Globalization and Productivity in the United States and Germany

Board of Governors of the Federal Reserve System International Finance Discussion Papers Number 595 November 1997 GLOBALIZATION AND PRODUCTIVITY IN THE UNITED STATES AND GERMANY Catherine L. Mann NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. References in publications to International Finance Discussion Papers (other than an acknowledgment that the writer has had access to unpublished material) should be cleared with the author or authors. Recent IFDPs are available on the Web at www.bog.frb.fed.us.

GLOBALIZATION AND PRODUCTIVITY IN THE UNITED STATES AND GERMANY Catherine L. Mann* Abstract: This paper investigates the impact of globalization on productivity growth and the procyclicality of productivity growth in manufacturing industries in the United States and Germany. For U.S. industries, the analysis suggests that changes in international demand affects productivity growth differently from changes in exposure to international competition. An increase in foreign demand for U.S. exports raises trend productivity growth, but to a lesser degree than does a similar demand shock from domestic buyers. On the other hand, whereas an increase in U.S. imports reduces trend productivity growth of U.S. industries, a loss of market share to imports is associated with gains to productivity growth. For Germany, neither international demand shocks nor exposure to international competition seem to be associated with productivity growth rates, perhaps because German industries experienced a smaller increase in exposure to international competition over the time period. Comparing the U.S. and German results suggests that "going global" may affect productivity growth rates more than simply "being global". As for the procyclical characteristics of productivity growth, the U.S. and German measures evidence different procyclical behavior. For many industries, both U.S. and German labor productivity growth rates exhibit some degree of procyclicality. For German industries, this procyclicality of productivity growth disappears with broader measures of productivity growth that include utilization of capital and intermediates inputs. For U.S. industries, the degree of procyclicality increases when productivity growth is measured on these broader bases. Moreover, in the United States, procyclicality appears to be accentuated by export demand growth and dampened by import demand growth. Keywords: globalization, trade, productivity, procyclicality * Catherine Mann is Assistant Director in the International Finance Division of the Federal Reserve Board. This paper was prepared for the conference: "Globalization, Technological Change, and the Welfare State" at the American Institute for Contemporary German Studies of the Johns Hopkins University on June 8-10, 1997. The views in this paper are solely the responsibility of the author and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or of any other person associated with the Federal Reserve System.

1 I. Introduction This paper investigates the impact of globalization on productivity in manufacturing industries in the United States and Germany. Using disaggregated data from the manufacturing sectors in Germany and the United States, I investigate two hypotheses of how globalization and productivity might be related. The first hypothesis investigates the relationship between globalization and changes in the productivity growth rate. The second hypothesis investigates the importance of globalization for the procyclical characteristics of productivity. Globalization is proxied by real exports and imports, both real volumes and as a share of output and apparent domestic consumption, respectively. Productivity is measured three ways: (1) Labor productivity; (2) A Solow residual from a calculation including labor and utilization-adjusted capital; (3) A Solow residual from a calculation including labor, materials, and utilization-adjusted capital. The time period analyzed is 1979 to 1995 for the United States and 1981 to 1994 for Germany. For U.S. industries, the analysis suggests that international demand growth affects trend productivity growth differently from the effect of greater international exposure. Increased export demand is associated with an increase in trend productivity growth; but this increase is less than is associated with an increase in domestic shipments. Thus, the positive correlation between productivity growth and the share of exports in output found by some other researchers is not corroborated by this study.1 On the other hand, while increased import growth is associated with lower trend productivity growth, an increase in import competition (measured by the share of imports in domestic demand) increases trend productivity growth. Thus, whereas increased imports apparently do not induce productivity enhancements, loss of market share to imports does. Comparing across the measures of productivity growth, labor markets do not bear the brunt of the reallocation of inputs necessary to achieve productivity gains; reallocations in the use of materials is particularly important for raising productivity growth rates. Capital utilization is more affected by the overall level of output, regardless of source or destination. 1See for example, the work of J. Bradford Jensen and Nathan Musick, Andrew B. Bernard and J.Bradford Jensen, and Martin Neil Baily and Jans Gersbach.

2 For Germany, no element of international demand or exposure seems to be related to productivity growth rates.2 One reason for this different behavior could be the difference in initial and subsequent international demand and exposure. German industries generally started the period under examination with a much higher share of output exported and a much higher share of domestic demand satisfied by imports. A smaller additional share of German output was exported over the period, and import penetration rates also grew more slowly. By the end of the period, however, the exposure of U.S. and German industries, as defined by exports as a share of output and imports as a share of domestic consumption, are fairly similar. Further examination of the German data suggests another reason why it may be difficult to find any relationship between productivity growth and globalization: average productivity growth over the period varies quite substantially across the alternative measures. For the U.S. calculations, the productivity measures corroborate each other as to which industries are ones with high average productivity growth rates; but, for Germany, the three measures of productivity growth do not move together. Instead, for many of the industries, the productivity growth measures other than labor productivity are negative, indicating a deterioration in the efficiency in the use of non-labor resources over time. As for the procyclical characteristics of productivity growth, the U.S. and German measures evidence different procyclical behavior. For many industries, both U.S. and German labor productivity growth rates have some degree of procyclicality. For German industries all procyclicality of productivity growth rates disappears with the broader measures; this would appear to be associated with the negative productivity growth of these broader measures for many industries, and is consistent with the results of other researchers.3 But, for the U.S. industries, the degree of procyclicality increases when productivity growth rates are measured on a broader basis to include capital and materials utilization. Moreover, the 2These results differ from those of Fitzenberger. He uses a sample period that begins in 1975. 3See Gebhard Flaig and Viktor Steiner.

