WORKING PAPER SERIES Corporate Governance, Entrenced Labor, and Economic Growt William R. Emmons and Frank A. Scmid Working Paper 2001-023A ttp://researc.stlouisfed.org/wp/2001/2001-023.pdf November 2001 FEDERAL RESERVE BANK OF ST. LOUIS Researc Division 411 Locust Street St. Louis, MO 63102 Te views expressed are tose of te individual autors and do not necessarily reflect official positions of te Federal Reserve Bank of St. Louis, te Federal Reserve System, or te Board of Governors. Federal Reserve Bank of St. Louis Working Papers are preliminary materials circulated to stimulate discussion and critical comment. References in publications to Federal Reserve Bank of St. Louis Working Papers (oter tan an acknowledgment tat te writer as ad access to unpublised material) sould be cleared wit te autor or autors. Poto courtesy of Te Gateway Arc, St. Louis, MO. www.gatewayarc.com
CORPORATE GOVERNANCE, ENTRENCHED LABOR, AND ECONOMIC GROWTH William R. Emmons Frank A. Scmid Federal Reserve Bank of St. Louis* 411 Locust Street St. Louis, MO 63102, U.S.A. November 8, 2001 ABSTRACT Te German system of codetermination contributes to te entrencment of labor. We sow in a two-period model of project coice tat entrenced labor leads to underinvestment and overstaffing. We provide empirical evidence tat German firms subject to codetermination wit equal representation of workers on supervisory boards during 1989-93 were, on average, overstaffed. In addition, te fraction of employees in codetermined firms as decreased over time. Te expanded reac of codetermination during te mid-1970s terefore may ave contributed to te deterioration of German economic growt performance beginning at about tat time troug underinvestment, overstaffing, and costly migration of business activity away from firms subject to codetermination. Keywords: Codetermination, corporate governance, economic growt JEL classification: G32, G38 * Te views expressed are tose of te autors and do not necessarily reflect official positions of te Federal Reserve Bank of St. Louis or te Federal Reserve System.
1 CORPORATE GOVERNANCE, ENTRENCHED LABOR, AND ECONOMIC GROWTH Te German system of worker codetermination (Mitbestimmung) dates from te early 1950s, altoug its social, political, and economic roots reac far back into te 19 t century (Kommission Mitbestimmung (Codetermination Committee), 1998, p. 29). Te system of codetermination was greatly expanded during te 1970s (Kommission Mitbestimmung, 1998, p. 21). Most importantly, te 1976 Codetermination Act extended to a large number of firms te rigt of employees to old one alf of te seats on te firm s supervisory board. Te supervisory board is te external (non-executive) board tat must approve all major financing and investment decisions proposed by te (internal) management board. Codetermination is not te only form of legal protection enjoyed by German workers, but it alone provides workers wit guaranteed access to large firms supervisory boards. Altoug worker representatives on supervisory boards cannot outvote sareolder-elected board members, teir presence increases employees public visibility and, in extreme cases, allows tem to create procedural delays. For example, drawn-out consultations logrolling could stall restructuring efforts or inibit takeover negotiations. We analyze a two-period model of a firm facing an investment decision in eac period. Te context is an economy wit legally entrenced labor. Te firm faces eiter ig or low demand for its product in te first period and similarly it will face eiter ig or low demand in te future (independent of te first-period state of demand). Te firm does not know te second-period demand conditions wen deciding on te scale of first-period operations. Tat is, te firm must make an investment decision in te first period, knowing tat iring decisions are asymmetric in tat incumbent labor in te second period does not oppose adding employees but migt oppose layoffs.
