. The metropoli tan areas are identified as high amenity (low wage, high rent), low amenity (high wage, low rent), high productivity (high wage, high rent), and low productivity (low wage, low rent). Classi fication of this sort provides information about the relative attractiveness to firms and households of the total bundle of attributes indigenous to each metropolitan area. I. A Model of Household and Firm Equilibrium In this section, we first present a model, based on the work of Roback (1982), of the effects of inter area differences in amenities and productivity on wages and rents. We then show how this model can be used to determine the relative importance of amenity and productivity differences as sources of factor price differentials across cities. Several simplifying assumptions are made in modeling the relationship between interarea differences in amenities and productiv ity and interarea differences in wages and rents. Workers are assumed to be identical in tastes and skills and completely mobile across cities. Sim ilarly, capital is assumed to be completely mobile and production technologies are assumed to be identical across firms.2 2 If people have different preferences, the value of certain areas will be understated in our approach, which uses a comparison of cost- of-living differences as an indication of the value individuals place on cit In this model, cities are character ized as bundles of attributes, which can affect the utility of households and the costs of production for firms. Individuals in these cities consume and produce a composite consumption good. The price of the good is determined by international markets and for convenience is normalized to one. Each worker supplies a single unit of labor independently of the wage rate. We assume that individuals work in the city in which they live, and we treat differences in leisure resulting from differences in intracity commuting as a site char acteristic.3 Equilibrium in this model is character ized by equal utility for identical workers and equal unit costs for firms across all regions. Workers choose the location that maximizes their utility, subject to an income con straint. Utility depends upon consumption of the composite commodity residential land (Zc), and amenities ( 5). Equivalently, the problem can be stated in terms of an indirect utility function, which is a function of wages ( rents ( r), and amenities Equilibrium for workers requires that utility is the same at all locations, or (1) The equilibrium relationship between wages, rents, and amenities for households can be determined by totally differentiating the indirect utility7function. In log form, this relationship can be stated as: (2) + 0. (X), V, (s). w), V(w,r;s)=V°. dlnV d ln w + dlnV dlnr dlnV = dlnw ds dlnr ds ds Using Roy’s identity, the marginal valuation of amenities in a city evaluated relative to the marginal utility of income is (3) - Ps = k xdlnr/ds dlnw/ds, W~ where Ps is the monetized value of the amenities, and kx is the portion of consumer income spent on land. Equation 3 states that individuals pay for amenities through reductions in real income in the form of higher land rents (which reduce income by times the increase in rents) and lower wage income. Firms are assumed to employ local residents and to use land to produce the compos ite commodity, according to a constantreturns-to-scale production technology. Under these assumptions, equilibrium for firms requires that unit costs are equal in all locations and equal to the price of assumed to be 1, (4) 1. kx X, X, C(w,r,s) = ies (see Roback [1982]). The second set of assumptions refers to the mobility of households and firms. W e assume that migration is costless and that, given the relative wages, rents, and site characteristics across cities, both firms and households have chosen locations such that they could not be made better off by relocating. If moving is not costless, we m ay have biased estimates of the attractiveness of areas. Individuals or 3 Roback's model ignores intracity commuting. Hoehn, et al. (1986) have pointed out that this leads to incorrect estimates of the value of other site characteristics. Since w e are not interested in deriv firms m ay perceive that they would be better off by moving, but if it is ing values for specific characteristics, but simply the net impact of these costly to do so they will move only if the extra benefits of moving out characteristics, our model is not subject to this criticism. W e therefore weigh the costs of moving. W e may then be over- or underestimating simply assume that intracity commuting is another site characteristic the attractiveness of an area since w e ignore the costs of moving. that reduces leisure time and therefore is a disamenity for workers. 17 The relationship between wages, rents, and site characteristics (s), which are con sistent with equilibrium for firms, can be expressed in log form as: (5) + + = 0. 3 The marginal value to firms of different locations is (6) ) where - is the price that firms are willing to pay to locate in one city rather than another, and and are the cost shares of land and labor, respectively. If the site characteristics of a city provide a net productivity advantage to firms, then firms will pay for this advantage in terms of higher wages and rents and will be positive. Wages and rents in each city are determined by the interaction of the location decisions of the households and firms. dlnC dlnw dlnC dlnr Inw ds dlnr ds dlnC ds Cs - - 6r(dlnr/ds -6u,(dlnw/ds), Cs Qr dw -Cs Inw Inr Combinations of and for which the unit costs of firms are equal are depict ed in figure lb. The value of site characteristics to firms is fixed along each quasi-isocost curve, and the curves shift up (down) as the site characteris tics of a city increase (decrease) the productivity of firms. The slope of the quasi-isocost curve is equal to the elasticity of substitution between land and labor, which from equation 6 is According to figure lb, site characteristics in city enhance productivity more than site character istics in city S ,. Each city is characterized by a bundle of amenities and site characteristics that are associated with a specific pair of isocost and iso-utility curves in figures la and lb. The inter section of any two curves for each city then determines relative wages and rents. In figure 2, equilibrium wages and rents in city Sj will be and ]. Using city Sj as a reference point - 0w/6 r. S2 wl r Equilibrium Conditions for Households and Firms 1A 1B SOURCE: Authors. FIGURE 1 Classification of Cities as Amenity or Productivity Cities The model described above is illustrated in figure 1. The upward sloping curves in figure la [labeled V(.)], show combinations of and for which utility is equal. The slope of these curves is the trade-off that households are willing to make between wages and rents for any given level of amenities. From equation 3, this trade-off is equal to the inverse of the budget share of land, [ Along each curve, the value of amenities is fixed and the curves shift up (down) as the amenities of one city are valued more (less) than the amen ities of other cities. The value of amenities in the city labeled is greater than the value of ameni ties in the city labeled since individuals are willing to pay higher rents at every wage rate. Inw Inr k \ S2 (which could be thought of as the average city), we can see how intercity differences in amenities and productivity will be reflected in differences in wages and rents. Consider a city that differs from 5, only in that the site characteristics of city provide a greater productivity advantage to firms than the site characteristics of city S,. In figure 2, this is illustrated by 2) lying above C(5'1). Assuming there is no difference in amenities between the two cities, we can see that equilibrium requires that wages and rents in city be high relative to city . These higher wages and rents re flect the amount firms are willing to pay to locate in city rather than Sj and, therefore, the pro ductivity value of relative to the average city. S2 S2 C(S S2 S2 S2 Productivity Differences and Equilibrium Wages and Rents SOURCE: Authors. FI GURE in amenities or productivity by examining the pattern of wages and rents across cities. If wage and rent differences primarily reflect amenity dif ferences across cities, we would see a negative relationship between wages and rents. If they reflect productivity differences, the relationship would be positive. Within the same framework, we can also classify individual cities on the basis of whether their wages and rents differ from the average because of above-average amenities, below-average amenities, above-average produc tivity, or below-average productivity. These classi fications are summarized in table 1 and figure 4. O f course, cities may differ in characteristics that affect both household utility and production costs. The problem of classifying cities by the relative magnitudes of these two effects becomes one of identifying the portion of the wage and rent differentials due to a shift in each curve. This can be done by identifying the combinations of and that would result from equal shifts of both curves and determining how wages and rents in each city fall relative to these shifts. The combinations of and that would result from equal shifts of both curves will form two lines with slopes that depend upon 1 and If j-1 (the slope of the