3 degree of procyclicality appears to be related to both export and import demand growth. Export growth accentuates procyclicality, particularly through an effect of capital utilization. Import growth appears to dampen procyclicality through the capital channel. The structure of the paper is as follows: The next section reviews the construction of the different measures of productivity and discusses how globalization might affect productivity. Section III describes the data and how each of the measures of productivity was implemented with the data. Section IV analyzes the results for the United States and for Germany. Section V offers some final remarks. II. Measures of productivity growth and the role for globalization Measures of productivity growth There is no standard measure of productivity growth, nor a standard by which to judge which of these might best measure technological change. Accordingly, I examine three measures. Each is the residual from the calculation of the difference between the log change in a measure of real output (a value-added measure—y or a gross-output measure—q) and the log change in measures of real inputs (including, variously weighted, labor—l, capital services—k, and materials—m). 1. Labor input measure of productivity: Labor = y -l Labor productivity is a good place to start. Since it has the fewest variables and since it does not require a decision about how to account for other inputs or the returns to scale in production, it is easiest to implement with the data and is less prone to misinterpretation--labor productivity growth is what it is. On the other hand, labor productivity growth is limited as it does not indicate how firms adjust inputs to achieve output gains, and it clearly is not a measure of technology. Many forces other than technological change can affect labor productivity–in particular a change in the capital-labor ratio. 2. Solow residual measure of productivity growth: Solow = y - a *l - (1-a )*l *K a is the labor share in nominal value added K is the capital stock; capital services are proportional

4 to K with factor of proportionality l . The Solow residual is the benchmark against which many other productivity measures are judged. Unfortunately, while clear as a theory, the assumption that capital services are proportional to the capital stock often yields unsatisfactory movements in the original Solow measure of productivity when implemented with data. 3. Labor and capital utilization-adjusted measure of productivity growth: K&L = y - a *l - (1-a )*k This measure of productivity is derived in Burnside et al.4 They assume that real gross output (q) is produced using value-added (y) and materials (m) in a Leontief specification. As shown above, value added output (y) is then produced using labor inputs (l) and capital services (k) according to constant returns to scale. The Leontief assumption implies that variation in the use of materials inputs cannot be a source of variation in productivity growth. 4. Labor, materials, and capital input-adjusted measure of productivity growth: KLM = 1/(1-g ) [q - b *l -g *m - (1-b -g )*k]; b is the labor share in nominal gross output g is the materials share in nominal gross output 1/(1-g ) adjustment to value-added basis5 This productivity measure extends the previous measuring to include the productivity gains that come from a reallocation of materials inputs, as well as of other inputs.6 Basu and Fernald find that one important reason for measures of aggregate productivity to change is that shocks cause resources to be reallocated across industries--from lower-valued uses to higher valued uses. This reasoning could apply equally well to the reallocation of resources across plants within an industry. Role for forces of globalization 4Craig Burnside, Martin Eichenbaum, and Sergio Rebelo (1995). 5This simple adjustment is correct if we assume that the markup of price over cost does not change; see discussion of this assumption is Basu and Fernald, pp 25-26. 6This specification can also be found in Burnside, et al.

5 Globalization, as proxied by trade, could affect the trend productivity growth rate and could affect the procyclicality of the productivity measures. The following decomposition of output points out two channels: the growth of real exports and real imports, and changes in the share of exports in output and in the share of imports in domestic demand. Total output (Q) equals the amount produced to satisfy domestic consumption (D) as well as to satisfy export demand (X). On the other hand, total output is reduced by the amount of domestic consumption that is satisfied by imports (M). Thus, the growth in total output can be calculated as: dQ/Q = dD/D * D/Q + dX/X * X/Q - dM/M *M/D However, for globalization per se to affect productivity, international sales or forces of international competition must elicit a different response from the firm than do domestic sales or domestic competition. Suppose a firm responds to demand shocks coming from exports or imports differently than it does to a shock to domestic demand. Changes in the growth rate of imports or exports would affect productivity growth differently than a similar change in growth of domestic demand. In addition, changes in the share of exports in output or in imports in domestic demand would affect the importance of the international shocks. Why should firms react differently to international shocks? Firms might respond differently to international shocks because these shocks convey different information about production techniques than do domestic shocks. Imported goods can be "reverse engineered", which teaches firms foreign production technology. In addition, if among import competing firms some are more efficient and cost-effective producers, these firms will survive import competition longer than other import competing firms, thus raising the average productivity growth rate of all firms that survive. On the export side, firms that export a high fraction of their product could have a more flexible and efficient production technology which increases their ability to meet foreign design demands. Moreover, firms that do export may be the most cost effective producers world-wide of a particular product. Bernard and Jenson suggest that, for the United States, firms that export are also the most efficient producers

6 and the technology leaders in an industry. On the other hand, international competition could hurt productivity growth. Increases in import market share might contribute to a deterioration of the productivity of the domestic industries as output falls--ultimately the domestic industry would disappear. Gearing up to export to many countries, although in the long-run contributing to higher productivity growth, could initially hurt productivity growth as firms figure out how to sell into markets with different standards and tastes. The differential response of firms to international forces can affect the procyclicality of productivity measures. To the extent that firms reallocate resources less in response to shocks that they perceive to emanate from the international dimension (either export shocks or import shocks), then procyclicality overall could be augmented or dampened, depending on whether the shock was positive or negative. To summarize, there are three hypotheses to test on the relationship between globalization and measures of productivity growth. The first two focus on the long-run, crosssectional relationship between increased globalization--measured either as a change in growth of exports or imports or as a change in the share of exports and imports--and changes in the trend rate of growth of productivity. The third focuses on the role for globalization to accentuate or damp the procyclicality of productivity and output. III. Data and Empirical implementation U.S. and German Data Table 1 lists the industries examined for the United States and for Germany. For the United States, all data are either annual or annualized monthly data for the time period 1978 to 1995. All are manufacturing industries and most produce either inputs or finished products destined for the manufacturing sector. Shipments, value-added output, and producer prices; factor inputs and factor

