The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
REGIONAL ECON OM IC ISSUES W orking P ap er S e r ie s Metro Area Growth from 1976 to 1985: Theory and Evidence William A Testa . FEDERAL RESERVE B A N K OF CHICAGO WP- 1989/1 M e t r o T h e o ry A r e a G ro w th a n d fro m 1 9 7 6 to 1985: E v id e n c e William A. Testa* Over the past 15 years, the process of regional change has been most un favorable to regions in the Midwest. In particular, the economy of the eastern part o f the Midwest—the East North Central Region (encompassing Ohio Indiana, Michigan, Illinois, and Wisconsin) has steadily declined. This region’s share of national employment fell from 20 percent to 17 per cent between 1972 and 1987. While the West North Central has fared somewhat better, agriculturally-oriented states such as Iowa and Nebraska, have generally paralleled the decline observed in the Midwest’s industrial belt. In response to lagging job opportunities in the Midwest, policymakers have put much effort into development programs intended to ignite robust eco nomic growth in the Midwest or, at least, to maintain the existing job base. This study is intended to help shape these programs by identifying the forces o f regional change, thereby suggesting more effective policy levers to stem the region’s decline. T h e o r i e s o f r e g io n a l c h a n g e In identifying potential policy levers, several economic theories explaining regional growth differences have been forwarded. In summary, it is prob ably safe to say that none of them have proven to be universal in explaining the U.S. experience. However, elements of all of them are consistent with the experiences of particular regions during particular time periods. In brief, the following theoretical frameworks are most prevalent in explaining the U.S. regional experience in recent decades. Neoclassical Theory: In its most basic form, this set of ideas considers factors of production—especially labor and capital—to be mobile across re gions. Labor will move to regions where real wage rates are highest. In *The author is Senior Economist at the Federal Reserve Bank of Chicago. The views and findings herein do not necessarily reflect the views of management of the Federal Reserve Bank. He thanks Joseph Crews, Alenka S. Giese and Natalie A. Davila for their research assistance and Stephanie Boykin for manuscript preparation. FRB CH ICAG O W orking Paper—January 1989 1 turn, capital will flow to regions of highest return. Under the most com mon assumption of one commodity which is produced by the same tech nology in all regions, differences in factor returns are determined by regional differences in endowments of labor and capital. So long as factor returns such as wages differ across regions, this theory would predict that labor will migrate to high wage regions of capital abundancy while capital would flow to low-wage regions o f labor abundancy.1 In examining the U.S. growth experience, this neoclassical growth mech anism has been related to the migration of the Southern poor to the facto ries o f the North during the 1950s and 1960s in search of high wages. Also, the strong capital investment observed in the South during the 1950s, 1960s, and 1970s can be partly attributed to low wages in that region. In general, the per capita income convergence observed across U.S. regions in this century has been related to neoclassical mechanisms. Export-Base Theory recognizes that regions will specialize in certain indus try products for reasons which are difficult to specify including historical accident, favorable location with respect to transport, endowment of a unique factor o f production, or increasing returns to scale of an industry or group of industries. These industry products or services of specialization form the “economic base” of the region and these goods are traded rather than nontraded. Export base theory posits that a region’s growth in em ployment and income can be best understood in terms of changes in de mand for the region’s traded goods by the rest of the world and nation or changes in competitive position which influence the quantity demanded. Empirically, divergences in regional growth can often be understood by a region’s mix of industries. In recent years, for example, the nation’s in creasing proclivity for foreign autos and domestic defense expenditure would be key features in explaining the relative decline in Michigan’s economy and the relative rise in the coastal economies. Product Cycle Theory suggests that industries and their attendant products undergo distinct stages or phases of a “life cycle” beginning with an inno vative stage. A t this initial stage, factor costs are of little relevance. Rather, specialized factor inputs such as entrepreneurial and innovative personnel, specialized business services, access to suppliers, access to emerging technological advances and ideas, and specialized financial ser vices are critical. A t the initial innovative stages and later in the early growth phases, the process technology of the industry cannot be transferred to another location. But later, as the process of production becomes standardized, the industry can become mobile in seeking locations where production costs are lowest. Because regions will maintain advantages in different factor costs, production will be spun out to low cost regions or even overseas. Examples from the U.S. experience have been the textile industry’s exodus from New England to the South Atlantic states and FRB CH ICAGO W orking Paper—January 1989 2 overseas. More recently, the production segment of the U.S. computer and the semiconductor industry has reportedly moved along the product cycle in such a fashion. While regions generally share the same fate as their industries, the recent New England experience has shown that a developed region’s capacity to innovate is not so permanently and easily eroded by the passing o f a single set o f industries through a product cycle. The technological legacy of a once-developed region may become the wellspring of a new set of indus tries. This observation has led others to assert that regions undergo waves or cycles o f growth and development.2 In other respects, the product cycle theory is not sufficiently developed to be useful as a predictive theory of regional growth. N o t all industries pro ceed through a product cycle (Malecki 1985) and evidence is lacking as to which industries the product cycle applies. Moreover, once a region has experienced decline (as its major industries move out through a product cycle), no guidance is offered on the conditions under which these regions will rejuvenate or if they will do so at all. Industrial Location Theory suggests that any regional growth theory must be reconciled with the micro decisions of each individual firm. In market economies, firms are thought to maximize their own profits (and minimize costs) in making capital investment decisions, including location alterna tives. In turn, firm location and expansion decisions accompany employ ment opportunities for a mobile labor force. While it is recognized that different types of firms respond to particular factor costs and conditions, the basic set of locational cost conditions can be generalized to include access to inputs and markets, labor, energy, transport, construction and lease costs, and those myriad services, regu lations, and costs which are often influenced by state and local govern ments. These costs, including tax levels and structure, service provision, environmental regulation, fiscal inducements and