curve) is equal to (the negative of the slope of the curve), the com binations of and resulting from equal shifts of both curves would coincide with the x and y axis. Assuming for illustration that this is the case, for any city with above-average wages and rents, the shift of the (productivity) curve must be greater than the shift of the (amenity) curve. Therefore, any city with wage and rent com binations in quadrant A in figure 4 is classified as a “high productivity” city, because the reason that this city’s wages and rents differ from those of the average city is the above-average productivity it affords firms. This above-average productivity is reflected in the ability of firms in these cities to pay above-average wages and rents. Similarly, cities with below-average wages and rents (quadrant C in figure 4) are clas sified as “low productivity” cities, since firms in these cities are compensated for the belowaverage productivity related to site characteristics with below-average factor costs. Above-average amenities in a city are associated with increases in rents and decreases in wages reflecting households’ will ingness to pay for the amenities. Quadrant D then identifies cities where the dominant factor determining relative wages and rents is high amenities. For cities in quadrant B, the dominant factor is their below-average amenity value. Inw 2 Consider another city, S3, that differs from only in that households find it to be more amenable. This relationship is illustrated in figure 3, where city is represented by which is to the left of Sj). If no productivity differences exist, [that is, 6’(5 1)= C(S3)], the dif ference in the households’ valuation of amenities across cities leads to lower wages and higher rents in the more amenable city, S Sx Inr Inr V(S5), Inw S} V( 3. Amenity Differences and Equilibrium Wages and Rents k x Vs Inw - 0w/0r. k 6u,/6r Cs Inr Cs Vs primary> SOURCE: Authors. Within this simple framework in which cities differ in either amenities or produc tivity, but not both, we can determine whether factor price differences reflect intercity differences 1 9 Classification of Cities rent D: High Amenity A: High Productivity These labels may be misleading in that what we are referring to as “high productiv ity” cities are not necessarily more or less attrac tive to households than the “high amenity” cities. A city like the one represented by point A in fig ure 5 is relatively attractive to both households and firms. This relationship can be seen by the positions of and relative to the average city. The effect that dominates, however, is the productivity effect, since the shift of the curve is greater than the shift of the curve. Another city like the one repre sented by point B may be less attractive to both firms and households than city A (again reflected in the relative positions of the amenity and pro ductivity curves). However, the dominant trait of city B is its amenities, which are above average. C(SA) Cs C. Low Productivity I B: Low Amenity w II. Estimation 4 Classification of Cities D irection o f Price Differential Rent Shift City Classification Wage High productivity High High Low productivity Low Low High amenity Low High Low amenity High Low C ( 5;) curve up C(S,) curve down ViS,) curve up ViS,) curve down SOURCE: Authors. TABLE Vs wage SOURCE: Authors. FI GURE V(SA) 1 Classification of Cities and the Relative Productivity and Amenity Effects The analysis is based on wage and rent data for a sample of recent movers drawn from the Public Use Microdata Sample of the 1980 Census of Popu lations. This subsample includes individuals who lived and worked in the same Standard Metropoli tan Statistical Area (SMSA) in 1980 and who changed addresses between 1975 and 1980. This subsample of movers was chosen because housing prices of recently acquired or rented dwellings more accurately reflect current land market conditions. The rent equation includes both owner-occupied and rental units for which posi tive values of unit or gross rent are reported. The dependent variable in the rent equation is gross monthly housing expenditures. For homeowners, the monthly housing expenditure is based on the value of the dwelling using 7.85 percent as the discount rate.4 The monthly housing expenditure is the sum of this imputed rent and monthly util ity charges. For renters, the monthly expenditure is gross rent (contract rent plus utilities). Individuals included in the wage sample had to meet the following criteria. Indi viduals had to be between the ages of 25 and 55; work more than 25 hours per week; not be selfemployed; and have positive wage and salary income. The dependent variable in the wage equation is average weekly earnings, which is calculated by dividing annual wage and salary income by the number of weeks worked. SOURCE: Authors. 4 The discount rate is from a study of the user cost of capital by Peiser and Smith (1985). Quality-Adjusted Wages A hedonic approach is used to estimate wage dif ferentials across SMSAs. This approach uses regression analysis to determine the value the market places on different worker characteristics. An individual’s wage is then predicted based on the value of his or her characteristics. The first step in constructing the wage indexes is to spec ify estimable equations that reflect appropriate individual characteristics of workers that could Estimates of Wage Equation Variables Mean Coefficient Intercept — 4.33 (50.19) -.083 (-5.00) -.161 (-11.57) .043 (5.16) .0007 (2.81) .043 (25.12) -.0008 Sex (Female =1 ) .42 Race (Black =1) .16 Education Education squared Experience Experience squared Part time Usual hours worked per week 15.55 250.37 10.29 192.33 .04 42.05 Head of household .64 Veteran .20 Sex x Race .08 Sex x (Marital status) .22 Sex x Experience Sex x (Experience squared) 4.10 76.82 Marital status .62 Union member .25 (-15.63) -.308 (-14.44) .006 (10.84) .111 (10.20) -.017 (1.53) .111 (5.47) -.058 (3.14) -.019 (-7.81) .0003 (3-54) .108 (9.62) .434 (14.12) affect wages. Our approach follows the human capital specification of individual wages set forth by Hanoch (1967) and Mincer (1974). Thus, we specify individual wages (expressed in loga rithms) as a function of education level (entered as a quadratic), potential experience (age minus years of education minus six, also entered as a quadratic), a binary variable indicating part-time employment status (less than 35 hours per week), and 42 binary occupation variables (with one omitted as a constant). Binary variables are also entered to account for gender, race, marital status, union affiliation, and whether or not an individual is a veteran.5 In addition, the gender variable is interacted with other characteristics in order to control for male/female differences in the rate of return to these attributes. The estimated coefficients of the wage equation are presented in table 2, except for the occupation variables, which are omitted for brevity. The estimated coefficients are as expected. Education and experience are valued positively in the labor market, while part-time, female, and nonwhite workers receive lower wages than their otherwise identical counterparts. We also find that individuals who are married, heads of households, and in highly unionized industries earn more than their counterparts. Females receive less return on experience than males. The predicted wage level for each worker in the sample is obtained by multiplying the estimated coefficients by each worker’s char acteristics. The predicted wage can be interpreted as the compensation a worker could expect to receive, given his or her characteristics, regardless of geographic location. Subtracting the predicted wage from the actual wage nets out the portion of the actual wage that is related to the individual worker’s characteristics. The skill-adjusted metro politan wage differentials are then obtained by averaging the wage residuals (actual minus pre dicted wage) for all workers in a particular metro politan area. Average wage differentials are calcu lated for each of 38 cities. The 38 metropolitan areas are chosen by including only those SMSAs for which 100 or more individuals in the sample were recorded as movers between 1975 and 1980. The quality-adjusted wage differentials are dis played in table 4. (42 Occupation Dummies) R-square No. observations Dependent variable: log (weekly earnings) .34 22,313 5.50 Note: Estimates derived from Public Use Microdata Sample. T-statistics in Rent Equation The method used to calculate quality-adjusted rent differentials is similar to the one used to cal culate quality-adjusted wage differentials. The log of monthly housing expenditures is regressed parentheses. SOURCE. Authors. The measure of unionization in the wage equation is the industry unionization rate taken from Kokkelenberg and Sockell (1985). 