7 prices; and trade values and trade prices are matched by industry code7; the same index of materials and components for manufacturing was used for all industries.8 For Germany, all data are annual for the time period 1981 to 1994. The set of German industries includes basic inputs, capital goods, and finished/consumer goods. Thus the sample of German industries includes more output destined for the household sector, as compared to the U.S. data sample. Gross-output, value-added output and producer prices; factor inputs and factor prices; and trade values and trade prices are matched by industry code; the materials input price index and energy price index are the same for all industries.9 Empirical implementation The empirical implementation of the calculations outlined in Section II is not completely straightforward and not entirely consistent across the two countries. For example, the measure of (the log change) in labor input is the (log change in) number of production workers for the United States, but is the (log change in) the wage bill less (log change in) average hourly wage for Germany. Which data to use to proxy for capital services is particularly difficult. Following Burnside et.al, I use (log change in) kilowatt hours for the U.S. industries. For Germany, capital services is proxied by the (log change in) energy bill less (log change in) energy price. Finally, an important data issue is that industry-specific producer prices were used to deflate both U.S. shipments and German gross-output (which is appropriate) and value-added output (which is not correct).10 Tables 2 and 3 7The trade data are end-use codes matched to the SIC codes of the industry and factor data. The output and input data are from the Annual Survey of Manufactures. The producer price and wage data are from the Bureau of Labor Statistics. The trade data are from the Department of Commerce. 8It would be superior to calculate industry-specific materials deflators using an input-output table, but data on disaggregated materials inputs and their prices are not available for the sample period investigated here. The NBER Productivity database has a fuller range of industries, but a less-up-todate time period. The more recent data were desirable given the focus on the effect of globalization. 9Data on energy usage shortened the time period. All data are from Statistisches Bundesamt. 10The lack of a value-added price index may be important for interpreting some of the results of the empirical implementation of the algebraic specifications of the previous section. The lack of a valueadded price index implies that the capital- and material-adjusted indexes differ more in principle than in

8 describe more precisely the series and transformation used in the implementation. practice. That is, in principle, we have a true measure of real value added, and that differs from real gross output. However, in practice, we have nominal value added and nominal gross output and only one price deflator. What this implies is that K&L and KLM, as implemented, differ in a specific way: KLM = K&L - g /(1-g ) (dP /P - dP /P ) matl matl output output If materials prices are rising faster than output prices, the capital and materials-adjusted productivity residual (KLM) will be smaller than the capital-utilization adjusted residual (K&L). If the relationship between the two prices is changing over time, and changing across industries, it could influence our interpretation of how industries achieve productivity gains through input reallocation.

9 IV. Results and Discussion Productivity Measures The first step is simply to look at the different productivity measures in several ways: time series, procyclicality, and cross-correlation. Charts 1 and 2 show, for selected U.S. and German industries respectively, the time series representations of the three productivity measures. As is common, the time series are quite volatile and it is difficult to discerne procyclicality or whether there is any trend in the productivity growth rates. Tables 4 and 5 show coefficients of procyclicality for each industry for each of the three measures of productivity growth for the United States and for Germany. The coefficient of procyclicality is derived from a simple regression of the productivity measure against a constant and (log change in) contemporaneous real shipments for the time period 1978 to 1995 for the United States, and against a constant and (log change in) real gross output for the time period 1981 to 1994 for Germany. For many industries, both U.S. and German labor productivity growth rates have some degree of procyclicality. For German industries any degree of procyclicality of productivity growth rates apparently disappears with the broader measures of producitivity growth. This may be associated with the negative productivity growth of these broader measures for many industries and, in any case, is consistent with the results of other researchers.11 For the U.S. industries, the degree of procyclicality increases when productivity growth rates are measured on a broader basis to include capital and materials utilization. 11See Gebhard Flaig and Viktor Steiner.

10 Productivity Means Measuring the trend rate of productivity growth is difficult, as is judging whether that trend rate has changed. Charts 3 and 4 show the sample means for the three productivity measures for each industry in the U.S. and Germany. The first observation from the U.S. productivity measures is that the means of the three measures are broadly consistent across industries—those industries with low means and those with high means are generally the same regardless of the productivity measure. The second observation is that across the different measures of productivity, labor productivity is often the highest of the three measures, except for industries with the highest rate of growth of productivity when capital and materials are accounted for, such as 365, 366, 367. The labor productivity measure would generally have a mean higher than the other two measures since all the residual between input and output is being ascribed to labor. When the other productivity growth measures, which account for the use and reallocation of multiple inputs, have a higher mean, it suggests that the shares of capital and/or materials inputs into the production process for these industries are falling. In contrast to the U.S., the means of the different German measures are not broadly consistent for an industry sector--there is no clear pattern of high-productivity or low-productivity sectors, regardless of the measure. What is more notable is the great degree to which, across industries, the three productivity growth measures differ from each other. Whereas the mean of the labor productivity measures are generally large and positive, some of the labor- and capital-adjusted productivity growth measures and many of the labor-, capital-, and materials-adjusted measures have negative means--suggestive of a declining rate of growth of productivity in the use of these inputs.12 A last way to describe the relationship between the three measures of productivity growth is to take the correlation of their sample means, as shown in Table 6. The means of the three measures of productivity are correlated with each other, but to differing degrees. The labor and labor- 12This observation of high labor productivity growth but negative growth rate of productivity when measured to include inputs other than labor is consistent with the rise in unemployment and the rise in the capital/labor ratio in Germany that others have observed in aggregate data. See for example, World Economic Outlook, IMF, May 1996, Chart 21.