financial subsidies, and labor insurance, are often included under the popular rubric of “ business climate” . Business climate factors are often the focus of state/local officials because these factors can be influenced via the state/local government process or in partnership with the private sector. In addition to cost oriented features, a separate set of growth factors can be included under the rubric business climate. State and local governments have become active in business promotion. Prominent examples of such “demand side” policies are export or trade missions overseas; lobbying the federal government for a more favorable federal funds flow or identifying early procurement contract opportunities for regional industries; and “buy local” programs to increase the home demand for the region’s products. FRB CH ICAGO W orking Paper—January 1989 3 E m p i r i c a l s tu d ie s a s a f r a m e w o r k Within the body of empirical literature, the econometric specifications of regional growth range from purely ad h oc or intuitive empirical equations which are loosely based market frameworks (Plaut & Pluta 1983; Wasylenko 1984; Browne et al 1980; Kieschick 1981; et, al.) to carefully developed foundations based on micro theory of firm location behavior, specific functional forms, and statistical models (Carlton 1979; Hodge 1981; Crihfield 1985; Bartik 1985). Nonetheless, the functional forms that are ultimately used for statistical estimation often end up being very similar. The dependent variable, or what the analyst is trying to explain, is usually measured by either the growth rate in output or the growth rates of the factor inputs o f labor and capital across regions.3 Regional variation in growth rates are then “explained” by such factors as wages, energy costs, state-local taxes, labor climate, access to markets, climate, and an everexpanding host o f measures. One characteristic of much empirical work has been to assume that regions are slow to equilibrate differences in cost factors. This means that “ static” specifications of the empirical models have been generally dis carded (i.e., an equation that would explain levels of employment based on concurrent levels of wages and other factors). Rather, one-term changes in the dependent variable are hypothesized to be chiefly determined by be ginning period differences in factor costs (Wheat 1973, 1987; Plaut and Pluta 1983; Wasylenko 1984). Beginning-period regional cost differences are represented, for example, by relative wages, unionization levels, taxes, energy costs, and the like. Most studies assume that these cost differences are persistent and fairly constant thoughout the subsequent period of growth. While one can easily imagine that growth would additionally be determined by concurrent changes in factor costs (Plaut & Pluta), such considerations are usually neglected because of a scarcity of data observa tions.4 A recent innovation in this literature recognizes the simultaneous equations bias in those formulations which use “change in employment” as the de pendent variable. The theoretical underpinning suggests that these specifi cations are measuring the regional derived demand for labor and, in turn, the demand for labor depends on relative wages. But if so, then the wage is simultaneously determined with local supply of labor so that inclusion of wages in the single-equation estimation will yield inconsistent parameter estimates.5 As a result, some studies have used more appropriate statistical techniques such as instrumental variables to correct for simultaneous equations bias (Crihfield 1985; Papke 1984).6 FRB CH ICAGO W orking Paper—January 1989 4 Similar to much o f the existing literature, we estimate a “disequilibrium” model where economic performance (or growth) across SM SAs (Standard Metropolitan Statistical Areas) can be observed to adjust to initial period relative factor costs. While our empirical specifications are largely ad hoc, (i.e. the equations will not follow directly from a formal model using spe cific functional forms of a production function), the general form of our equations are similar to those grounded in economic theory (Engle 1974; Crihfield 1985; Papke 1984). The econometric specification chosen is in linear form; percentage change in employment is a linear function of beginning period levels of input costs (C) and other growth factors (OTH). Each metro area accounts for one observation so that the database can be considered as cross-sectional rather than time-series. P C H E M P i = a + bCi + c(077Z}) + e{ The linear form rather than log-linear is chosen insofar as some observa tions of employment change are negative, particularly for the manufactur ing industry. Accordingly, the logarithm o f these observations cannot be calculated. N o other functional forms were attempted. In an alternative specification, output growth substitutes for employment growth as the dependent variable. Output is measured by percentage change in value added in manufacturing-value added measuring the sum o f factor payments which is, in practice, roughly equivalent to the value of manufacturing output less purchased inputs. The employment growth equations represent labor demand equations for a region. Theory suggests that we attempt some statistical techniques to account for the simultanity bias in these employment equations. In fact, corrections for these consid erations have been made. However, the results reported here are of the ordinary least squares variety. It is noted that our results are robust with respect to estimating methods.7 The observations are drawn from the 75 largest SM SAs (Appendix I). SM SAs are economic regions having a common pool of labor, common statewide regulations, and common resources. While these areas have been criticized as economic units due to their proximity to other urban areas with which they are closely entwined, metro areas are more reasonable as econ omies than either counties, cities, or states—other geographic units for which data for important variables are available. The analysis was conducted for large industry categories: total employ ment, manufacturing and nonmanufacturing employment, and also manu facturing output. Owing to regional differences in business cycle timing and severity, a period of some length is required to capture secular growth FRB CH ICAGO W orking Paper—January 1989 5 trend differences. Growth in employment was measured in percentages for the 1976-85 period from C ou n ty Business P atterns data. Unfortunately, manufacturing output for metro areas has not been reported since 1982, so tills six-year time period is regrettably short. The 9-year period can be considered sufficient to capture the effects of regional cost differences on regional growth trends. A period that is too long may violate the assump tion that observed regional cost differences are constant over the period of study. Some analysts have used longer periods (Crihfield 1984; Wheat 1973, 1986) and others have used shorter periods (Plaut and Pluta 1983 and Wasylenko 1984). S t a tis tic a l r e s u lts The equations reported in Tables 1 and 2 are ordinary least squares esti mates where the dependent variable is expressed in percentage change from the beginning year to the endpoint year. The equations were checked for the common cross-sectional data problems o f multicollinearity and heteroskedasticity.8 The overall statistical results explain much of the variation in metro area growth over the 1976-85 period. “Explained variations” of between .4 and .5 are reported for the employment