21 Estimates of Rent Equation Variables Intercept Dwelling rented (=1) Central city (=1) x rental Mean Coefficient .53 .14 9.93 (248.36) .084 (1.35) -.05 (-3.29) .021 (1.70) Number of floors x rental 1.10 .122 (5.43) -.056 (-2.62) Attached dwelling (=1) x rental .06 .06 (2.41) .027 (1.17) Year dwelling built x rental 3-65 -.06 (-17.98) -.018 (-4.94) Number of rooms x rental 7.07 .11 (22.80) -.032 (-5.64) Number of bedrooms x rental 4.25 .10 (9.96) .011 (1.03) .14 .06 (3.70) -.027 (-.83) .52 .12 (9.13) .038 (2.82) .91 .12 (6.35) -.058 (-4.14) .96 -.046 (1.62) 2.92 -.003 (-.65) .007 (1.41) 2.72 .179 (32.03) -.056 (-6.73) Well water (=1) x rental Central air conditioning (=1) x rental Central heating (=1) x rental Dwelling other than condominium (=1) Number of units at address x rental Number of bathrooms x rental against housing attributes. These characteristics include the number of rooms, number of bed rooms, number of bathrooms, and separate binary variables indicating location of the dwelling in the central city, and whether or not the dwelling is a single structure, has central air conditioning and/or heating, is connected to a city sewer sys tem, and has well water. The year the dwelling was built is entered to proxy the vintage. Dwell ing characteristics are interacted with rental status in order to account for differences in the valua tion of these attributes between rented and owner-occupied dwellings. Coefficient estimates are reported in table 3- The results are as expected. Larger, newer dwellings with central air and heating and that are located outside the central city have higher market value than otherwise identical homes. In general, attributes of rentals are valued less than otherwise identical owner-occupied dwellings. The predicted rent is calculated by multiplying the estimated coefficients by the housing characteristics of each household. The quality-adjusted rent differentials presented in table 4 are the differences between the actual and predicted house values. By including a number of housing characteristics in the rent equation, the difference between actual and predicted house values can be interpreted to reflect primarily land values in specific geographical locations. Thus, qualityadjusted rent differentials relative to the national average reflect differences in city land values, which are due primarily to the capitalized effects of differences in site characteristics. Land Shares City sewer connection (=1) x rental .87 Lot size less than one acre (=1) x rental .92 -.130 (8.72) .185 (8.07) Elevator (=1) .04 .065 (2.45) R-square .053 (4.27) .004 (.18) .63 No. observations 16,017 Dependent variable: log (house value) 11.07 Note: Estimates derived from Public Use Microdata Sample. T-statistics in parentheses. The entry “ x rental” indicates that the rental dum m y variable has been interacted with the variable listed immediately above it. SOURCE: Authors. In addition to the quality-adjusted wage and rent differentials, our classification of cities requires estimates of the share of household income spent on land }) and the ratio of the income shares of land and labor in production (0r/0 M,). These values are not readily available for each specific metropolitan area. Thus, we use national estimates and assume that the portion of house hold income spent on land and the ratio of labor income to land income in production are con stant across metropolitan areas and equal to the national average. The budget share of land is calcu lated by multiplying the fraction of income spent on housing (27.0 percent in our sample) by the ratio of land value to the total value of the house (estimated to be 19 6 percent).6 From these esti mates, land’s share of household income ( x) is (k k 6 The ratio of land value to total house value w as estimated by Roback using F H A housing data. Unfortunately, the census data used in this study cannot be used to make a new estimate. (dr/d w) kx 5.3 percent. The ratio is calculated by sub tracting our estimate of from the ratio of the to tal income to land (6.4 percent of national income) relative to total labor income (73 percent of national income).7 The ratio of these income shares is 8.8 and the estimate of is 3-5. SMSAs lie in the “productivity” quadrants than in the “amenity” quadrants. This is confirmed by a positive correlation coefficient of 0.46. The rela tively small value of the coefficient suggests that the relationship is not the same across all SMSAs. We now proceed to determine whether deviations from the average wages and rents for individual SMSAs primarily reflect a) above-average amenities, b) below-average amen ities, c) above-average productivity, or d) belowaverage productivity. In order to determine the combi nations of wages and rents that fall into each of these categories, we must first determine the wage and rent combinations that form the boun daries for these categories. These boundaries are determined by the combinations of wages and rents that would result from equal shifts of the and curves relative to the average SMSA Using the estimates of land shares discussed above, we find that for all practical purposes these boundaries coincide with the x and y axis in figure 6.8 A listing of cities in each category is presented in table 5. Most of the SMSAs fall 0r/6 u, III. Classification of Cities As discussed in section I, we can determine whether wage and rent differentials reflect varia tions in productivity or amenities across SMSAs by examining the pattern of wage and rent differen tials across SMSAs. If intercity wage and rent dif ferentials primarily reflect amenity7differences, we should observe a negative relationship between wages and rents. If they primarily reflect produc tivity differences, the relationship should be positive. The quality-adjusted wages and rents for the SMSAs in our sample are presented in figure 6. It appears from figure 6 that there is a slight positive relationship between wages and rents in our sample. Using the same amenity and productivity quadrants found in figure 4, more Vs Cs 2 3 Standard Metropolitan Statistical Areas Included in Sample 0.3075 • rent differential . Los Angeles, CA • •I Boston, MA * • Newark, NJ San Diego, CA # San Francisco, CA • Anaheim, CA ^an Jo s e ’ CA * Nassau, NY New York, NY Washington, D.C. • • Miami, FL Milwaukee, WI • • Portland, OR • Fort Lauderdale, FL • Denver, CO • • • • Sacramento, CA • Phoenix, AZ Minneapolis, MN Seattle, WA Houston, TX • Riverside, CA ------- T " 0.0129 ,• Chicago, IL Philadelphia, PA ----------------------------- •-Detroit, M I Cleveland, OH • • Tampa, FL * • Dallas, TX Columbus, OH • -0.2029 -0.1194 • Salt Lake City, UT • New Orleans, LA Baltimore, MD * • Pittsburgh, PA St. Louis, MO * Cincinnati, OH Kansas City, MO • Atlanta, GA Indianapolis, IN San Antonio, TX 0.0000 wage differential SOURCE: Authors. FI GURE 0.1493 6 7 The estimate of labor compensation is taken from the national income account data reported in Table B-23 of the Economic Report of the President (1987). Unfortunately, the national income accounts do not include land income as a separate category of income. 8 The exact boundaries are tw o lines that pass through the origin, one with a slope of -.003, the other with slope 333. W e classified Our estimate of land's share of income is taken from Mills and Hamilton cities based on these boundaries, but the classifications do not change if (1984). one uses the x and y axis as reference points. Quality-Adjusted Rent and Wage Differentials3 M etropolitan Area________________ Anaheim, CA Atlanta, GA Baltimore, MD Boston, MA Chicago, IL Cincinnati, OH Cleveland, OH Columbus, OH Dallas, TX Denver, CO Detroit, MI Ft. Lauderdale, FL Houston, TX Indianapolis, IN Kansas City, MO Los Angeles, CA Miami, FL Milwaukee, WI Minneapolis, MN Nassau-Suffolk, NY New Orleans, LA New York, NY Newark, NJ Philadelphia, PA Phoenix, AZ Pittsburgh, PA Portland, OR Riverside-San Bernardino, CA Sacramento, CA St. Louis, MO Salt Lake City, UT San Antonio, TX San Diego, CA San Francisco, CA San Jose, CA Seattle, WA Tampa, FL Washington, D.C. Rent Quality-Adjusted ________ W age .281 -.145 -.075 .220 .104 -.082 -.053 -.126 -.103 .036 .013 .039 .023 -.172 -.155 .261 .076 .100 .073 .240 -.110 .145 .195 -.013 -.029 -.079 .059 .016 -.014 .085 -.099 -.203 .148 .308 .269 .048 -.142 .116 .078 .014 .031 -.001 .081 .064 .108 -.074 .001 -.013 .149 -.029 .142 .041 -.037 .049 -.112 -.002 .065 .077 -.079 .036 .045 .017 -.047 .047 -.027 -.008 -.047 .019 -.081 -.105 -.014 .073 .125 .047 -.119 .103 a. Quality-adjusted differentials are obtained by subtracting the predicted cities receive compensation for this low amenity value in the form of above-average wages and below-average rents. SMSAs that can be characterized as “high productivity” include Chicago, Houston, Los Angeles, and San Jose, among others. For these cities, both wages and rents are above aver age, suggesting that the firms in these cities are compensated for high factor costs by other loca tional characteristics of these cities. SMSAs like Tampa, New Orleans, and San Antonio can be characterized as “low productivity.” Firms in these areas are compensated for the belowaverage productivity value of site characteristics in the form of lower wages and rents. Classifying SMSAs according to the dominant effect of their site characteristics does not mean that a high-productivity city has no amen ity value. It simply means the city is dominated by its productivity characteristics. Using equations 3 and 6, we can develop relative rankings of cities within the productivity groups by amenities and within the amenity groups by productivity. The ordering of cities in table 5 reflects this sort of cross classification. For example, of the highproductivity cities, New York, Los Angeles, and Seattle are considered more amenable than Chi cago, Houston, and Detroit. O f the high-amenity cities, Boston is more attractive to firms than Miami. The classifications of some cities are questionable, especially for cities near the boundaries. For some cities like Boston