11 and capital-utilization adjusted residuals are highly correlated for both the United States and Germany. Once materials usage is included, the correlations of the mean of the productivity measures drops substantially for Germany. For the United States, all three measures are highly correlated. Patterns of globalization Because patterns of globalization are central to the analysis, Table 7 for the United States and Table 8 for Germany show how the two measures of globalization have changed for the two countries and for the industries. Each Table shows for each industry the share of exports in output (initialXsh) and imports in domestic demand (initialMsh) for the first year of the sample as well as the change in percentage points in those shares to the end of the sample period. German industries generally had higher initial shares of output exported and of domestic demand satisfied by imports, as compared to U.S. industries. Almost as a consequence of this higher exposure initially, a smaller additional share of German output was exported over the period and import penetration rose less than it did for the U.S. industries. From lower initial exposures, U.S. industries experienced a much higher increase exports as a share of output and imports as a share of domestic demand. By the end of the sample period, the exposure of U.S. and German industries was similar, with German exposure somewhat higher. Globalization and productivity growth rates The first hypothesis asks, for an industry, whether globalization is related to productivity growth rates through either of two channels: Does industry productivity respond differently to international demand shocks than to domestic demand shocks? Are the industries with the greatest increase in productivity growth rate those with the greatest increase in international exposure (e.g. increase in exports or imports as a share of their output)? Tables 9 and 10 show for the United States and Germany the output from regressions that are based on the discussion in Section II. The empirical approach splits the sample period, calculates the means of the variables in the two sub-sample periods, and differences the means.

12 Splitting the sample and differencing yields proxies for how trends in productivity growth, domestic demand growth, export and import growth, and export and imports shares changed between the first and the second half of the sample; this also helps to eliminate industry-specific factors. A separate regression for each productivity measure is shown. Thus the regressions take the form (with variable names in italics): [difference(mean(productivity growth rate))] = (labor, K&L, KLMva) a* [mean (share of exports in output) in the first half of the sample] (Xsht0) +b* [mean (share of imports in domestic demand) in the first half of the sample] (Msht0) +c* [difference(mean(growth rate of domestic shipments))] (dlDomship) +d* [difference(mean(share of exports in output))] (difmnXsh) +e* [difference(mean(share of imports in domestic demand))] (difmnMsh) +f* [difference(mean(growth rate of exports))](dlXgrow) +g* [difference(mean(growth rate of imports))](dlMgrow) Controlling for the initial exposure to export and import competition (a and b), does productivity growth on average respond differently to mean growth of external demand than to mean growth domestic demand (e.g., are the coefficients on f and g significantly different from zero and different from c); and how does productivity growth change as the share of exports in output changes (d) and as import market share changes (e)? Table 9 shows results for the United States. First a high percentage of the variation in the data is explained by the variables. Second, an increase in mean growth of domestic shipments is associated with an increase in mean productivity growth rate. An increase in export demand also is associated with an increase in the productivity growth rate (0.40), but to a lesser extent than the increase in domestic demand (0.57); thus increased dependence on exports as a share of output apparently reduces mean productivity--as the negative sign on the export share variable also suggests. An increase in mean import growth negatively affects mean productivity growth (-0.10), although the

13 deterioration in productivity growth is substantially less than it would be given a decline in domestic demand (-0.57); thus increased import competition apparently increases productivity--as the positive sign on the import market share variable also suggests. Altogether, these results suggest that losses in market share to imports stimulates producers to find more efficient ways of combining resources thus raising mean productivity growth. Export demand does support productivity, but apparently exporting is hard work. A high dependence on exports as a share of output is associated with lower mean productivity growth perhaps because it is difficult to efficiently use resources to meet varied foreign standards and tastes. For Germany (Table 10), very little of the variation in the data is explained by either the domestic or globalization variables. It may be that German industry does not distinguish between domestic and international demand shocks because there has been relatively little increase in globalization over the sample period. The fact that domestic demand variables are also not important in explaining variations in productivity growth suggests that factors such as domestic regulation may be the driving force behind changes in trend productivity growth. This would be consistent with the very different behavior of the three measures of productivity growth.

14 Trade and the procyclicality of productivity and output The last hypothesis outlined in Section II was that the procyclicality of productivity and output might be affected by the interaction of industries’ different responses to globalization with the importance of international exposure. Tables 4 and 5 showed the cofficient of procyclicality for each of the productivity growth measures for the U.S. and German industries. Tables 11 and 12 show the results of a simple regression of these procyclicality coefficients against the globalization variables: [cross-section of industry coefficients of procyclicality] = (labor, K&L, KLMva) a* [mean(share of exports in output) in first half of sample] (Xsh-t0) +b* [mean(share of imports in domestic demand) in first half of sample] (Msh-t0) +c* [mean(growth rate of exports))](dlXgrow) +d* [mean(growth rate of imports))](dlMgrow) Not surprisingly, there is little evidence that labor procyclicality in Germany is related to international demand or exposure (Table 11). However, there is some evidence that procyclicality evidenced in U.S. data (Table 12) might be related to international factors--principally through the channel of capital usage. Strong export demand growth increases the procyclicality of the K&L measure, whereas strong import growth would appear to dampen the coefficient of procyclicality of this measure. One explanation for this behavior is that U.S. firms base capital decisions on expected domestic demand. Higher export demand must be met with the same capital--thus raising procyclicality. Higher import demand smooths out what would be peaks in domestic demand--thus dampening procyclicality.