equations. These statistics are sub stantial for cross-sectional type analysis. In comparison, the manufacturing output equation does not perform as well, exhibiting an R 2 slightly under .3. This may be due to the shorter time period for which output data is available. Leonard Wheat (1986) has criticized Plaut and Pluta (1983) and others for using a short time period for this type of model because “cyclical effects, strikes, random spurts, and other short-run anomalies overshadow long-run trends” . The equations reported were arrived at from experimentation and iteration—there is no pretense that these are the outcomes of single “roll of the dice” as set forth from a structural model. A cross-section database was constructed covering many potential factors of importance which were gleaned from existing studies. In the course of this process, considerable care was taken in choosing and constructing variables that were thought to measure the concept that theory would suggest had an influence on regional growth differences. Some variables were ultimately dropped because collinearity between independent variables degraded the estimates of re maining coefficients. (See Table 3 for correlation coefficients of all the re tained variables.) In clear cases of bivariate collinearity, the variable that was retained had the greatest economic content and the most straightfor ward interpretation. For example, unionization variables were dropped in favor o f wages. FRB CH ICAGO W orking Paper— January 1989 6 Table 1 O L S reg ressio n equatio n: Em p lo ym en t and output g ro w th in m an u factu rin g Percent change in manufacturing employment (1976 to 1985) Percent change in manufacturing output (1976 to 1982) 1.04** (3.34) Labor Costs (WM76MFG) .95** (2.27) -0.15** (-2.76) Intercept -.18** (-2.45) -130.35* (-3.72) -149.11* (-3.22) Access to Technology (TECH) .04 (1.13) .07 (1-36) Defense Spending Per Capita (DOD) .0001 (1.21) -.00002 (-.14) Educational Expenditure Per Pupil (EDEXP) .0002* (1-78) .0001 (60) Tax Growth Per Capita (CHTX) -.003** (-2.06) -.0001 (-.07) Unemployment Insurance (UIMAN) -21.95 (-1.84) Market Maturity (MARKET) Export Orientation (XMFGEMP) R2 2.88 (-18) .03** (2.27) .03* (1.71) .44 .29 ‘ Significant at the 10 percent level. “ Significant at the 5 percent level. FRB CH ICAGO W orking Paper—lanuary 1989 7 Table 2 O L S reg ressio n equation: To tal and n o n m an u factu rin g em ploym ent g ro w th 1976 to 1985 percent change in total employment percent change in nonmanufacturing employment 1.01** (5.63) 1.03** (6.15) Labor Costs (WM76MFG) -0.007** (-3.71) -0.007** (-4.02) Market Maturity (MARKET) -91.66** (-3.73) -52.40** (-2.27) Access to Technology (TECH) 0.04 (1.52) 0.03 (1.29) Defense Spending Per Capita (DOD) 0.0002** (2.31) 0.00002** (2.29) Educational Expenditure Per Pupil (EDEXP) 0.0001 * (1.76) 0.0001 * (1.84) Intercept Tax Growth Per Capita (CHTX) -0.002** (-2.28) R2 0.48 -0.0001 ** (-2.22) 0.41 ‘ Significant at the 10 percent level. “ Significant at the 5 percent level. FRB CH ICAGO W orking Paper—january 1989 8 Table 3 C orrelation m atrix (2) (3) (5) (6) (8) (9) -0.11 -0.54 0.43 -0.12 -0.51 0.40 -0.45 -0.49 0.40 0.05 0.39 0.01 -0.11 1.00 (4) (7) (10) (13) (12) 1.00 -0.38 -0.35 -0.29 0.65 -0.41 -0.39 -0.13 1.00 -0.34 -0.24 0.01 1.00 0.73 -0.10 1.00 -0.10 1.00 3. PCNM 4. PCVA 0.68 0.67 5. WM76MFG 6. UPLTW 7. CHTX 8. MARKET 9. DOD -0.04 0.32 0.14 -0.04 0.33 0.41 0.03 0.13 -0.03 0.32 0.25 0.07 0.14 0.08 0.22 0.26 0.33 0.22 0.17 0.46 0.30 0.47 0.17 -0.24 0.03 0.00 -0.39 -0.29 -0.30 0.02 0.10 0.01 0.04 0.12 0.39 0.05 0.34 0.27 0.47 0.19 0.19 0.39 1.00 0.85 0.05 1.00 0.98 1.00 2. PCMFG -0.41 0.08 1.00 0.92 (11) 0.14 1.00 d) 1.00 1. PCTOT 0.17 10. EDEXP 11. TECH 12. UIMAN 10 .0 13. XMFGEMP Glossary of variables in regression equations CHTX Percent change in per capita state and local taxes from fiscal 1976-77 to fiscal 1984-85 EDEXP Education expenditure per pupil in A.D.A. 1976-77 DOD Per capita procurement and payroll by the Department of Defense in 1977 MARKET Ratio of value added (in $ millions) in manufacturing to population in the metro area TECH Total number of scientists and engineers engaged in research and development per 1,000 of the population, 1974 UPLTW Index of average hourly earnings of unskilled plantworkers, 1975-76 WM76MFG Average hourly wages, all manufacturing industries, 1976 XM FG EM P Percent of total m anufacturing em ploym ent related to exports, 1 9 7 6 PCMFG Percent change in manufacturing employment, 1976-1985 PCTOT Percent change in total employment, 1976-1985 PCNM Percent change in non-manufacturing employment, 1976-1985 PCVA Percent change in value added in manufacturing, 1976-1982 UIMAN Average statewide unemployment insurance rate (as a fraction of total wages) for 1975, 1976, and 1977 in the manufacturing sector FRB CH ICAGO W orking Paper—January 1989 9 D i s c u s s i o n o f fin d in g s An interpretation and discussion of the regression coefficients is presented below. L a b o r co sts Labor costs are invariably considered in statistical studies of growth. This is not surprising since labor costs comprise a large share of production costs. For example, manufacturers paid out 48 percent of value added to employee compensation in 1984. This share is possibly higher for service industries which tend to be more labor intensive than manufacturing. Sig nificant regional wage differences have been widely observed by researchers (See A C IR 1980 for a review) even though wages have displayed some convergence over the course of the century. Moreover, a significant body of research finds no real regional wage differences once education and ex perience of workers is taken into account (Dickie and Gerking 1987). Opinions remain somewhat divided on the importance of wage differences on regional growth disparities but evidence strongly suggests that wages do matter. The heavy weighting of labor-related costs in business climate rankings indicates that the popular wisdom equates high wage rates with poor business climate. For example, the Grant-Thornton annual M a n u fa ctu rin g C lim ates S tudy (1988) assigns over 40 percent of its factor weights to labor costs—and this excludes labor productivity and availability indica tors. Econometric studies beginning with Victor Fuchs (1962) have impli cated wages as affecting manufacturing employment growth across regions. Leonard Wheat’s analysis stands out as a well-known study rejecting the importance of wage differences. However, his study does report a state’s degree of unionization as highly significant (see also Bartik 1985) and these two variables are highly correlated in his study (rho = .73), as well as in our own study. Roger Schmenner’s extensive interview studies of manu facturing branch plant decisions identified labor-related costs as very im portant (1982). The estimations in Tables 1-2 report labor costs to be highly significant over the study period. As calculated in elasticity form (at the mean values of observations), employment growth was most responsive to labor costs for both the manufacturing and nonmanufacturing sectors (Table 4). The labor cost measures in the estimating equations represent the costs of hourly workers rather than salaried professionals. This suggests that the attraction of locations in the South for low-cost routinized production op erations, such as branch plants and production facilities of mature indus tries, continues to be a major force in regional growth disparities into the 1980s. FRB CH ICAGO W orking Paper— January 1989 10 Table 4 P o in t e la s t ic it ie s