and Mil waukee, rents are considerably higher than aver age, but wages are so close as to be indistinguish able from the average. As a result, we cannot be confident in our classification of these cities as high productivity or high amenity, although we can be fairly confident that they are not lowamenity or low-productivity cities. Philadelphia and Riverside are examples of cities that are so close to the average in both wages and rents that their classifications may also be meaningless. estimate from the actual value. The reference point for these estimates is the sample average. IV. Conclusion SOURCE: Authors. TABLE 4 within expected classifications. For instance, Miami, Denver, Portland, Ft. Lauderdale, and San Diego are classified as high-amenity cities, since these cities are characterized by below-average wages but above-average rents, both of which reduce the income of households. In cities like Baltimore, Cleveland, Pittsburgh, and Atlanta, wages and rents primarily reflect the below-average amenity value to households of these cities. Households in these In this paper, we have utilized the relationship between regional wage and rent differentials to identify cities by the net effect of their bundle of site characteristics on firms and households. We have found that, on average, firms respond more to site characteristics than households, as is revealed in the relatively large contribution of demand effects to determining regional wage dif ferentials. Nevertheless, the amenity (or house hold) component of the total regional differential is also significant. Thus, regional wage differen tials result from the interplay of the forces of supply and demand and exist even though indi viduals move freely in response to factor price References Classification of Cities High Productivity L ow Productivity New York, NY Newark, NJ Los Angeles, CA Seattle, WA San Francisco, CA Minneapolis, MN Anaheim, CA Nassau-Suffolk, NY Chicago, IL Washington, D.C. San Jose, CA Houston, TX Detroit, MI Tampa, FL San Antonio, TX Salt Lake City, UT New Orleans, LA Columbus, OH Sacramento, CA Phoenix, AZ Kansas City, MO High Amenity Low Amenity Boston, MA San Diego, CA Milwaukee, WI Denver, CO Riverside, CA Portland, OR Ft. Lauderdale, FL Miami, FL Cleveland, OH Cincinnati, OH Pittsburgh, PA Philadelphia, PA Baltimore, MD St. Louis, MO Indianapolis, IN Dallas, TX Atlanta, GA Amer ican Economic Rei’iew, Eberts, Randall W., and Joe A. Stone. “Metropoli tan Wage Differentials: Can Cleveland Still Compete?” Federal Reserve Bank of Cleveland, Quarter 2 (1986), pp. 2-8. Economic Review, Hanoch, Giora. “An Economic Analysis of Earn ings and Schooling,” vol. 2, no. 3 (Summer 1967), pp. 310-29. Journal of Human Resources, NOTE: Productivity cities are listed from the most amenable to the least. Amenity cities are listed from the most productive to the least. SOURCE: Authors. TABLE Bellante, Don. “The North-South Differential and the Migration of Heterogeneous Labor,” vol. 69, no. 1 (March 1979), pp. 166-75. Hoehn, John P., Mark C. Berger, and Glenn C. Bloomquist. “A Hedonic Model of Interre gional Wages, Rents and Amenity Values,” University of Kentucky Working Paper No. E-91-86 (1986). Johnson, George E. “Intermetropolitan Wage Dif ferentials in the United States,” in Jack E. Tri plett, ed., Chicago: University of Chicago Press, 1983, pp. 309-30. The Measurement of Labor Cost. Kokkelenberg, Edward C., and Donna R. Sockell. “Union Membership in the United States: 1973-1981,” vol. 38, no. 4 (July 1985), pp. 497-543. Industrial and Labor Relations Revieu\ 5 differentials. Thus, so long as regions differ in the amount and quality of their site-specific character istics, wage differentials will continue to exist. Urban Mills, Edwin S., and Bruce W. Hamilton. Glenview, 111.: Scott Foresman and Company, 3rd ed. (1984). Economics. Schooling Experience, and Earn Mincer, Jacob. New York: National Bureau of Economic Research, Distributed by Columbia University Press, 1974. ings. Peiser, Richard B., and Lawrence B. Smith. “Homeownership Returns, Tenure Choice and Inflation,” vol. 13, no. 4 (Winter 1985), pp. 343-360. American Real Estate and Urban Economics Association Journal, Roback, Jennifer. “Wages, Rents and the Quality of Life,” vol. 90, no. 6 (December 1982), pp. 1257-78. Journal of Political Economy, 25 FSLIC Forbearances to Stockholders and the Value of Savings and Loan Shares by James B. Thomson Jam es B. Thomson is an economist at the Federal Reserve Bank of Cleveland. The author thanks Lynn Downey for excellent research assis tance, and Randall Eberts, Christopher Jam es, William Osterberg, and Walker Todd for helpful comments and suggestions. 26 Introduction Policies of forbearance to stockholders of insol vent firms by federal deposit guarantors represent a wealth transfer from federal deposit-insurance agencies, and ultimately from federal taxpayers, to the stockholders of the insured institutions. Kane (1985, 1986), Pyle (1986), and Thomson (1987) discuss theoretical determinants of the value of forbearances to stockholders of financial institutions by the Federal Deposit Insurance Corporation (FDIC) and the Federal Savings and Loan Insurance Corporation (FSLIC). Brickley and James (1986) show empirically that the stockmarket returns of thrifts increase with the exten sion of FSLIC capital forbearances. This paper investigates the rela tionship between the market and book values of the firm’s equity. It demonstrates that the market value of a thrift is positively related to its book value and to the value of its unbooked assets. We argue that one of the major unbooked assets of a thrift is its FSLIC insurance guarantee. Measures of FSLIC forbearance policy are shown to be related to the market value of the thrifts whose market values exceed their book values. Section I of this paper discusses the relationship between the market value and book value of a firm. It outlines the reasons that these values may diverge and argues that FSLIC guarantees are one of the unbooked assets valued by the market. Section II gives a brief overview of the empirical evidence and theoretical arguments regarding the value of federal deposit guarantees and forbearances. Section III describes the data, the sample selection criteria, the regression experiment used to test the forbearance hypothe sis, and the empirical results. The conclusions and policy implications of the paper are pre sented in section IV. I. The Relationship Between Market and Book Values The book value of a firm’s equity is measured as the difference between the book value of the firm’s assets-in-place and the par value of its lia bilities. The book value of assets may not equal their market value for three reasons. First, the accounting conventions used by most firms carry assets at their par, or acquisition, value and do not reflect subsequent changes in the market value of the assets. The market value of the assets would include these unbooked gains and losses. Sec ond, because book values tend to include only assets-in-place, they do not measure the value of options for future business that are unique to the firm.1 Finally, to avoid taxes, burdensome regula tions, or restrictive debt covenants, some firms may engage in activities that are not carried on their books. The assets (liabilities) associated with these activities would not show up in book measures of assets (liabilities), but would none theless be reflected in their market values. 1 Myers (19 77) and Warner (19 77) argue that the market value of the firm's assets includes both the market value of the assets-in- place and the market value of the firm’s options for profitable future business opportunities. Therefore, if the firm carried its assets-in-place at market value, the book value of the firm would understate its market value. On the other side of the ledger, the firm carries its liabilities at par. Like the assets, the liabilities’ market value includes unbooked changes. The market value of the firm’s liabilities also includes off-balance-sheet financing and other types of contingent liabilities not reflected in book values (see Bennett [1986] and Forde [1987]). Therefore, the book value of the firm’s equity will differ from its market value if the errors in the book measures of the firm’s assets and lia bilities do not completely offset one another. Unbooked Losses and Gains in Thrift Portfolios The market value of a thrift institution’s assets can be separated into the market value of its assets-inplace and the market value of its charter. The mar ket value of the assets-in-place may not equal their book value because the accounting procedures that thrifts and their regulators use to calculate book values do not take into account unrealized gains and losses on the thrift’s asset portfolio. For example, thrifts hold a large vol ume of fixed-rate mortgages, whose market values fluctuate inversely with interest rates. When inter est rates rise, the market values of the mortgages decrease while the face value of the mortgage portfolio remains constant. Because thrifts are not forced to recognize capital losses on the mort gages until they are sold (or until the customer defaults), an increase in interest rates causes the book value of the mortgage portfolio to exceed its market value and the market value of the assets-in-place to be less than their book value. Another source of unbooked capi tal gains and losses in the thrift’s portfolio