15 V. Final Remarks As an overall summary, these results suggest that for the United States but not for Germany the forces of international demand growth and competition have important and independent effects on productivity growth rates. U.S. firms respond differently to globalization than to domestic forces. One difference between the U.S. and German firms is on the extent to which international exposure increased for the U.S. firms--both exports as a share of output and imports as a share of domestic demand. Increased global competition may spur increases in productivity growth more than just simply facing the same degree of international competition. Besides being interesting in their own right, these results feed into several important debates ongoing in the profession and in the policy community: first, the debate on the effect of globalization on labor markets and second, the debate on the origin of procyclicality. Many studies of the impact of trade on labor markets assume that globalization and productivity growth are independent forces.13 Most of these econometric or factor analyses of employment or relative wage dispersion measure the effect of globalization with import prices or import flows, but do not account independently for productivity growth. Instead, the substantial residual unexplained effect in the regression is termed “technological change”. If trade and productivity growth are interdependent forces affecting the labor market, then including only one in a simple regression will bias that coefficient: up to the extent that the two forces covary positively or down to the extent that they covary negatively. The results in this paper for the United States indicate that international demand as well as international competition affect trend productivity growth rates. Thus, the next step in research on the effects of international forces on U.S. labor markets should include export as well as import variables and should explicitly take account of productivity growth. 13For Germany, Fitzenberger op. cit. For the United States see for example: Eli Berman, John Bound, and Zvi Griliches; George Borjas and Valerie Ramey; George Borjas, Richard Freeman, and Lawrence Katz; Robert Feenstra and Gordon Hanson; Paul Krugman and Robert Lawrence; Robert Lawrence and Matthew Slaughter; Edward Leamer; Jeffrey Sachs and Howard Shatz; Adrian Wood.

16 The results in this paper also are important for the research that addresses the role of technology vs. other factors in affecting, on the one hand, the procyclicality of productivity and output and, on the other hand, long-term growth. Researchers have examined various hypotheses for the source of procyclicality in U.S. data including profit margins, labor effort, capital utilization, variation in the shares of inputs, and resource reallocation across industries.14 This paper suggests that global forces of competition and technology transfer may be different from domestic forces in generating procyclical productivity, at least in the United States. 14See for example Susanto Basu; Basu and John Fernald; Mark Bils; Craig Burnside, Martin Eichengreen, and Sergio Rebelo; Ricardo J. Caballero and Richard K. Lyons; Matthew Shaprio; Robert M. Solow.

17 References Baily, Martin Neil and Jans Gersbach (1995), "Efficiency in Manufacturing and the Need for Global Competition," BPEA--Microeconomics. Basu, Susanto (1996), "Procyclical Productivity: Increasing Returns or Cyclical Utilization," Quarterly Journal of Economics, August , p.719-751. Basu, Susanto and John Fernald (1997), "Aggregate Productivity and Aggregate Technology," manuscript, Federal Reserve Board, February. Berman, Eli, John Bound, and Zvi Griliches (1994), "Changes in the Demand for Skilled Labor within U.S. Manufacturing Industries: Evidence from the Annual Survey of Manufactures," Quarterly Journal of Economics, February. Bernard, Andrew B. and J.Bradford Jensen (1995), "Exceptional Export Performance: Cause, Effect, or Both?", manuscript, December. Bils, Mark (1992), "Measuring Returns to Scale from Shift Practices in Manufacturing," manuscript, University of Rochester. Borjas, George and Valerie Ramey (1994), "Time-Series Evidence on the Sources of Trends in Wage Inequality," American Economic Review, May. Borjas, George, Richard Freeman, and Lawrence Katz (1991), "On the Labor Market Effects of Immigration and Trade," NBER WP no. 3761, June. Burnside, Craig, Martin Eichenbaum, and Sergio Rebelo (1995), "Capital Utilization and Returns to Scale," in B. Bernanke and J. Rothenberg, eds., NBER Macroeconomics Annual. _____,_____,_____ (1996), "Sectoral Solow Residuals," European Economic Review, p 861-869. Caballero, Ricardo J. and Richard K. Lyons (1992), "External Effects in U.S. Procyclical Productivity," Journal of Monetary Economics, vol 29, pp 209-226. Feenstra, Robert and Gordon Hanson (1995), "Foreign Investment, Outsourcing, and Relative Wages," NBER WP no. 5121, May. Fitzenberger, Bernd (1996), "Wages, Prices, and International Trade: Trends Across Industries for an ’Export Champion’," manuscript University of Konstanz, October. Flaig, Gebhard and Viktor Steiner (1993), "Searching for the ’Productivity Slowdown’: Some Surprising Findings from West German Manufacturing," Review of Economics and Statistics. Jensen, J. Bradford and Nathan Musick (1996), "Trade, Technology, and Plant Performance," ESA/OPD 96-4, U.S. Dept of Commerce, Economics and Statistics Administration, February. Krugman, Paul and Robert Lawrence (1994), "Trade, Wages, and Jobs," Scientific American, April. Lawrence, Robert and Matthew Slaughter (1993), "Trade and U.S. Wages: Great Sucking Sound or

18 Small Hiccup?," BPEA--Microeconomics-BPEA. Leamer, Edward (1996), "In Search of Stolper-Samuelson Effects on Wages" in Susan M. Collins, ed. Imports, Exports, and the American Worker, The Brookings Institution. Sachs, Jeffrey Sachs and Howard Shatz (1994), "Trade and Jobs in U.S. Manufacturing," BPEA.. Shapiro, Mathew (1996), "Macroeconomic Implications of Variations in the Workweek of Capital," BPEA vol2, p.79-119. Solow, Robert M. (1957), "Technological Change and the Aggregate Production Function," Review of Economics and Statistics, vol 39, p 312-320. Wood, Adrian (1995), "How Trade Hurt Unskilled Workers," Journal of Economic Perspectives, Summer.