2 Entrenced labor creates a old-up problem. Tat is, sareolders may suffer a loss in te second period because labor cannot be dismissed easily even toug demand for te firm s product is low. Anticipating tis, firms decisions may deviate from te first-best efficient allocation of resources in te first period. We model te old-up problem by introducing entrenced labor and uncertainty in te future demand conditions facing te firm. It turns out tat if initial demand conditions are strong, depending on te firm s tecnology and project profitability, te firm eiter underinvests in te first period or is overstaffed in te second period, on average. Tat is, in a cross section of firms, codetermination causes underinvestment and overstaffing. To investigate te model s prediction tat codetermined firms tend to be overstaffed, we examine wage and staffing levels in a large dataset covering 250 German firms during te 1989-93 period, about one alf of wic were subject to so-called 1976 codetermination (equal representation of workers on supervisory boards). We find tat average wage levels (standardized by sales per employee) are not unusual for equal-representation firms. However, staffing levels (again adjusted by sales) are indeed significantly iger on average for codetermined firms. Tis is consistent wit te model s prediction. Over time, economic activity may migrate away from firms tat face legislative constraints, suc as labor entrenced by codetermination (Jensen and Meckling, 1979). If codetermination added value to te firm, codetermination would emerge as a Coase solution in a bargaining process between sareolders and labor. Te fact tat structures like codetermination ave not been adopted by firms tat were not required to do so indicates tat its stipulations represent binding constraints on action. Te loss of efficiency arising from codetermination tat is imposed on firms presumably makes tem grow more slowly tan firms not subject to its rules, or to srink outrigt. Te Codetermination Commission found tat te codetermination-free zone of te German economy indeed as grown over time, consistent wit Jensen and
3 Meckling s prediction (Kommission Mitbestimmung, 1998, p. 27). German firms subject to codetermination also migt represent attractive takeover targets for foreign firms if, by transferring legal jurisdictions, some of te target firm s operations or decision-making could escape te reac of te codetermination laws, releasing latent economic value. Te presence of underinvestment and overstaffing problems in codetermined firms, togeter wit inefficiency-induced migration of economic activity away from firms subject to codetermination, may be sufficient to detract from overall economic performance. Te growt performance of te German economy deteriorated noticeably beginning in te 1970s in comparison wit te United States and Japan, altoug it is not obvious wat caused te deterioration. Our results suggest tat entrenced labor is a plausible candidate to explain at least part of Germany s slow economic growt in recent decades. CORPORATE GOVERNANCE IN GERMANY Several features of te German corporate world are noteworty. Like oter countries on te European continent, Germany s equity market capitalization is low relative to GDP. Bank lending is a more important source of external finance in Germany tan in some oter advanced economies. Excange-listed companies exibit igly concentrated ownersip structures and ig family ownersip (see Franks and Mayer, 2000). Worker codetermination is anoter important caracteristic of German corporate governance, giving a strong role to labor representatives in corporate decision-making. Workers in firms wit more tan five permanent employees ave te rigt to form a works council, wic as far-reacing power on sop-floor issues concerning operations, ealt and safety, to name but a few. Corporations wit more tan 500 permanent employees are subject to codetermination at te supervisory board level. One tird of te seats on te supervisory board go to labor representatives. Corporations wit more tan 2,000 employees must allocate one alf of board
4 seats to labor representatives. Te cairman, wo is generally a sareolder representative, controls a tie-breaking vote. A MODEL OF ENTRENCHED LABOR We study a representative firm s investment decisions in a two-period model tat incorporates entrenced labor. Te model draws inspiration from Hart (1995) and Myers (2000), bot of wom abstract from many features of actual firms in order to igligt critical aspects of corporate governance. Our focus in tis paper is ow te bargaining power of labor at te beginning of te second period, togeter wit uncertainty about future demand conditions, can create a old-up problem. We sow tat te old-up problem affects te firm s first-period coice of investment intensity and its second-period staffing levels. Te firm is a contract between two groups of people, te sareolders (owners) and labor (workers). We assume tat eac group acts in a coordinated manner to furter its own interests; tere are no conflicts witin eiter group. Te sareolders purcase operating assets and ire labor, creating te firm. Te firm produces a single product ( widgets, wic are te output of te firm s single project ). Te firm may operate eiter at large or small scale in eiter period by investing more or less in pysical capital and iring more or fewer workers, respectively, as described furter below. At te beginning of eac period, demand for widgets during tat period becomes known. Demand for widgets is eiter ig or low. If demand is low and te project is operated at large scale, tere is surplus output, wic cannot be sold for any positive price. If demand is ig in te first period, te sareolders pay K to buy te operating assets. If demand is low, te required initial investment is K ( K K ). Te labor input of te firm, L, is determined by a l l Leontief production tecnology, L = α K, were L is te number of employees. Te desired number of workers are offered contracts on a take-it-or-leave-it basis and all oter workers