are real estate holdings. Thrifts tend to carry real estate on their books at acquisition price, which may not equal the current value of the real estate. The real estate portfolios of many thrifts are likely to be carried on their books at a discount from market value, which may cause the book value of the thrifts to be less than their market value. The Value of Thrift Charters The charter value of a thrift reflects the value of its unbooked assets.2 We can divide the value of the thrift’s charter into five categories. The first is the value of business relationships built over time. Kane and Malkiel (1965) argue that long standing customer banking relationships have 2 Buser, Chen, and Kane (1981) maintain that the FD IC attempts to preserve the value of the banking charter when disposing of a failed bank, by using the charter's value to reduce the disposal costs. If the bank is disposed of via a purchase-and-assumption transaction, the pur chase premium paid by the bank acquiring the failed bank reflects the value of the charter to the acquiring institution. value because they lower the information and contracting costs associated with doing business. The reduction in the cost of servicing long standing customers is available only to the servic ing thrift and is a source of profitable future busi ness opportunities. Firm-specific options for profitable future business opportunities are the second source of the charter’s value. These options may be available to the thrift because it has developed expertise in servicing a particular segment of the market. The third source is monopoly rents that may accrue to the thrift from restrictive branching laws and other regulations that restrict competition. The fourth source of the charter’s value is access to Federal Home Loan Bank Board (FHLBB) advances. The FHLBB makes secured loans to member thrifts at subsidized rates. These advances represent both a direct subsidy and an in expensive source of backup liquidity. The fifth com ponent of the charter’s value is federal deposit guarantees. Kane (1985, 1986) maintains that the mispricing of deposit insurance and the use of forbearance policy by federal deposit guarantors has made the value of deposit guarantees an important source of thrift charter values. II. FSLIC Subsidies, Forbearances, and the Market Value of Thrift Institutions A new and growing body of literature addresses the value of federal deposit insurance subsidies and forbearances to insured depository institu tions. Kane (1985, 1986) argues that the aggre gate net worth of the thrift industry, net of the value of deposit guarantees and forbearances, is negative. Pyle (1986) shows that the the use of capital forbearances increases the value of deposit guarantees. Brickley and James (1986) empirically demonstrate a positive relationship between the adoption of a capital forbearance policy by the FHLBB and the market value of thrift institutions. Ronn and Verma (1986) show that estimates of the fair value of deposit guarantees are extremely sensitive to assumptions regarding the forbear ance policy the FDIC employs when disposing of failed banks. Thomson (1987) breaks down the value of the deposit guarantee into three compo nents: the value of the guarantee on insured deposits, the value of a conditional guarantee on the uninsured deposits, and the value of a condi tional guarantee of the stockholders’ claim on the residual future earnings of the insured institution. This paper is concerned with the value of forbear ances to the stockholders of insured institutions. The federal deposit insurance agen cies extend forbearances to stockholders of insol vent institutions in two ways. The first, and politi cally preferred, method is to allow the institutions to operate after they are discovered to be insol- 27 28 vent.3 The de jure failure of a federally insured bank or thrift is an event timed by the regulators. The extension of explicit or implicit guarantees to the claims of uninsured depositors and general creditors of the insolvent bank or thrift removes the incentives of these individuals to force the closing and reorganization of the institution.4 A forbearance policy that does not at least close out the position of stockholders in insured depository institutions that are found to be insolvent has value to the stockholders (see Thomson [1987]). It represents an option on the future residual earnings of the institution. The behavior of the stock of Beverly Hills Savings and Loan (BHSL) of California is evidence that this type of forbearance has value. At the end of March 1985, roughly one month before it was closed by the FHLBB, the stock of BHSL had a market value of $19.21 million, while the book value of its equity was -$58,091 million.5 The second way stockholders re ceive forbearances from the federal deposit insurer is when the federal deposit guarantor uses open-bank assistance to handle the failure (or to head off the imminent failure) of an insured institution.6 In this case, the federal de posit guarantor may preserve some or all of the value of the stockholders’ claim on the residual future income of the institution. For example, when the FDIC bailed out the Continental Illinois Bank and Trust Company of Chicago (Continental) in 1984, it gave the original stockholders warrants allowing them to purchase shares in the reorganized insti tution. The estimated value of these warrants was approximately $155 million (close to 20 percent of the estimated equity value of the reorganized Continental) on the day after the bailout package was announced. 3 Net worth certificates and capital forbearances are two of the tools that politicians and industry regulators use to forestall the closing of insolvent institutions (see Nash [19 8 7] and McTague [19 87]). 4 a guarantee of the market value of their claim at the time the 5 ment program on April 2 5 ,19 8 5 . A t that time, the book value of The deposit guarantor must provide the uninsured depositors with institution is discovered to be insolvent on a market-value basis. The B H S L was admitted to the F H L B B ’s management consign its assets w as $2,939 billion, and its June 6 ,1 9 8 6 , the reported TNW TNW Empirical Issues The Data The sample consists of 43 thrifts that meet the fol lowing criteria. First, to measure the market value of equity, we had to be able to obtain stock price and share data on the thrifts from Data Resources Incorporated’s (DRI) Security Price File from March 1984 to the end of June 1986. Second, the thrifts had to be insured by the FSLIC. Third, balance-sheet and income-statement data had to be available from the FHLBB’s Quarterly Reports of Condition and Income. Finally, to remove the effects of nonthrift subsidiaries from the results, we excluded all thrift holding companies. The requirement that the thrifts’ stock must trade on the market restricts the sam ple to the largest firms in the industry. For exam ple, at the end of June 1986, the average size (measured in total assets) of the thrifts in our sample was $1,895 billion.7 This is considerably larger than the size of the average thrift in the population. Therefore, one should be careful in generalizing the results of the tests on this sam ple to the population. We do not expect the other sample selection restrictions to materially affect the results.8 To construct proxy variables for our tests, we draw on theoretical arguments (see Beaver, et al. [1970], Bowman [1979], Myers [1977], and Unal and Kane [1987]); empirical findings (see Barth, et al. [1985], Benston [1986], w as -$58,091 million. On ment program occurred when interest rates were falling. The one-year secondary market Treasury bill rate w as 8.22 percent on April 2 6 ,19 8 5 , 6, III. of B H S L w as -$540 million. In fact, the decline in B H S L 's net worth under the F H L B B ’s management consign and 6 .14 percent on June The probability that federal deposit guarantors will extend forbearances to stock holders of insolvent insitutions is a function of constraints on the guarantors’ ability to reorganize insolvent institutions. Kane (1986) places these constraints into four categories: political and legal constraints, information constraints, staff con straints, and funding constraints reflected in the implicit and explicit reserves of the insurance fund. Sprague’s (1986) account of the FDIC’s decision to bail out Continental makes it clear that the first three constraints played a major role in that bailout. Barth, et al. (1985) show that the ability of the FSLIC to close insolvent thrift institu tions is directly related to the solvency of the FSLIC insurance fund. 7 The largest (smallest) thrift in the sample at the end of June 1986, measured in terms of total assets, w as $10,551 billion ($164,226 million). 1986. Thus, it is fairly clear that the posi tive market value of B H S L before its closing was not due to unrealized capital gains on B H S L ’s portfolio. 