o f g r o w t h f a c t o r s (as evaluated at the mean) Manufacturing employment Nonmanufacturing employment Labor Cost 4.9 Labor Cost 1.4 Educational Spending 4.4 Educational Spending .5 Markets 1.9 Markets .3 Tax Growth 1.4 Tax Growth .3 Unemployment Insurance Tax 1.4 Defense Spending .1 Export Orientation 1.1 Technology* Access .1 Technology* Access .5 Defense Spending* Per Capita .3 *The underlying coefficient is not statistically significant at the 10 percent level using a two-tailed test. FRB CH ICAGO W orking Paper— January 1989 11 M a rk ets By market influence, we mean the relative balance in the beginning period between the demand for goods (and services) and the supply (Wheat 1986). Strong demand relative to accessible supply will exert a market pull to at tract suppliers to a more proximate location to the metro area. In Leonard Wheat’s recent study, a local market variable is constructed based on a state’s ratio of personal income to manufacturing employment (and also on distance to the manufacturing belt) as a proxy for demand to supply imbalance. Unlike the approach of Plaut and Pluta (1983), who weight these two market components of supply and demand by their dis tance from each and every state, Wheat asserts that it is local demand/supply rather than national market pull that exerts the most in fluence on growth. Following Wheat, and after some experimentation with both approaches, we also settle on a variation of the “local market pull” , using manufactur ing value added per metro area resident as our market measure. Although the point elasticity of the market variable ranks only third among growth factors (Table 4), the market variable enters first in a stepwise regression equation. One cannot decompose the explained variation in the regression attributable to each variable. However, the market variable, when entered alone, accounts for approximately one-half of the explained variance of the overall regression. N o other variable (alone) performs in this fashion so that, at an intuitive level, it appears that regional differences in market pull accounted for much of the interregional differences in growth over the sample period. As a matter o f conjecture, the market variable is thought to account for two distinct influences in the equations. First, enhancements in transportation—especially the advent of cheap and fast truck transport over the course of this century—is thought to have magnified the market pull of populous regions such as the South (Chinitz and Vernon 1960). With rail as the dominant mode of shipment, the relative costs of long haul shipments from the core manufacturing belt was fairly low compared with short haul transport. This is because, with rail transport, terminal costs are fairly high. With truck transport, terminal costs are much lower so that short haul transport from factory to market compares more favorably with long haul transport. As a result, the coming of interstate truck transport greatly enhanced the attractiveness of building branch manufacturing plants closer to the markets of final destination. This implies that, regardless of any migration of people to the South and West in this century, strong forces of market pull have been exerted because of changing transport technology and investment in infrastructure (highway). FRB CH ICAGO W orking Paper—January 1989 12 A second influence behind market pull has been the migration of people to warmer climates which has accompanied the rising incomes of retirees and the attraction o f population to the resource-rich Western states. Some studies such as Wheat’s have accounted for these two influences separately. In the current study, our sole market variable, value added per capita, is pulling double duty. It must also be noted that the two influences are simultaneous; growth in supply attracts population growth in search of jobs which, in turn, further enhances market pull. Some analysts have modelled this process as a si multaneous system of job and population growth (Steinnes 1984). U n e m p l o y m e n t in s u r a n c e ta x e s State-by-state differences in unemployment insurance systems greatly con cern many business groups and state chambers of commerce. This is espe cially so for those industry groups, such as construction and manufacturing, which tend to pay higher-than-average U I rates. The employment volatility of these industries is usually reflected in higher tax rates because state tax rates are “experience rated”—based on the unemployment history o f indi vidual firms. Accordingly, U I tax rates will often comprise a higher frac tion o f wage costs for firms in manufacturing and other plant-type industries. Concern over unemployment insurance costs are expressed by manufactur ers in the Grant-Thomton annual study of manufacturing climates. Input into this business climate ranking is provided by 36 associations represent ing manufacturers around the country. In the 1988 edition of the study, average benefits per covered worker are given a weight of 5.1 percent of the overall index and the net worth of the state unemployment insurance trust fund is weighted at 4.6 percent. Few statistical studies of regional growth differences consider U I tax rates. A statistical study by Roger W. Schmenner (1987) and others’ using an hi erarchical or two-stage sequential approach, examines the plant location decisions of 114 branch plant openings by Fortune 500 manufacturing firms during the 1970s. In the model, the unemployment insurance tax rate is measured by average unemployment compensation benefits paid per em ployed worker. This is a fairly cost-relevent measure reflecting current plus expected U I system liabilities to employers. However, little evidence is found in the Schmenner study that U I costs are influential in the decision to open branch manufacturing plants. FRB CH ICAGO W orking Paper— January 1989 13 Using employment data from 1973 to 1980, a study by Michael Wasylenko investigates the factors surrounding differences in growth among the 48 states on the U.S. mainland (1984). Major industry sectors under consid eration include manufacturing, transportation, administrative and auxiliary employment, wholesale trade, retail trade, services, finance-insurance-realestate, and total (employment). The measure of unemployment insurance is reported to be dropped from inclusion in the final results with no expla nation. Presumably, the variable displayed a perverse sign or collinearity with another variable(s). It should be noted that the particular measure of unemployment burden on employers, which is the average benefit paid to a worker receiving benefits, is not well chosen. A state with very gener ous benefits could burden firms very slightly if that state is experiencing high growth and low unemployment. This would tend to lessen the popu lation of unemployed workers and hence concomitant tax rates. A 1985 study by Timothy Bartik examines how corporate location decisions for new branch plants (using the same database as Roger Schmenner) are influenced by unionization, taxes, and other characteristics of states. The study results show no detrimental impact of high state U I tax rates on plant location. In fact, the sign o f the U I variable is unexpectedly positive for one of the reported estimating equations. In contrast to these existing studies, our results indicate that U I taxes neg atively influence employment growth over the 1976-85 period. These findings hold true for the manufacturing sector where, because the tax rates are often higher, one would most expect that high tax