5 remain outside te firm wit no bargaining power or influence on te firm s actions. We assume te employees work a fixed number of ours, wic, along wit te wage, w, is determined outside te model. Tese assumptions are meant to resemble Germany s collective wage bargaining arrangement (Fläcentarifvertrag), wic determines wages and standard weekly work ours. For simplicity, we assume tat te wage is fixed for two periods: w w0 w 1. Figure 1 illustrates te timing of important decisions in te model. [Figure 1] Te only uncertainty in te model is weter demand for te widget in te second period will be ig or low. If demand in te second period is ig, te firm s profit is maximized by operating at large scale, ( K = K, L= α K ), wereas if second-period demand is low, te optimal scale is small ( K = K, L= α K ). For simplicity, we assume capital does not l l depreciate. Given optimal coice of pysical capital and labor, te sareolders end-of-period return after paying wages of w α K turns out to be (1 κ ) r K, were r is te marginal opportunity cost of capital and κ is te project s value added per unit of capital, or profitability. Given ex ante uncertainty about second-period demand and te possibility of adjusting production capacity (and labor input) at tat time, tere may be surplus (net income) over wic te sareolders and labor can bargain. Tis is te essence of te old-up problem: labor may be in a position to extract some or all of te surplus of te firm at te beginning of te second period. We first discuss te case in wic sareolders possess all bargaining power and ten te case in wic labor as all te bargaining power. Our interest is in te implications of labor bargaining power for employment and investment. Sareolders possess all bargaining power in te second period We begin by assuming tat sareolders possess all bargaining power at te beginning of period two, wen uncertainty is resolved and recontracting can occur. In oter words, labor is not
6 entrenced and must accept any take-it-or-leave-it offer from sareolders. At t 0, te beginning of te first period, first-period demand for widgets is revealed to all. If demand is low, te sareolders invest te amount K l,0 to build te operating assets and ire te amount of labor determined by te Leontief tecnology, α K l,0. Tis implies a wage bill at te end of te first period equal to w α K l,0. At t 1, te beginning of te second period, te firm learns second-period demand for widgets. If demand remains low, te firm operates exactly as it did during te first period. Te operating assets are liquidated and labor retires at te end of te second period. If, on te oter and, demand is ig during te second period (after being low during te first period), sareolders invest te additional amount K additional employees numbering α ( K K ) l. K at t 1 and ire Now consider sequences in wic demand is ig in te first period. If demand is ig during te second period, as well, te firm continues operating at large scale. If demand turns out to be low during te second period, ten te firm at t 1 liquidates capital in te amount K and lays off workers numbering α ( K Kl ). Because workers ave no bargaining power, tey ave no influence on te project size cosen at te beginning of te second period or on te level of employment. Tus, tere is no pat dependency in te model because te firm s project coice at te beginning of te first period as no bearing on investment at te beginning of te second period. l K l Labor possesses all bargaining power in te second period Now suppose labor is entrenced. In our model, tis means tat workers make a take-it-or-leave-it offer to sareolders at t 1, te beginning of te second period. Te bargaining power of labor at t 1 is limited in several ways. First, sareolders generally ave te rigt to liquidate te firm at t 1 and invest te proceeds, K, at te opportunity cost (te return on
7 an outside investment) during te second period. We assume te sareolders cannot liquidate te firm at t 1 and immediately reestablis it wit a new workforce. If sareolders could undertake suc sam reorganizations, labor of course would ave no bargaining power. Second, labor s bargaining power extends only to matters tat are negotiable. We assume te wage rate and ours of work are determined outside te model troug an industry-wide collective bargaining agreement, so te only negotiable matter is second-period layoffs. Incumbent workers are represented on te supervisory board (rater tan potential new workers), so te only time labor uses its bargaining power is wen sareolders want to downsize. If demand for te widget in te second period remains uncanged (at eiter te ig or low level), te sareolders propose no canges and consequently tere are no negotiations. Similarly, wen demand increases from low to ig, labor as no incentive to resist expansion because te newly ired workforce does not affect te income of te incumbents. Bargaining occurs only wen demand drops from ig to low because firm s profit-maximizing workforce drops by te amount α ( K Kl ) to te reduced level of α K l. We assume tat labor suffers a loss if laid off; oterwise, labor as no incentive to bargain. Te loss migt be due to switcing costs or lack of alternative employment at te same wage. For instance, employed labor migt earn a premium over te alternatives of early retirement or unemployment compensation. We assume te loss suffered by labor from being laid off equals a fraction η of te current wage, w. Tus, if te workforce adjusts to te optimal level at t 1 in response to a drop in demand for te widget, te loss to labor amounts to η w α ( K K ) (in terms of t 2 values). Note tat, in a competitive labor market, η would l equal zero labor loses noting from being laid off. In competitive markets, labor would be fully insured against losses in income eiter troug unemployment insurance or, as approximated by te United States, troug a robust job market in wic laid-off workers quickly find new jobs