8 To test the sensitivity of the results to survival bias, w e replicate the cross-section regression experiments using a sample that includes all firms in the sample with complete information for that quar 6 On December 4 ,1 9 8 6 , the FD IC announced that it had set up ter. Because the number of firms varies across quarters, w e do not formal guidelines for the use of open-bank assistance in handling attempt to pool this sample. Overall, the results over the larger sample troubled and failed banks (see McTague [1986]). support the paper’s main results. Results from the SMVAM Regressions3 (U sing GAAP Net W orthb ) Quarter 1984 1985 1986 N um ber MKTVAL0 TN W d 1 43 39506.73 60361.41 2 43 37452.41 3 43 4 k Ue R2 14006.781 (4.288)e 0.4224 l n (16.243) 0.7749 62386.19 12627.56+ (3-317) 0.39792++ (14.936) 0.7039 39170.32 63704.42 10493-50+ (2.695) 0.45015++ (13.553) 0.7502 43 40985.42 65153.79 6560.23 (1.482) 0.52837+ + (10.479) 0.7707 1 43 47330.67 67189.88 5920.71 (1.332) 0.61631 ^ (8.714) 0.8269 2 43 55950.50 71629.05 5789.46 (1.092) 0.70029n (6.107) 0.8324 3 43 51973.94 74923.77 10687.99* (2.044) 0.55104n (9.660) 0.7742 4 43 62388.50 77475.47 8304.28 (1.232) 0.69808++ 0.7772 (5.173) 1 43 79638.41 84108.60 10490.68 (1.354) 0.82212++ (2.937) 0.8180 2 43 83701.49 85911.07 21001.71** (1.968) 0.72982++ 0.6939 (3.569) 2 9 a. M odel: MKTVAL = Ug + kTN W + e. f Significantly different from zero at 1%. b. Net worth com puted using generally accepted accounting procedures. f t Significantly different from on e at 196. c. Average market value o f thrift stock (0 0 0 ’s). * Significantly different from zero at 5%. d. Average b o o k value o f thrift equity (0 00’s). ** Significantly different from zero at 10%. e. T-statistics in parentheses. SOURCE: Author. TABLE 1 Brickley and James [1986], and Lee and Brewer [1985]); and the deposit-forbearance literature (see Kane [1986], Pyle [1986], Ronn and Verma [1986], and Thomson [1987]). The following proxy variables are constructed from stock-market data and balance-sheet and income data. MKTVAL = market value of the thrift’s stock. is the product of the price of the thrift’s stock and the number of shares outstanding, or the market value of equity. net worth according to generally accepted accounting principles. is the book value of equity. = proxy variable for liquidity. is nondeposit liabilities divided by total book liabilities. = proxy variable for diversification of assets. is the sum of nonmort gage loans and contracts and direct investments, divided by mortgage loans and contracts. proxy variable for solvency and a measure of capital adequacy. is divided by total book assets. MKTVAL TNW = LLQ DIV TNW LIQ DLV TNWA = TNW TNWA Empirical Tests of the Forbearance Hypothesis To test the forbearance hypothesis, we use the Statistical Market-Value A ccounting M odel (SMVAM) of Unal and Kane (1987): (1) + + Equation 1 is the basic SMVAM regression where is the value of the thrift’s stock and is the book value of the thrift’s equity. Unal and Kane interpret the slope coefficient, as the market’s value of $1 of book equity, and as the market’s value of unbooked equity. In other words, times is the portion of market value accounted for by assets-in-place, and is the portion of market value accounted for by the charter. If booked assets and liabilities are marked-to-market, then the theoretical value of is one; and if all assets and liabilities are carried on the books, the theoretical value of is zero. If the charter value net of FSLIC forbearances and guarantees is positive (negative), FSLIC forbear ances and guarantees will increase (decrease in absolute value terms) the size of Equation 1 is estimated over the cross-section of firms in the sample for each quar ter. As seen in table 1, is positive in every MKTVAL = Ue kTNW e. MKTVAL TNW k k, Ue TNW Ue k Ue Ue. Ue Proportion of Stock-Market Value Explained by Charter Value3 Quarter 1984 1 2 3 4 1985 1986 1 2 3 4 1 2 Num ber Ue /MKTVALh T-Billc GNMAd 43 43 43 43 0.35454 0.33716 0.26789 0.16006 0.0952 0.0987 0.1037 0.0806 0.1270 0.1414 0.1308 0.1254 43 43 43 43 43 43 0.12509 0.10347 0.20564 0.0852 0.0695 0.0710 0.0710 0.0656 0.0621 0.1268 0.1154 0.13311 0.13173 0.25091 0.1129 0.1070 0.0944 0.0957 a. Charter value is measured by the intercept term, LJe , in the SMVAM regressions. b. MKTVAL is the average stock-market value o f the firms in the sample. c. Annual equivalent yield o n 3-month Treasury bills traded on the secondary' market (from Interest Rates tables in selected Federal Reserve Bulletins, 1984-1986). d. Average net yields on Governm ent National Mortgage Association, mortgage-backed, fully m odified pass-through securities, assuming 12-year prepayment on 30 p o o ls o f FHA/VA mortgages (from Interest Rates tables in selected Federal Reserve Bulletins, 1984-1986). SOURCE: Author. TABLE 2 quarter. However, it is not significantly different from zero in five of the 10 quarters. Table 2 shows the percent of stock-market value accounted for by the estimated charter value, The value of the charter, which includes the FSLIC forbearances, ranges from a high of 35.4 percent in the first quarter of 1984 to a low of 10.3 percent in the second quarter of 1985. In other words, the charter is a nontrivial compo nent of stockholder equity. The per-dollar value the market places on book equity, appears in the sixth column in table 1. This value ranges from a low of 40 cents on the dollar in the second quarter of Ue. 3 0 k, Pooling and Cross-Equation Equality Restrictions for the SMVAM Regressions3 Uei U Ue io Test: = 0, e2 = 0, ............... F( 10,410) = 5.3392896+ Test: i = 2 ............ = F(9,410) = 0.62610870 Test: = 1, & 2 = 1, ............ , £io = 1 F( 10,410) = 102.89425+ Test: 2 = ............ = F(9,410) = 8.4505921+ Ue Ue = 0 Ue io k\ k\ - k k\o a. SMVAM Regression M odel: MKTVAL = Ue + kTN W + e. f Significant at the 1% level. SOURCE: Author. 1984 to a high of 82 cents on the dollar in the first quarter of 1986. In all quarters, is positive and significantly different from one at the 1 per cent level. As expected, there appears to be an inverse relationship between and the level of interest rates. The general upward trend in from the first quarter of 1984 to the second quarter of 1986 coincides with the downward trend in interest rates over this period. Table 3 presents the results of joint tests of the SMVAM coefficients and tests of pool ing restrictions. A seemingly unrelated system of equations, with each quarter estimated as a separate regression, is used to perform the tests. We reject the joint restriction that is zero in every equation at the 1 percent level, but we cannot reject the restriction that is equal across equations. For the slope coefficient, we reject both the cross-equation equality restriction and the joint restriction that equals one in every quarter at the 1 percent level. Overall, the results of the joint tests and the pooling restric tions support the forbearance hypothesis. Although the results of the SMVAM regressions are consistent with the forbearance hypothesis, the SMVAM specification does not provide a direct test of the forbearance hypothe sis. Recall that a thrift charter may have value exclusive of deposit insurance subsidies and for bearances because the charter also contains the net value of all unbooked assets and liabilities. Moreover, estimates of could be positive and significant when the value of FSLIC forbearances and guarantees is zero. Estimated could be insignificant (or negative and significant) when the value of FSLIC forbearances and guarantees is positive and significant. k k k Ue Ue k Ue Ue k, Results from the MSMVAM Regressions3 (U sing GAAP Net W orthb ) Quarter 1984 1985 1986 U 1 Us k 1395.67 (0.193)c 0.40942t (17.592) -19394.22 (-0.721) 109864.881+ (4.318) 124226.48 2 12627.61 (1.302) 0.42172+ (13-652) -66681.97* (-2.175) 49812.35* (1.934) 96534.29 (0.798) 0.7649 3 4231.84 (0.388) 0.46008+ (11.881) -40394.91 (-1.122) 27471.48 (1.094) 190151.80 (1.345) 0.7806 4 -5007.80 (0.454) 0.53010+ (9.679) -25806.27 (-0.775) 7272.61 (0.262) 320283.78* (2.021) 0.8080 1 -4171.15 (-0.383) 0.62986+ (7.879) -2115.69 (-0.065) -43843.94 (-1.508) 303247.14** (1.941) 0.8610 2 -8187.52 (-0.597) 0.70668+ (5.408) 15703.69 (0.398) -44814.60 (-1.443) 337791.44** (1.751) 0.8575 3 -8847.13 (-0.645) 0.55141+ (9.093) 8946.48 (0.238) -30860.57 (-0.874) 443532.27* (2.413) 0.8178 4 -23639.28 (-1.265) 0.67333+ (5.129) 8533.00 (0.177) 23836.61 (0.514) 599654.62* (2.374) 0.8097 1 -30194.55 (-1.464) 0.77275+ (3-332) 35213.95 (0.618) 35936.96 (0.618) 681527.26* (2.417) 0.8425 2 26518.43 (1.073) 0.75192+ (2.758) 73120.49 (0.849) -189767.15* (-2.131) 76236.99 (0.253) 0.7293 a. M odel: MKTVAL = Ue + kTN W + f t LIQ + f t / W + foTN W A + e. 0.8493 (1.483) 3 1 f Significantly different from on e at 1%. b. Net worth com puted using generally accepted accounting procedures. f t Significantly different from zero at 1%. c. T-statistics in parentheses. * Significantly different from zero at 5%. ** Significantly different from zero at 10%. SOURCE: Author. TABLE 4 A careful reexamination of the results in tables 1 and 2 indicates that the positive sign on in every quarter is due, at least in part, to the positive value of FSLIC guarantees and for bearances. There is an inverse relationship between and / the the market value of book equity increases, charter value as a percent of decreases. The value of for bearances and guarantees should be inversely related to On the other hand, the value of the charter exclusive of FSLIC forbearances and guarantees is expected to be positively correlated with This suggests that FSLIC forbearances and guarantees are a large enough portion of that changes in their value dominate the pattern of across quarters. To test the forbearance hypothesis more directly, we modify equation 1 to include the variables and to proxy for FSLIC forbearance policy: Ue k Ue MKTVAL. As MKTVAL k. k. Ue Ue LIQ, DIV, (2) TNWA MKTVAL = Ue + kTNW + f t LIQ DIV + f t TNWA + e. + LIQ, The first forbearance proxy, measures liquidity. Because the closing of an insolvent institution is an event timed by the reg ulators, insolvency is a necessary, but not suffi cient, condition for the forced closing of a thrift by its regulator. Given the growing insolvency of the FSLIC insurance fund and the large number of market-value and book-value insolvent thrifts (see Barth, et al. [1985] and U.S. General Accounting Office [1987]), the liquidity of the thrift affects the probability that FSLIC forbear ances will be extended to stockholders. Insolvent thrifts (those that are not running up large losses) tend to be closed when illiquid, especially when they are insolvent according to market-value accounting, but not book-value accounting. Ceteris paribus, the more liquid the thrift, the less likely a liquidity crisis will cause the FHLBB to close the thrift. There fore, the value of FSLIC forbearances should be positively related to liquidity. By construction, as increases, the thrift’s liquidity decreases. Consequently, f t should have a negative sign. The second forbearance proxy, is a measure of diversification in the asset portfo lio. includes both direct investments and nonmortgage loans and contracts. In March 1985, the FHLBB issued a formal regulation that restrict ed direct investments to less than the minimum of 10 percent of total assets and twice the amount of capital. This regulation, which was in effect LIQ DIV, DIV Pooling and Cross-Equation Equality Restrictions for the MSMVAM Regressions3 Ue i U Test: = 0, e 2 = 0, ............ , F( 10,380) = 0.5559142 Test: i - e 2 ............ = F(9,380) = 0.61687062 Test: = 1, 2 1, ............ F( 10,380) = 90.82540+ Ue io = 0 Ue U Ue io k\ k = , kio = 1 k\ DIV TNWA, TNWA TNWA k\o Test: = & 2 = ............ = F(9,380) = 6.8009228+ Test: /3i,i - 0, /?i,2 = 0, ............ , 0i,io = 0 F( 10,380) = 0.98109793 Test: 0i,i - /?i,2 = ............ = 0i,io F(9,380) = 0.82046518 32 its policy statements emphasizing mortgage lend ing during this period, 0 2 should be negative in the sample period from March 1985 on. Converse ly, could also be a proxy for management quality. 9 That is, the market may view a decrease in the thrift’s reliance on mortgages as an indica tion of the quality of management. This diversifi cation (management quality) explanation would make 0 2 positive before March 1985. After that time, the sign of 0 2 should be negative if the forbearance hypothesis holds. The third forbearance variable, proxies for solvency. Note that is solvency measured by book, not market, values. This means that a thrift with positive could be insolvent on a market-value basis.10 The value of deposit-guarantor forbearances depends on market solvency, not on O n the other hand, the probability of forbearance is a function of FHLBB-mandated capital requirements of 3% or more) are based on book values. FSLIC forbearances are extended to any institution that meets the minimum capital guide lines, and they may be extended to institutions with deficient capital ratios. Therefore, we use as our proxy for solvency because the probability of forbearance is a positive function of The sign on 03 should be positive. The results from the regressions on equation 2 are reported in table 4. Joint tests of the regression coefficients and pooling tests for the small sample appear in table 5. For all quarters, the estimates of are not significantly different from zero in the modified SMVAM (MSMVAM) regressions. In fact, we cannot reject the joint restriction that is zero in every quar ter or the cross-equation equality restriction on In the SMVAM regressions, estimated is significantly different from zero in five of the 10 quarters, and we reject the joint restriction that is zero. However, estimates are not affected by the inclusion of the forbearance prox ies. Estimated is positive and significantly less than one in every quarter, and we cannot reject the restriction that s m v a m = &m s m v a m in any quarter. Furthermore, both the joint test that equals one in every quarter and the cross equation equality restriction on are rejected at the 1 percent level for both the SMVAM and the TNWA TNWA ( TNWA TNWA TNWA Test: 0 2 ,1 - 0, 0 2 , 2 = 0, ............ , F( 10,380) = 2.8303445* Test: 0 2 ,1 Test: 0 3 ,1 Test: 0 3 ,1 = 0 2 , 2 = ............ = F(9,380) = 2.8692565* 0 2 ,1 0 = 0 Ue 0 2 ,1 0 Ue - 0, 0 3 ,2 = 0............... .. F( 10,380) = 2.9404988* - 0 3 ,2 - ............ = F(9,380) = 0.98635699 0 3 ,1 0 = 0 0 3 ,1 0 a. MSMVAM Regression M odel: MKTVAL = Ue + kTN W + PiLlQ + /32Z W + faTN W A + e t Significant at the 1% level. * Significant at the 5% level. SOURCE: Author. Ue. Ue Ue k k k k throughout the remainder of the sample period, applies only to nationally chartered thrifts, and not to the FSLIC-insured, state-chartered thrifts. The FHLBB is strongly opposed to direct investments by thrifts because it believes such investments increase the losses to the FSLIC fund when an insolvent thrift is closed (see Benston [1986]). Therefore, we expect there to be an inverse relationship between FSLIC forbearances and the level of direct investment. Given the FHLBB’s policy regarding direct investment and k 9 In economics, we assume that management is a scarce resource. Therefore, firms with high-quality management will have a higher market value than firms with lower-quality management. This, of course, assumes that the market for managerial talent is not perfectly competitive. ”1 The difference between market-based and accounting-based JL \ J measures of solvency can be quite large. A TNWA of 3 per cent is often used as a proxy for the solvency threshold on a marketvalue basis. MSMVAM regressions. The difference (similarity) in the behavior of ( between the SMVAM and the MSMVAM regressions is consistent with the forbearance hypothesis. The coefficients on the forbearance proxies themselves present a mixed set of conclu sions. The coefficient on /3i, is negative and significant in the second quarter of 1984, support ing the forbearance hypothesis. However, /3i is not significantly different from zero in any other quarter, and we cannot reject the joint restriction that equals zero in every quarter. Therefore, the overall performance of does not provide strong support for the forbearance hypothesis. 11 The poor performance by may be due in large part to sample selection bias. The thrifts in this sample are the largest in the industry and are likely to have greater access to national capital markets, and therefore greater sources of liquid ity, than the average thrift in the population. The results for the diversification (management quality) variable, are also mixed, is positive and significant in the first two quarters of 1984 and negative and significant in the second quarter of 1986. Moreover, /fe is positive in six of the 10 quarters in table 4. The cross-equation equality restriction on >82 and the joint restriction that is zero in every equation are both rejected at the 5 percent level. On the surface, the seemingly con flicting evidence provided by DIV seems to refute the forbearance hypothesis. But a closer inspec tion of the results indicates that this is not the case. Recall that the FHLBB policy restricting direct investment did not go into effect until the first quarter of 1985. Therefore, the positive and significant (insignificant) 's in the first (second) two quarters of 1984 are consistent with both the management-quality hypothesis and the forbearance hypothesis. Moreover, in table 4, 2 is positive but not significant twice, and negative and signif icant once, after the FHLBB took a stand against direct investment and against diversification of the asset portfolio away from mortgage-based assets. In fact, if we split the sample according to this policy change, we cannot reject the cross equation equality restriction on /fe in the preand post-policy change periods. However, in the first period we reject the joint restriction that /fe equals zero at the 1 percent level, but we cannot reject it in the second period. Ue k) LIQ, fii (3i LIQ DIV, fc fc (I n "I jl The poor performance of the liquidity proxy w as not due to proxy variable construction. Similar results were obtained with other specifications of Although Ld TNWA is TNW LIQ. scaled by total book assets, there is almost no correlation between of the quarters in either sample. TNWA and TNW for any O f all of the forbearance proxies, the solvency proxy, provides the strongest evidence supporting the forbearance hypothesis. 