rates deter employ ment expansion. Unlike those measures used in the previous studies, our measure of the UI tax burden is industry-specific and mirrors the employer’s cost perspective. Still, this variable merits further investigation in that it is probably suscep tible to simultaneous bias. Slow growth causes high U I tax rates. While we account for this by choosing beginning period values of U I tax, the causation we measure could be reversed to the extent that growth in a re gion is serially correlated from one period to the next (i.e. slow growth in the prior period accounts for high initial values of U I tax rates which, in turn, are correlated with slow growth over the subsequent period of study). O r ie n ta tio n to m a n u fa c tu r e d e x p o r ts Throughout the 1970s, and peaking in 1980, the international trade share of U.S. output climbed steadily upward—imports and exports alike (Hervey 1986). Subsequently, merchandise exports fell off rapidly under the weight o f a rising dollar and significant import penetration. As a percent of FRB CH ICAGO W orking Paper—January 1989 14 GNP-output, merchandise exports had fallen to roughly the same level by 1984 as they had been in 1976. In addition to their importance to the nation’s economy, manufacturing exports have been demonstrated to greatly effect job generation in state economies. A t least two studies have documented a significant relationship between a state economy’s export orientation and economic growth (Crihfield 1985; Manrique 1987). It is not surprising, then, that state policies have placed greater resources in recent years into stimulating state exports abroad. For example, one analyst reports that between 1976 and 1980 alone, the number o f overseas offices maintained by state governments tripled (Posner 1981). A t least one study has uncovered a link between state export promotion activity and actual state export activity (Coughlin and Cartwright 1987). In our model, metro area export orientation is measured by the percent of an area’s manufacturing employment directly related to exports in 1976. The coef ficient on this variable is found to be statistically significant in both the manufacturing employment equation and in the manufacturing output equation. Consequently, the role of exports in state economic growth merits some attention as a factor that can be influenced by state and local policy. A c c e s s t o t e c h n o lo g y The high tech boom of the late 1970s and early 1980s furthered public awareness o f the importance o f technology to regional development. Those flourishing regional economies which we know so well in California and Massachusetts serve as a frequent reminder that a region’s technological base (and institutions) are important to economic growth—even to the ex tent that new industries can arise from the infrastructure legacy of longdeparted manufacturing industries. Technological factors are now recognized in business climate studies. The recent Ameritrust/SRI “ Indica tors of Economic Activity” lists nine measures o f regional technological capacity. The importance of technology in regional economic revival has even been noted overseas: “Places in America where modern manufactur ing has taken root and grown fastest tend to have three things in common: a handful of firms strong in one particular field; technical expertise on tap at a nearby engineering school or big government laboratory; and imag inative local bankers and investors” .9 To date, the importance of technology access to regional growth remains anecdotal rather than statistical (Markusen and Hall 1985; Sirbu et al 1976; Office o f Technology Assessment 1984). However, the strength of the re cent regional growth success stories, along with the well-documented his- FRB CH ICAGO W orking Paper—January 1989 15 tones o f the importance o f technology in their success, suggests that metro area accessibility to technology through joint ventures with universities and government labs, private consulting with university faculty, and interaction among industrial R & D facilities should possibly be included in future sta tistical studies. In attempting to measure a metro area’s access to technology, comprehen sive and condensable measures are not plentiful. A special survey con ducted by the National Science Foundation for 1974 reported on the number of scientists and engineers who are actually engaged in research and development activity by metro area. The coefficient of this measure proved to be quite robust over the course of alternative estimating equations. While the coefficients are not significant as reported in Tables 1-2, the co efficient sign remains consistently positive across industry sectors. However, while the presence of R & D activities can be measured, individual program initiatives that attempt to accelerate technological transfer from lab to market are not accounted for in these measures. Measurement re finements which account for public policy influence may yield more (or less) significant results. S t a t e -lo c a l ta x e s a n d s p e n d in g “ ...relative growth in manufacturing employment from 1939 to 1953 has not been highest where per capita state and local tax collections are lowest... ...relative growth in manufacturing employment from 1939 to 1953 has not been highest where increases in per capita state and local tax col lections have been held lowest...” Clark C. Bloom— 1956 “ ...an inverse relationship exists between changes in state relative tax burdens and state relative economic growth...” Robert J. Genetski— 1983 “ ...economic growth varies inversely with the burden of state and local government taxes; the fastest growing states, by and large, are states with relatively low tax rates....Even more important, changes in tax burden are strongly inversely related to economic growth...” Richard K. Vedder— 1981 The above passages exemplify the long-standing debate over the role of state-local taxes (and spending) in economic growth. Evidence and argu ment are as diametrically opposed today as 30 years ago. It would not be difficult to unearth one hundred or more statistical studies with the evi dence weighing significantly on either side. Despite the apparent conflict in the literature, we know more than the body of conclusions from these studies suggests. A look at the problems inherent FRB CH ICAGO W orking Paper—January 1989 16 in answering the question “do taxes matter?” helps to understand the con flicting results which have emerged. Many statistical studies, such as the ones cited above, perform only simple, one-by-one correlations between tax levels and economic growth. Because there are many influences on differential regional growth (and more im portant ones as well), the impact, if any, of tax levels or tax growth on re gional economic growth will be seriously distorted by such methods. However, it is not only statistical technique that has given rise to the con flicting evidence in the literature, but also the complexity of the question and the fact that taxes are not the primary determinant of regional growth differences. State and local taxes are usually a small fraction of total costs. Estimates of 3-4 percent of total costs are common. For this reason, some major research efforts have felt justified in neglecting taxation altogether in statistical studies (Fuchs 1962). However, some researchers have recently argued that, while taxes are in deed a small part of total costs, differences in taxes are larger relative to profits and thus they do influence relative rates of return to capital (and