19 Table 1: Included Industries United States Germany Industry Code Industry Code Basic steel products 331 Stone, clay, glass products 25 Foundry products 332 Non-ferrous metal product 28 Non-ferrous metal products 333 Chemicals 40 Fasteners, stampings 34567 Logs and planks 53 Ordnance 348 Pulp and plywood 55 Engines, turbines 351 Rubber products 59 Farm machinery 352 Various steel products 302 Construction equipment 353 Railroad, metal girders 31 Metalworking machinery 354 Farm machinery 32 Special industrial machinery 355 Autos 33 General industrial machinery 356 Electric home appliances 36 Office &computing machinery 357 Optics and clocks 37 Service industry machinery 358 Iron and steel sheets 38 Electrical industrial apparatus 362 Office machinery &computers 50 Household appliances 363 Musical equipment 39 Lighting and wiring products 364 Fine ceramics 51 TV and radio sets 365 Glassware 52 Communication equipment 366 Finished wood products 54 Electronic components 367 Finished paper products 56 Misc. electrical machinery 369 Books 57 Autos and parts 371 Finished plastic products 58 Ships and boats 373 Shoes 61 Scientific and medical eqpt. 381234 Textiles 63 Photo supplies and eqpt. 386 Apparel 64

20 Table 2: Implementation for U.S. Data For each industry code; industry code subscripts not shown 1. Labor productivity: labor = y -l y: log-change in nominal value-added less log-change in producer price l: log-change in production employment 2. Labor and capital utilization-adjusted residual: K&L = y - a * 1 - (1-a ) * k y: log-change in nominal value-added less log-change in producer price; a : ratio of wage bill for production employees to nominal value added l: log-change in production employment k: log-change in kilowatt hours 3. Labor, materials, and capital-utilization adjusted residual: KLM = 1/(1-g ) * (q - b * l -g * m - (1-b -g ) * k) ; q: log-change nominal gross-output less log-change in producer prices b ratio of wage-bill for production employees to nominal gross output l: log-change in production employment g : ratio of materials costs to nominal gross output m: log change in nominal materials costs less log change in (common) deflator for materials in manufacturing k: log change in kilowatt hours

21 Table 3: Implementation for the German Data For each industry code; industry code subscripts not shown 1. Labor productivity: Labor = y -l y: log-change in nominal value-added less log-change in producer prices; l: log-change in nominal personnel costs less log-change in wages. 2. Labor and capital-utilization adjusted residual: K&L = y - a * 1 - (1-a ) * k y: log-change in nominal value-added less log-change in producer price; a : ratio of wage bill to nominal value added l: log-change in nominal personnel costs less log-change in wages k: log-change in nominal energy input cost less log-change in (common) energy index 3. Labor, materials, and capital-utilization-adjusted residual: KLM = 1/(1-g ) * (q - b * l -g * m - (1-b -g ) * k) q: log-change nominal gross-output less log-change in producer prices b ratio of wage-bill to nominal gross output l: log-change in nominal personnel costs less log-change in wage g : ratio of materials costs to nominal gross output m: log-change in nominal materials costs less log-change in (common) deflator for materials in manufacturing k: log change in nominal energy input costs less log-change in (common) energy index

-- 22 -- Chart 1: Productivity Measures for Selected Industries: United States Productivity Measures: United States Productivity Measures: United States Industry 333 Industry 352 80 20 60 40 10 20 0 0 -20 -40 -10 -60 -80 -20 1979 1981 1983 1985 1987 1989 1991 1993 1995 1979 1981 1983 1985 1987 1989 1991 1993 1995 K&L labor KLMva K&L labor KLMva Productivity Measures: United States Industry 363 Productivity Measures: United States Industry 357 20 200 15 150 10 100 5 50 0 0 -5 -50 -10 -100 1979 1981 1983 1985 1987 1989 1991 1993 1995 1979 1981 1983 1985 1987 1989 1991 1993 1995 K&L labor KLMva K&L labor KLMva Productivity Measures: United States Industry 386 Productivity Measures: United States Industry 371 40 20 20 10 0 0 -20 -10 -40 -60 1979 1981 1983 1985 1987 1989 1991 1993 1995 -20 1979 1981 1983 1985 1987 1989 1991 1993 1995 K&L labor KLMva K&L labor KLMva

-- 23 -- Chart 2: Productivity Measures for Selected Industries: Germany Productivity Measures: Germany Productivity Measures: Germany Industry 28 Industry 33 80 20 60 10 40 0 20 -10 0 -20 -20 -40 -30 -60 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 -40 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 K&L labor KLMva K&L labor KLMva Productivity Measures: Germany Productivity Measures: Germany Industry 32 Industry 50 20 30 10 20 0 10 -10 0 -20 -10 -30 -20 -40 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 -30 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 K&L labor KLMva K&L labor KLMva Productivity Measures: Germany Productivity Measures: Germany Industry 36 Industry 51 15 15 10 10 5 5 0 0 -5 -5 -10 -10 -15 -20 -15 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 K&L labor KLMva K&L labor KLMva