12 /fe is positive in every quarter and is significant in six quarters. The significance of in every quar ter from the last quarter of 1984 through the first quarter of 1986 coincides with the time period when the FSLIC fund was shrinking as a result of massive losses in the thrift industry (see U.S. General Accounting Office [1987] and Barth, et al. [1985]). The joint restriction that $ 3 equals zero in every equation is rejected at the 5 percent level. However, we cannot reject the cross equation equality restriction on $ 3 . Even though the results were somewhat disappointing when we look at the forbearance proxies individually, the overall results are encouraging. Looking at table 4, we see that in every quarter except the third quarter of 1984, is not significantly different from zero, and at least one of the forbearance proxies is significantly different from zero and correctly signed. Moreover, we obtain these results using a sample that is likely to be biased against support ing our maintained hypothesis. That is, our sam ple is drawn from the largest firms in the industry, and it is likely that we undersample the part of the industry for whom the FSLIC forbearance policy has the most value. TNWA, Ue IV. Conclusions and Policy Implications Deposit-insurance guarantees and forbearances have value. The value of FSLIC deposit guarantees and forbearances is reflected in the market value of thrift institution stocks. Proxies for FSLIC for bearances and forbearance policy are shown to be related to thrift charter values. The empirical results of this paper support Kane’s (1986) argu ment that FSUC forbearances and guarantees are an increasingly important source of thrift charter value. Our results also support Thomson’s (1987) theoretical result that the extension of forbear ances to stockholders of insolvent institutions increases the value of stockholders’ equity. Because deposit-insurance forbear ances to stockholders increase the value of the stockholders’ position in the firm at the expense of the federal deposit guarantor, and ultimately the federal taxpayer, the federal deposit-insurance agencies should always close out the position of the stockholders when reorganizing insolvent insti tutions. Capital forbearance programs, such as those utilized by the FHLBB in dealing with thrift insolvencies and those being used by bank regula tors for agricultural and energy lenders, result in a bailout of deposit institutions’ stockholders by the federal taxpayer. Our results support the concept of the management consignment program current ly used by the FHLBB to reduce the unintended 3 3 value of deposit-insurance subsidies. However, our results also indicate that the FDIC should re think its capital forbearance and open-bank assis tance policies, unless the bailouts of existing man agements and shareholders of failed and failing banks are the intended results of those policies. Benston, George J. “An Analysis of the Causes of Savings and Loan Association Failures,” New York University Graduate School of Business Administration, Monograph Series in Finance and Economics, vol. 4/5 (1986). Bowman, Robert G. “The Theoretical Relation ship Between Systematic Risk and Financial (Accounting) Variables,” vol. 34, no. 3 (June 1979), pp. 617-30. Journal of Finance, Brickley, James A., and Christopher M. James. “Access to Deposit Insurance, Insolvency Rules and the Stock Returns of Financial Institu tions,” vol. 16, no. 3 (July 1986), pp. 345-71. Journal of Financial Economics, Buser, Stephen A., Andrew H. Chen, and Edward J. Kane. “Federal Deposit Insurance, Regula tory Policy, and Optimal Bank Capital,” vol. 36, no. 1 (March 1981), pp. 51-60. Jour nal of Finance, 3 4 References Barth, James R., and Martin A. Regalia. “The Evolv ing Role of Regulation in the Savings and Loan Industry,” Paper prepared for the Cato Institute Conference on the Deregulation of Financial Services, Washington, D.C., February726-27,1987. Barth, James R., R. Dan Brumbaugh, Jr., Daniel Sauerhaft, and George H.K. Wang. “ThriftInstitution Failures: Causes and Policy Issues,” Proceedings of a Conference on Bank Struc ture and Competition, The Federal Reserve Bank of Chicago, May 1985, pp. 184-216. Beaver, William H., Paul Kettler, and Myron Scholes. “The Association Between Market Determined and Accounting Determined Risk Measures,” vol. 4 (1970), pp. 654-82. The Accounting Review, Flannery, MarkJ., and Christopher M. James. “The Effect of Interest Rate Changes on the Com mon Stock Returns of Financial Institutions,” vol. 39, no. 4 (September 1984), pp. 1141-53. Journal of Finance, _________ “Market Evidence on the Effective Maturity of Bank Assets and Liabilities,” vol. 16, no. 4, part 1 (November 1984), pp. 435-45. nal of Money, Credit, and Banking Jour Forde, John P. “Off-Balance-Sheet Items are Piling Up: Total Commitments Exceed $1 Trillion at Top Five Banks,” May 1, 1987, p. 1. American Banker, Kane, Edward J. “Appearance and Reality in De posit Insurance: The Case for Reform,” vol. 10(1986), pp. 175-88. of Banking and Finance, Journal The Gathering Crisis in Federal De posit Insurance. Cambridge, MA: MIT Press, _________ Bennett, Barbara. “Off Balance Sheet Risk in Banking: The Case of Standby Letters of Credit,” Federal Reserve Bank of San Francisco, Winter 1986, pp. 19-29. Economic Revieu\ Bennett, Dennis E, Roger D. Lundstrom, and Donald G. Simonson. “Estimating Portfolio Net Worth Values and Interest Rate Risk in Savings Institutions,” Proceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, May 1986, pp. 323-46. 1985. _________, and Chester Foster. “Valuing Conjectu ral Government Guarantees of FNMA Liabili ties,” Proceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, May 1986, pp. 347-68. _________, and Burton G. Malkiel. “Bank Portfo lio Allocation, Deposit Variability, and the Availability Doctrine,” vol. 79, no. 1 (February 1965), pp. 113-34. of Economics, The Quarterly Journal Lee, Cheng-few, and Elijah Brewer. “The Associa tion Between Bank Stock Market-Based Risk Measures and the Financial Characteristics of the Firm: A Pooled Cross-Section Time-Series Approach,” Proceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, May 1985, pp. 285-315. Lee, Cheng-few, and Morgan J. Lynge, Jr. “Return, Risk and Cost of Equity for Stock S&L Firms: Theory and Empirical Results,” Journal of the American Real Estate and Urban Economics Association, vol. 13, no. 2 (1985), pp. 167-80. Ronn, Ehud I., and Avinash K. Verma. “Pricing Risk-Adjusted Deposit Insurance: An OptionBased Model,” vol. 41, no. 4 (September 1986), pp. 871-95. Journal of Finance, Rosenberg, Barr, and Phillip R. Perry. “The Fun damental Determinants of Risk in Banking,” in Sherman J. Maisel, Ed., Chicago: Uni versity of Chicago Press and National Bureau of Economic Research, 1981. Risk and Capital Ade quacy’ in Commercial Banks. Bailout: An Insider’s Account of Bank Failures and Rescues. New York: Sprague, Irvine H. Basic Books, Inc., 1986. McTague, Jim. “Volcker Joins Opposition to House FSLIC Bill,” May 1, 1987, p. 2. American Banker, _________ “FDIC Broadens Its Guidelines on Failure Aid,” December 5, 1986, p. 2. American Banker, Myers, Stewart C. “Determinants of Corporate Borrowing,” vol. 5, no. 2 (November 1977), pp. 147-75. Journal of Financial Economics, _________ “The Relation Between Real and Financial Measures of Risk and Return,” in Irwin Friend and James Bicksler, Eds., Cambridge, MA: Bal linger Publishing Co., 1977, pp. 49-80. and Return in Finance. Risk Nash, Nathaniel. “Wright in F.S.LI.C. Reversal,” April 29, 1987. The New York Times, Pyle, David H. “Capital Regulation and Deposit Insurance,” vol. 10(1986), pp. 189-201. foum al of Banking and Finance, Thomson, James B. “The Use of Market Informa tion in Pricing Deposit Insurance,” 1987 (forthcoming). Money, Credit, and Banking Journal of Unal, Haluk, and Edward J. Kane. “Off-BalanceSheet Items and the Changing Market and Interest-Rate Sensitivity of Deposit-Institution Equity Returns,” Proceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, May 1987 (forthcoming). United States General Accounting Office. “Thrift Industry Forbearance for Troubled Institutions 1982-1986,” Briefing Report to the Chairman, Committee on Banking, Housing, and Urban Affairs, United States Senate, May 1987. Warner, Jerold B. “Bankruptcy Costs: Some Evi dence,” vol. 32, no. 2 (May 1977), pp. 337-47. foum al of Finance, 3 5 Third Quarter Working Papers Working Paper Notice The Federal Reserve Bank A s of January 1, 1987, w e no of Cleveland has changed its longer send method of distribution for the ing Paper series Work produced by the Bank's Research Department. Working Papers of charge to those who request to indi viduals as part of a mass mailing. Our current Working Papers will be listed on a quarterly basis in each issue of the Economic Review. Indi viduals m ay request copies of spe cific 3 6 Q "7 Amenities and the Returns to O / \J y Human Capital by Patricia E. Beeson. The author examines whether regional differences in the returns to human capital imply structural differences in regional labor markets, and finds that regional wage differen tials represent compensation for regional differ ences in amenities. Working Papers listed by com them. A regular mailing list for Working Papers, however, will not be maintained for personal subscrib ers. Libraries and other organiza tions m ay request to be placed on a mailing list for institutional subscrib ers and will automatically receive pleting and mailing the attached Working Papers as form below. Papers will be sent free published. they are O ""7 "I Monetary Policy in an Economy O / JL \ J with Nominal Wage Contracts by Charles T. Carlstrom. The author incorporates a standard nominal wage contracting model into the neoclassical optimal growth framework. It is shown that optimal monetary policy can be either procyclical or countercyclical, depending on the utility function, and that with logarithmic utility, the optimal policy rule will be to target the nominal interest rate. Please complete and detach the form below and mail to: Federal Reserve Bank of Cleveland Research Department P .O . Box 6387 Cleveland, Ohio Check item(s) 44101 Please send the following Working Paper(s): □ 8709 □ 8710 Send to: ____ Name Please print ______ Address City State Zip Code