profit) by location (Papke 1984). Measurements of “ business taxes relative to business income” (Wheaton 1983) have also been shown to be larger than the taxes-to-total-cost measurements which were often cited in earlier studies (Cornia, Testa, Stocker 1978). Subsequently, researchers have carefully measured tax rates as they influence the price o f capital and they have entered them into statistical forms intended to explain location of capital investment. A t least one study found that, in using this careful measure (and correcting for simultaneous equation bias), location of in vestment expenditures is significantly related to the after-tax return on a marginal investment (Papke 1984). A second reason why tax levels are not thought to be important, and where statistical studies err, is that higher levels of taxation are frequently associ ated with higher levels of spending for services which, in turn, benefit businesses directly (e.g. highways and sanitation) or indirectly by enhancing quality of life (e.g. education and recreation) and thereby lower the level of wages necessary to compensate the workforce (Hoehn, Berger, and Blomquist 1987). As a result, it is not surprising to find some studies re porting that tax levels enhance economic growth rather than exert a fiscal drag (Romans and Subrahmanyam 1979; Plaut and Pluta 1983). In considering that public services can have value, one would also expect businesses to value certain types of services more than others. This has been accounted for in empirical studies by including variables to measure the composition of state-local government expenditures (Plaut & Pluta 1983; Newman 1983; Romans and Subrahmanyan 1979; Wasylenko 1984; Helms 1985). The idea here is that highway and education and infrastructure FRB CH ICAGO W orking Paper—January 1989 17 spending will more significantly benefit business than welfare spending and recreation. By accounting for these spending patterns in the estimating equation, the influence of tax levels can presumably be measured more ac curately. A third reason for the conflicting evidence on the tax-growth relation, and one which has gained recent popularity in tandem with federal tax reform, is that state and local tax structures are important rather than simply tax levels. Accordingly, studies focusing on tax levels alone will tend to be mis-specified. Some analysts contend that these differences in tax structure differ by region so as to cause differences in economic growth across re gions (Vedder 1981; Wasylenko 1984). Tax structure differences must then be accounted for in statistical studies which have, in fact, included variables such as the marginal corporate income tax rate (Kieschnick 1981) and the percentage of revenue raised from individual income taxes (Waslyenko 1984). The statistical results reported in Table 1-2 show that tax growth is signif icantly related to regional growth. One interpretation is that those metro areas that were not able to hold their initial tax burdens in check ultimately paid a price in terms of lower subsequent growth. This result is very close to those results reported by others who have cor related personal income growth by state along with the growth in “taxes per $1000 of personal income” (Genetski 1982; Vedder 1981). The latter studies have found strong negative correlations between personal income growth and measured growth in tax burden. However, such results have been strongly criticized as displaying reverse casualty. Over relatively short time periods, such as the length of a business cycle or less, one would find that slow-growing regions might necessarily experience increasing tax ef fort. As income falls, public expenditure needs fall less rapidly, driving up the tax rate. But such an observation hardly implies a direction of causality from tax burden to growth. In deference to these criticisms of the existing tax growth literature, statelocal taxes were measured on a per capita basis in the estimations presented here. A region experiencing economic decline would not experience the dramatic short term drop in population (so much as income) so that there would not tend to be an automatic increase in tax burden in response to lagging growth. For this reason, we believe our results to be more mean ingful than those others, such as Wasylenko (1984), which have measured tax burden using income-type measures in the denominator. In alternative and unreported specifications, the best available measures of tax burden levels were also entered into the empirical work including A C I R s measure of tax burden and William Wheaton’s careful measurements of business tax/business income. That tax levels did not turn out to be a significant FRB CH ICAGO W orking Paper—January 1989 18 variable in our estimations is a bit difficult to explain, (although Michael Wasylenko reports a similar result in his recent examination o f state eco nomic growth). The most straightforward explanation is that differences in taxes reflect monies needed to pay for regional differences in demand for local public goods. If so, variables reflecting regional differences in taste would need to be included if tax levels were to display significant coeffi cients. Other studies have found that the composition o f public spending also af fects economic growth. For example, Romans and Subrahmanyam report that transfer payments per dollar of state income are negatively related to growth. Michael Wasylenko reports education spending as a fraction of state income to be positively correlated with growth. We part slightly with Waslyenko by specifying the educational spending (elementary and sec ondary) variable more closely to service output—i.e. educational spending per pupil. Similar to Wasylenko’s recent analysis o f state economic growth, we find that the education coefficient has been positive and significant in accounting for metro area growth.1 0 F e d e r a l s p e n d in g The search for explanations of differential rates o f regional growth fre quently leads to the uneven geographic incidence of federal spending across the U.S. landscape. An extensive study by the Advisory Commission on Intergovernmental Relations (1980) documents the markedly changing in cidence o f federal spending away from Midwest and toward the South and West over the period from 1952 up through the mid-1970s. Coupled with a strong growth in the level of federal spending in the post-WW II era, the federal government is often accredited or blamed for an implicit industrial targeting that favors the Sunbelt (Markusen 1986). Among major categories of federal spending, defense spending grew most rapidly during the period of study; the defense budget growth has out stripped G N P growth in every year from 1978 to 1986. Moreover, defense outlays occupied almost 28 percent of federal government outlays in 1986. For these reasons, we chose per capita outlays by the Dept, of Defense as an important measure of federal spending incidence in metro areas. This component of the federal budget was found to exert a positive and significant impact on employment growth over the 1976 to 1985 period. Whether or not such job gains were offset or augmented by other federal spending and regulatory programs cannot be answered with our limited data set. FRB CH ICAGO W orking Paper—January 1989 19 C o n c l u s i o n s a n d p o li c y im p lic a t io n s Using metro area economies as observations, a cross-sectional study of growth over the 1976 to 1985 period is able to identify several key elements that account for regional growth differences in recent years. Regional dif ferences in wages and education exert strong hypothetical point impacts on metro area growth. Meanwhile, in terms of actual impacts on growth over the 1976-85 period, regional differences in market pull were highly influen tial. Among policy variables that can be manipulated by state and local offi cials; unemployment insurance, tax growth, educational spending, a state’s propensity to exports overseas, and technology can be listed as potentially important. However, several significant influences, including wages and the market pull of developing regions, will be more difficult for slow-growing regions to manipulate. These factors can possibly be maneuvered by tighter reins on alternative policy instruments. For example, educational im provement can potentially improve labor productivity, thereby offsetting labor cost disadvantages in some regions. F o o tn o te s 1 Results of the theorem are modified under differing assumptions about factor mobility, transport costs, differences in technology, and multifactor production. 