-- 24 -- Table 4: Coefficient of Procyclicality--United States productivity measure labor K&L KLM Basic steel products 331 0.309 * 0.648 * 0.097 Foundry products 332 0.13 0.098 -0.054 Non-ferrous metals 333 0.4 * 1.75 * -0.231 Fasteners, stampings 34567 0.295 * 0.01 0.062 Ordnance 348 0.269 * 0.106 0.162 Engines, turbines 351 0.222 * 0.378 * 0.468 * Farm machinery 352 0.216 * 0.243 * 0.375 * Construction eqpt 353 0.178 * 0.678 * 0.499 * Metalworking mach. 354 0.283 * 0.417 * 0.416 * Special ind. mach. 355 0.163 0.165 0.388 General ind. mach. 356 0.251 * 0.354 * 0.471 * Offic&comp. mach. 357 0.928 * 0.186 1.74 Service ind. mach. 358 0.113 0.301 0.449 Electrical ind. apparatus 362 0.236 * 0.212 0.657 * Household appliances 363 0.258 * 0.61 * 0.602 * Lighting & wiring prod. 364 0.165 0.686 * 0.801 * TV and radio sets 365 0.483 * -0.09 -0.15 Communication eqpt. 366 0.554 * 0.831 * 1.48 * Electronic components 367 0.298 * 0.717 * 0.616 * Misc. electrical mach. 369 0.434 * 0.181 0.945 * Autos and parts 371 0.299 * 0.345 * 0.394 * Ships and boats 373 0.148 0.169 0.317 Scientific & medical eqpt. 381234 0.991 * 0.047 1.54 * Photo supplies & eqpt. 386 0.867 * -0.535 0.256 coefficient from simple regression of productivity measure against real shipments:1978-1995 * significant at 10%

-- 25 -- Table 5: Coefficient of Procyclicality--Germany productivity measures labor K&L KLM Stone, clay, glass prod. 25 0.443 * 0.386 0.221 Non-ferrous fabrications 28 0.351 * 0.348 0.081 Chemicals 40 -4.663 -2.13 -2.061 Logs and planks 53 0.486 * 0.38 -0.263 Pulp and plywood 55 0.461 0.594 0.009 Rubber products 59 0.574 * 0.514 0.579 Various steel prod. 302 0.286 * 0.279 -4.010 Metal framing & rails 31 0.363 * 0.221 0.057 Farm machinery 32 0.408 * 0.486 0.614 Autos 33 0.424 * 0.458 0.643 Electrical home appl. 36 0.148 0.397 0.404 Optics and clocks 37 0.286 * 0.399 0.452 Iron and steel sheets 38 0.237 * 0.504 0.363 Office mach. & computers 50 0.045 0.001 0.042 Musical eqpt. 39 0.625 * 0.467 0.512 Fine ceramics 51 0.198 0.48 0.519 Glassware 52 0.292 * 0.255 -0.038 Wood products 54 0.303 * 0.377 0.123 Paper products 56 0.361 * 0.682 * 0.519 Books 57 0.318 0.585 0.401 Plastic products 58 0.157 0.425 -0.078 Shoes 62 0.462 0.533 0.945 Textiles 63 0.313 * 0.319 0.161 Apparel 64 0.222 * 0.085 0.337 coefficient from simple regression of productivity measure against real gross output:1981-1994 * significant at 10%

-- 27 __ )cr E L

-- 28 -- Table 6: Productivity Measures: Correlations of Period Means Germany labor K&L KLM labor 1 K&L 0.94367 1 KLM 0.238155 0.322159 1 United States labor K&L KLM labor 1 K&L 0.95858 1 KLM 0.860876 0.875192 1

-- 29 -- Table 7: Trade Exposure--United States industries codes initialMsh initialXsh diffMsh diffXsh Basic steel products 331 8 2 18 7 Foundry products 332 37 9 29 34 Non-ferrous metal products 333 10 4 18 24 Fasteners, stampings 34567 1 1 3 3 Ordnance 348 3 7 23 46 Engines, turbines 351 2 11 25 43 Farm machinery 352 10 15 13 24 Construction eqpt 353 6 18 24 23 Metalworking mach. 354 10 9 18 11 Special industrial mach. 355 20 13 17 21 General industrial mach. 356 8 12 24 20 Office&computing mach. 357 6 14 44 30 Service industry mach. 358 2 7 12 10 Eletrical industrial apparatus 362 6 7 20 14 Household appliances 363 6 5 19 9 Lighting and wiring products 364 4 11 20 14 TV and radio sets 365 38 11 39 44 Communication eqpt. 366 11 7 16 16 Electronic components 367 11 11 40 29 Misc. electrical machinery 369 10 8 47 33 Autos and parts 371 15 6 19 12 Ships and boats 373 1 1 36 32 Scientific and medical eqpt. 381234 5 14 9 17 Photos supplies and eqpt. 386 9 9 26 14

-- 30 -- Table 8: Trade Exposure--Germany industries codes initialMsh initialXsh diffMsh diffXsh stone,clay,glass 25 11 11 3 5 non-ferrous fabrications 28 44 39 21 20 chemicals 40 23 38 12 9 logs/planks 53 34 17 6 8 pulp/plywood 55 45 29 9 15 rubber 59 21 24 11 7 various steels 302 10 23 8 5 railroad/metal framing 31 5 16 6 5 farm mach 32 17 45 11 7 autos 33 16 38 14 10 elec home appliance 36 19 31 13 6 optics/clocks 37 32 46 20 15 iron/steel sheet 38 17 25 5 9 office mach/computers 50 42 21 35 45 musical eqpt 39 52 56 15 10 fine ceramics 51 38 33 9 12 glassware 52 21 27 7 7 wood products 54 11 12 11 6 paper products 56 8 14 7 11 books 57 6 16 4 4 plastic products 58 16 25 6 6 **shoes 62 47 19 26 21 textiles 63 41 39 33 29 apparel 64 39 26 22 11