2 See Douglas E. Booth, “Regional Long Waves and Urban Policy,” Urban Studies. Vol. 24 No. 6, December 1987, pp. 447-459. 3 In the bibliography, see references to Crihfield, Steinnes, Fuchs, Borts and Stein, Papke, Wasylenko, Kieschnick, Plaut and Pluta, Wheat, ACIR, Kahley, Newman, and Browne. One exception remains—the work of Carlton who formulates a sta tistical model using conditional logit analysis on the probability of firm birth and expansion in any given region (Carlton 1979). Similarly, other studies have bor rowed this basic framework and have estimated it using more refined statistical specifications (Bartik 1985). 4 One exception is Crihfield (1985) who had a sufficient number of data observa tions to include both the initial period level of relative costs along with changing relative costs. 5 It is more accurate to say that most empirical work purportedly measures shifts over time in demand for labor and supply of output. FRB CH ICAG O W orking Paper—January 1989 20 Equations explaining output per se do not suffer from this simultaneous equations bias because the price of output (i.e. demand for output facing a single small re gion) can be assumed to be fixed for a small region selling to a national or inter national market. 6 Using the instrumental variables approach, John Crihfield used six to seven variables to identify labor demand (i.e. to shift the supply of labor), depending on the specification chosen. The variables included state income taxes as a frac tion of state personal income, local prices as reflected in housing rents, local government expenditures as a fraction of local personal income, state government expenditures as a fraction of state personal income, nominal social security pay ment in the locality, the local unemployment rate, and local real wages in 1960. Leslie Papke chose the unemployment rate and the percent of workforce unionized to create an instrument for wages in that study (1984). 7 Instruments for wages were constructed using tax effort, unionization, unem ployment insurance system generosity, and unemployment rate (see Appendix II). 8 In such cases, like the present, where the size of the observations varies mark edly, there may be reason to suspect some heteroskedascity in the error terms. Accordingly, the residuals were plotted against the population of the SMS A in 1976. No heteroskedasticity was evident. Bartlett’s test was performed over the top and bottom one-third of this ranked sample. The hypothesis that the error variances were equal could not be rejected at the 5 percent significance level. As in all cross-sectional samples, multicollinearity lowers the efficiency of the parameter estimates. As seen by the correlation coefficients of the independent variables, bivariate collinearity does not appear to be a severe problem. In addi tion, analysis of the type practiced by Belsley-Kuh-Welsch suggested that severe collinearity was not present in the reported equations. 9 Automation Alley: “Rust bowls can regain their shine by playing to their in dustrial strengths,” Economist , April 11-17, 1987. 1 The educational spending per pupil variable has been criticized as measuring a 0 single input among many in the production of education rather than an output of education. Unfortunately, output variables are difficult to measure. In this study, the percent of the adult population with at least a high school education was attempted as a replacement for educational spending. These latter results were consistent with the results reported herein. FRB CH ICAGO W orking Paper— January 1989 21 B ib lio g r a p h y Advisory Commission on Intergovernmental Relations, R egion al G row th : H istorica l P ersp ective , Washington, D.C., June, 1980. Ameritrust and SRI, Indicators o f E con om ic C a p a city , December, 1986. Bartik, Timothy V., “ Business Location Decisions in the United States: Estimates o f the Effects o f Unionization, Taxes, and Other Charac teristics of States” , Journal o f Business and E con om ic Statistics Jan. 1985, pp. 14-22. Belsley, David A., Edwin Kuh, and Roy E. Welsch, R egression D iagn ostics , John Wiley & Sons, New York, 1980. Bloom, C. C., S ta te and L o ca l T a x D ifferentials and the L ocation o f M a n ufacturing, Studies in Business and Economics, Bureau of Business and Economic Research, University of Iowa, No. 5, 1956. Borts, George H. and Jerome L. Stein, E con om ic G row th in a F ree M a rk e t, Columbia University Press, New York, 1964. Brown, Lynn E., et. al., “ Regional Investment Patterns,” N ew England E co n o m ic R ev iew , Federal Reserve Bank of Boston, July/Aug. 1980, pp. 5-23. Carlton, Dennis W., “Why New Firms Locate Where They Do: An Econometric M odel” Interregional M o v e m e n ts and R egion al G row th , pp. 13-50. William C. Wheaton ed., The Urban Institute, Washington, D.C., 1979. Carlton, Dennis W., “The Location and Employment Choices of New Firms: An Econometric Model With Discrete and Continuous Endogenous Variables,” The R eview o f E con om ics and S tatistics , Vol. 65 (August 1983), pp. 440-444. Chinitz, Benjamin, and Raymond Vernon, “Changing Forces in Industrial Location,” H arvard Business R ev iew , Vol. 38, 1960 pp. 126-136. Cornia, Gary C., William A. Testa and Frederick D. Stocker, S ta te-L o ca l Fiscal Incentives and E con om ic D evelo p m en t , The Academy for Con temporary Problems, Columbus, OH, 1978. FRB CH ICAGO W orking Paper—January 1989 22 Coughlin, Cletus C. and Phillip A. Cartwright, “An Examination of State Foreign Export Promotion and Manufacturing Exports,” Journal o f R eg ion a l S cien ce, Vol. 27, No. 3, pp. 439-450. Crihfield, John, A n E m pirical A n a lysis o f R egion a l D em a n d and Sup p ly F unctions, Ph.D. Dissertation, University o f Chicago, Chicago, IL 1985. Dickie, Mark, and Shelby Gerking, “Interregional Wage Differentials: An Equilibrium Perspective,4 Journal o f R egion al S cien ce, Vol. 27, No. 4, 1987. Engle, Robert F., “A Disequilibrium Model of Regional Investment,” Journal o f R eg ion a l S cience, Vol. 14, No. 3, 1974, pp. 367-376. Fuchs, Victor, Ch anges in the L oca tion o f M anu factu rin g in the U. S . Since 1 9 2 9 , Yale University Press, New Haven, C T , 1962. Genetski, Robert J. and Lynn Ludlow, “The Impact of State and Local Taxes on Economic Growth,” H arris E con om ics, December 17, 1982, pp. 1-15. Grant Thornton, M anu factu rin g C lim ates S tu d y , Grant Thornton, Chicago, 1988. Hall, Peter, and Ann R. Markusen, “High Technology and Regional-Urban Policy,” in Peter