-- 31 -- Table 9: Globalization and Productivity Growth--United States regression of diff(mean(productivity) against initial(mean(Xsh), initial (mean(Msh), diff(mean(domship grow)), diff(mean(Xsh),diff(mean(Msh), diff(mean(dlxgrow), diff(mean(dlMgrow) raw:e3..e26, raw:h3..h26,diffs:m3..m26,diffs:e.3..e26,diffs:h3..h26, diffs:k3..k26, diffs:l3..l26 labor Regression Output: Constant 0 Std Err of Y Est 2.032226 R Squared 0.825153 No. of Observations 24 Degrees of Freedom 17 Xsht0 Msht0 dlDomshipdifmnXsh difmnMsh dlXgrow dlMgrow X Coefficient(s) -0.08255 0.02891 0.566972 -0.5662 0.485583 0.400074 -0.10004 Std Err of Coef. 0.078892 0.061748 0.089528 0.2119 0.161901 0.089612 0.044946 K&L Regression Output: Constant 0 Std Err of Y Est 3.16434 R Squared 0.543081 No. of Observations 24 Degrees of Freedom 17 Xsht0 Msht0 dlDomshipdifmnXsh difmnMsh dlXgrow dlMgrow X Coefficient(s) 0.136883 -0.08598 0.635796 0.328755 -0.30469 0.54832 -0.12365 Std Err of Coef. 0.122842 0.096147 0.139403 0.329945 0.252093 0.139533 0.069985 KLM Regression Output: Constant 0 Std Err of Y Est 2.668591 R Squared 0.762406 No. of Observations 24 Degrees of Freedom 17 Xsht0 Msht0 dlDomshipdifmnXsh difmnMsh dlXgrow dlMgrow X Coefficient(s) -0.1056 0.069677 0.545038 -0.554 0.600585 0.344174 -0.12539 Std Err of Coef. 0.103596 0.081084 0.117563 0.278253 0.212598 0.117673 0.059021

-- 32 -- Table 10: Globalization and Productivity Growth--Germany regression of diff(mean(productivity) against initial(mean(Xsh), initial (mean(Msh), diff(mean(domship grow)), diff(mean(Xsh),diff(mean(Msh), diff(mean(dlxgrow), diff(mean(dlMgrow) raw:g5..g28,raw:f5..f28,diffs:J5..J28,diffs:g5..g28,diffs:f5..f28,diffs:i5..i28,diffs:h5..h28 labor Regression Output: Constant 0 Std Err of Y Est 7.93466 R Squared 0.226863 No. of Observations 24 Degrees of Freedom 17 Xsht0 Msht0 dlDomshipdifmnXsh difmnMsh dlXgrow dlMgrow X Coefficient(s) -0.25443 0.222109 0.208607 -0.16158 0.092658 0.139999 0.373653 Std Err of Coef. 0.17329 0.16007 0.699699 0.845924 0.732571 1.058104 0.801301 K&L Regression Output: Constant 0 Std Err of Y Est 3.857884 R Squared 0.056735 No. of Observations 24 Degrees of Freedom 17 Xsht0 Msht0 dlDomshipdifmnXsh difmnMsh dlXgrow dlMgrow X Coefficient(s) 0.012161 0.023544 0.261696 0.104734 0.018568 0.000652 0.298692 Std Err of Coef. 0.084255 0.077827 0.340198 0.411294 0.356181 0.514457 0.389598 KLM Regression Output: Constant 0 Std Err of Y Est 15.99659 R Squared 0.247686 No. of Observations 24 Degrees of Freedom 17 Xsht0 Msht0 dlDomshipdifmnXsh difmnMsh dlXgrow dlMgrow X Coefficient(s) -0.34399 0.577237 1.445606 -2.16482 0.573453 1.868323 -2.78291 Std Err of Coef. 0.34936 0.322708 1.410621 1.705416 1.476892 2.133181 1.615455

-- 33 -- Table 11: Globalization and Procyclicality--Germany regression of the industry-cross section of procyclicality coefficients against init(Xsh), init(msh), dlXgrowth, dlMgrowth period: 1981-1994 Regression Output: labor Constant 0.773427 Std Err of Y Est 1.099999 R Squared 0.057254 No. of Observations 24 Degrees of Freedom 19 X Coefficient(s) -0.02287 0.010664 -0.06886 0.006341 Std Err of Coef. 0.022807 0.020275 0.161124 0.129864

-- 34 -- Table 12: Globalization and Procyclicality--United States regression of industry cross-section of procyclicality coefficients against initXsh,initMsh,mean(dlXgrowth), mean(dlMgrowth) period: 1979 to 1995 labor Regression Output: Constant 0.197467 Std Err of Y Est 0.233701 R Squared 0.264865 No. of Observations 24 Degrees of Freedom 19 Xsht0 Msht0 dlXgrow dlMgrow X Coefficient(s) 0.006593 -0.00185 0.018912 0.001808 Std Err of Coef. 0.011651 0.006986 0.023663 0.021455 K&L Regression Output: Constant 0.881096 Std Err of Y Est 0.341276 R Squared 0.469373 No. of Observations 24 Degrees of Freedom 19 X Coefficient(s) 0.005652 -0.03226 0.135648 -0.11692 Std Err of Coef. 0.017013 0.010202 0.034555 0.031332 KLM Regression Output: Constant -0.01344 Std Err of Y Est 0.352291 R Squared 0.592679 No. of Observations 24 Degrees of Freedom 19 X Coefficient(s) 0.032085 -0.01305 -0.00203 0.045555 Std Err of Coef. 0.017563 0.010531 0.03567 0.032343