Hall and Ann Markusen eds., Silicon Lanscapes, Allen R. Unwin Inc., Winchester, M A , 1985. Helms, L. Jay, “The Effect of State and Local Taxes on Economic Growth: A Time-Series Cross-Section Approach,” The R eview o f E con om ics and Statistics, 1985, pp. 574-582. Hervey, Jack L., “The Internationalization of Uncle Sam” , E con om ic P e r spectives, Federal Reserve Bank of Chicago, Vol. X, No. 3, May/June 1986, pp. 3-14. Hodge, James H., “A Study of Regional Investment Decisions” in J. Vernon Henderson ed., R esearch in Urban E con om ic, JAI Press, 1981, pp. 1-65. Hoehn, John P., Mark C. Berger, and Glenn C. Blomquist, “A Hedonic Model of Interregional Wages, Rents, and Amenity Values,” Journal o f R eg ion a l S cien ce, November 1987, pp. 605-620. FRB CH ICAGO W orking Paper—January 1989 23 Kahley, William J., C om parative A dvantage and S tate E m p loym en t C h ange, Working Paper 86-2, Federal Reserve Bank of Atlanta, Jan. 1986. Kieschnick, Michael, T a x es and G row th : Business Incentives and E con om ic D e v elo p m en t, Council o f State Planning Agencies, Washington, D.C., 1981. Malecki, Edward J., “Industrial Location and Corporate Organization In High Tech Industries,” E con om ic G eorg a p h y, October, 1985. Manrique, Gabriel G, “ Foreign Export Orientation and Regional Growth in the U.S.,” G row th and C h an ge, Winter 1987, pp. 1-12. Markusen, Ann, “Defense Spending and the Geography o f High Tech In dustries,” in John Rees ed., T ech n olog y, R egion s, and P o licy, Rowman & Littlefield, Totowa, N.J., 1986. Newman, Robert J., “Industry Migration and Growth in the South” , R e view o f E con om ics and Statistics, 65, 1983, pp. 76-86. Olson, Mancur, The R ise and D eclin e o f N a tion s , Yale University Press, New Haven, C T , 1982. Papke, James A., and Leslie E. Papke, M ea su rin g D ifferential S ta te -L o ca l T a x Liabilities and Their Im plications F o r Business Investm ent and P lant L oca tion , Center for Tax Policy Studies, Purdue University, West Lafayette, IN, 1986. Papke, Leslie E., “The Influence of Taxes on the Location of Manufactur ing Activity: New Evidence” in Indiana's R evenue S tructure: M a jo r C om p on en ts and Issu ed P art II, James A. Papke ed., Center for Tax Policy Studies, West Lafayette, IN, 1984, pp. 115-130. Plaut, T. R., and J. E. Pluta, “ Business Climate, Taxes and Expenditures, and State Industrial Growth in the U.S.,” Southern E con om ic Journal 50, 1983, pp. 99-119. Posner, Alan R., “The States and Overseas Export Promotion,” M S U B usiness Topics, vol. 28, Summer, pp. 43-49. Romans, Thomas and Gonti Subahmanyam, “ State and Local Taxes, Transfers, and Regional Economic Growth,” Southern E con om ic Journal, October 1979, pp. 435-44. Schmenner, Roger W., M a k in g Business L ocation D ecision s, Prentice-Hall Inc., Englewood Cliffs, N.J., 1982. FRB CH ICAGO W orking Paper—January 1989 24 Sirbu, Jr., M. A., R. Tretel, W. Yorsz, and E. B. Roberts, The Form ation o f a Technology Oriented Com plex: Lessons fro m N orth American and European Experience, C P A Report 76-78, Center for Policy Alterna tives, Massachusetts Institute o f Technology, 1976. Steinnes, Donald N., “Business Climate Tax Incentives, and Regional Economic Development,” Growth and Change, April, 1984, pp. 38-47. Vedder, Richard K ., Joint Econom ic Committee R eport, State and Local Economic Development Strategy: A “ Supply Side Perspective” , Con gress o f the United States, 97th Congress, 1st Session, Washington, D .C., 1981. Wasylenko, Michael, The E ffe ct o f Business Clim ate On Em ploym ent G row th: A R eport To The M innesota T a x Study Commission, June 28, 1984. Wheat, Leonard F., “The Determinants of 1963-77 Regional Manufactur ing Growth: Why The South and West Grow” , Journal o f Regional Science, Vol. 26, No. 4, 1986, pp. 635-659 Wheaton, William C., “Interstate Differences in the Level of Business Taxation,” National T a x Journal, March 1983, pp. 83-94. FRB CH ICAGO W orking Paper—January 1989 25 Appendix I Metropolitan areas included in the statistical analysis (ranked by 1976 SMSA population) 1976 SMSA Population 1976 SMSA Population New York Chicago Los Angeles Philadelphia Detroit Boston Oakland (w/S.F.) San Francisco (w/Oak.) Washington, D.C. Dallas (w/Ft. Worth) Fort Worth (w/Dallas) Houston St. Louis Pittsburgh Baltimore Minneapolis Newark Cleveland Atlanta Columbus Anaheim San Diego Miami Denver Seattle Milwaukee Tampa-St. Petersburg Cincinnati Buffalo Kansas City Riverside Phoenix San Jose Indianapolis New Orleans Portland Hartford San Antonio 9,605,000 7,003,800 6,981,500 4,784,500 4,414,500 3,930,400 3,156,400 3,156,400 3,056,500 2,603,400 2,603,400 2,389,900 2,367,300 2,313,800 2,144,100 2,042,300 1,990,000 1,955,200 1,849,300 1,806,600 1,776,000 1,655,900 1,465,400 1,442,400 1,431,500 1,428,500 1,427,100 1,379,100 1,322,400 1,290,300 1,262,900 1,257,300 1,210,100 1,156,800 1,117,000 1,103,600 1,059,900 993,600 FRB CH ICAGO W orking Paper—January 1989 Rochester Sacramento Louisville Fort Lauderdale Memphis Providence Dayton Salt Lake City Birmingham Albany Norfolk Toledo Greensboro Oklahoma City Nashville Jacksonville Akron Syracuse Scranton Gary-Hammond Allentown Charlotte Orlando Tulsa Richmond Omaha Jersey City Grand Rapids Greenville Raleigh-Durham West Palm Beach Tucson Fresno Oxnard-Ventura Knoxville Harrisburg Austin 978,700 902,960 895,300 886,300 880,500 862,500 835,200 802,500 802,300 799,700 787,300 782,100 776,200 772,900 769,700 716,100 663,900 648,000 643,600 635,900 627,500 605,800 601,400 596,300 596,100 579,800 579,700 567,000 534,700 483,700 480,500 467,300 460,800 459,500 443,200 432,600 410,800 26 Appendix II In stru m en tal va ria b le s reg ressio n equ atio n : Em ploym ent and o u tp u t g ro w th in m an u factu rin g Percent change in manufacturing employment (1976 to 1985) Percent change in manufacturing output (1976 to 1982) Intercept 1.12** (2.94) 1.26** (2.46) Labor costs (W M 7 6 M FG ) -.1 7 * * (2.10) - .24** ( 2.03) Market maturity (M A R K ET ) A ccess to technology (T EC H ) 137.7** (3.88) -1 5 3 .4 ** ( 3.26) .04 (-91) .07 (1.30) Defense spending per capita (D O D ) .0002 (1.34) .00001 ( .0 7 ) Educational expenditure per pupil (ED EX P ) .0002 (1.37) .0005 ( 28) Tax growth per capita (C H TX ) - .0 0 2 (1.43) .0007 ( 44) Unemployment insurance (U IM A N ) Export orientation (X M FG EM P ) R2 - 1 4 .9 (- 1 .8 4 ) .03** (2.14) .41 11.16 (67) .03 (1.51) .28 ‘ Significant at the 10 percent level. “ Significant at the 5 percent level. FRB CHICAGO Working Paper—lanuary 1989 27 Appendix II (cont'd) In stru m e n ta l v a ria b le s r e g re ssio n e q u a tio n : T o ta l an d n o n m a n u fa c tu rin g e m p lo y m e n t g r o w th 1976 to 1985 Percent change in total employment Percent change in nonmanufacturing employment 1.10** (5.67) 1.10** (5.97) Labor costs (W M 76M FG ) -0 .0 0 8 ** ( - 3 .6 3 ) -0 .0 0 7 ** ( - 3 .4 8 ) Market maturity (M A R K ET) -9 7 .8 1 ** ( - 4 .0 5 ) -6 2 .9 2 ** (- 2 .7 5 ) A ccess to technology (T EC H ) 0.04 (1.59) 0.03 (1.33) Defense spending per capita (D O D ) 0.0002** (2.36) 0.00002** (2.37) Educational expenditure per pupil (ED EX P ) 0.0002* (1.81) 0.0001 * (1.84) Intercept Tax growth per capita (C H TX ) R2 -0.001 * (- 1 .7 5 ) -0 .0 0 0 1 * ( - 1 .6 4 ) 0.47 0.38 ‘ Significant at the 10 percent level. “ Significant at the 5 percent level. NOTE: The wage variable is created as an instrument by regressing unemployment rate, tax burden, Ul generosity, and unionization on the wage index, UPLTW and WM76MF6. FRB CHICAGO Working Paper—January 1989 28