8 paying te worker s marginal product. Labor as no use for bargaining power in a competitive market, so it is no surprise tat te United States as neiter codetermination nor entrenced labor. Te value at t 2 of labor income lost by being laid off, η w α ( K K ), is te l maximum amount labor can credibly bargaining over. Te value at t 2 of net income (income in excess of opportunity cost of capital) to te firm is κ r K l, wic puts a limit on te amount labor can extract from sareolders. Tus, te maximum amount labor can gain from bargaining at t 1 (expressed in terms of t 2 values) is: Min{ κ r K, η w α ( K K )} l l. Note tat tis amount migt not be enoug to maintain te workforce at te ig-demand level in a second-period low-demand state. All else equal, te more profitable a project is (te iger κ is), te more likely labor can extract a sufficient amount from sareolders to keep all workers employed. At te same time, te more profitable a project is, te more likely te sareolders second-period net income will not be exausted by wage payments. In te case were labor as (some or all) bargaining power and demand is ig in te first period, running te firm at large scale during te first period is not necessarily optimal. To see tis, consider wat will appen if demand drops to te low level in te second period. Labor will use its bargaining power to extract part or all of te firm s second-period net income, κ r K l (expressed in t 2 terms). Suppose te firm is a listed stock corporation wose investors are diversified and wo make investment decisions in a risk-neutral manner. Tis means tat te sareolders simply maximize expected final (t 2 ) wealt. Let π be te probability tat demand falls to te low level during te second period. Te value added during te first period is invested outside te firm at
9 te opportunity cost, r. If sareolders coose to operate at small scale during te first period, teir expected final wealt at t 2 will be: l, EW [ ] (1 κ ) r K (1 r)+ r ( K K) (1 r) l l π [(1 κ ) (1 r) K +(1+ r) ( K K )] l l (1 π ) (1 κ ) (1 r) K, were te sareolders wealt at te outset is K. If, on te oter and, sareolders coose to operate at large scale during te first period, teir expected final wealt (at t 2 ) is:, EW [ ] (1 κ ) r K (1 r) + π [(1+ κ ) (1 r) K + (1 r) ( K K )] l l π Min{ κ r K, η w α ( K K )} l l (1 π ) (1 κ ) (1 r) K. Te difference in expected final wealt between operating te firm at small scale and large scale, respectively, is: l,, EW [ ] EW [ ] κ r ( K K) (1 r) l π Min{ κ r K, η w α ( K K )} l l Clearly, te sareolders will underinvest in te first period i.e., coose to run te project at l,, small scale in spite of ig first-period demand if (and only if) E[ W ] EW [ ]>0. On te l,, oter and, for E[ W ] EW [ ] 0, te sareolders will operate during te first period at large scale. Tey ten are faced wit te risk (wit probability π ) tat te firm will be
10 κ r Kl overstaffed during te second period by te amount Min{, ηα ( K Kl )}. Tese w considerations lead to te following two testable ypoteses: Hypotesis 1 Codetermined firms tend to underinvest. Hypotesis 2 Codetermined firms tend to be overstaffed. We provide numerical examples in te next section to illustrate te general conclusions stated ere. Te following section discusses empirical evidence tat bears on tese questions. NUMERICAL EXAMPLES Te examples illustrate te firm s coice of investment scale at t 0, te beginning of te first period, in te case of ig first-period demand; te case of low first-period demand is trivial. We assume te following values: K = 100; K = 80; r = 0.1; η = 0.2; α = 0.25; w = 1; π = 0.5. We consider in turn ig and low values for te profitability of te project: κ 0.1 or 0.3. l For low profitability (κ 0.1), te difference in expected final wealt between running te project at small scale and running it at large scale during te first period equals: l,, EW [ ]- EW [ ] = 0.22 + 0.5 Min{0.8;1} = 0.18 Clearly, te firm will coose to run te project at small scale, wic implies underinvestment in te first period by te amount K K l =20. Tere is no overstaffing during te second period, because no layoffs ever would be needed. Te loss to te sareolders (valued at t 2 ) wic also is te loss in social welfare amounts to κ r ( K K ) = 0.2. l
11 If on te oter and te profitability of te project is ig (κ 0.3), te difference in expected final wealt resulting from small-scale operation during te first period instead of large-scale operation is: l,, EW [ ] EW [ ] 0.66 0.5 Min{2.4;1} = 0.16 In tis ig-profitability case, te firm cooses to run te project at large scale during te first period despite te existence of entrenced labor and te rational expectation of a old-up problem if second-period demand is low. No underinvestment occurs, but te firm is overstaffed on average during te second period. Tere are no layoffs; te firm operates wit te maximum amount of excess labor because its operations are igly profitable. Te loss to te sareolders (valued at t 2 ) equals w α ( K K ) = 5. Tere is no loss to society, because te wealt extracted by labor comes from sareolders (i.e., it is producer surplus). l Te two foregoing examples illustrate cases of underinvestment and overstaffing, respectively. Underinvestment always creates a loss to society, wile overstaffing never does. Overstaffing is simply a wealt transfer from te sareolders to labor and as no efficiency implications. EMPIRICAL EVIDENCE Gorton and Scmid (2000) sow tat equal representation, wen compared to one-tird representation, reduces stock market (i.e., equity) value by 27 percent. Using a different metodology and a different sample, FitzRoy and Kraft (1993) found in an early study tat equal representation depresses a typical firm s value added by 19.7 percent. 1 Tese studies indicate tat equal representation places a drag on te performance of German corporations. Removing equal representation requirements presumably would add value.
12 Overstaffing Little evidence as been presented to date tat explains te reasons wy equal representation (in comparison to one-tird representation) depresses corporate performance. A 1998 report by te bipartisan Codetermination Committee (Kommission Mitbestimmung) offers insigts into ow labor uses its power, and ow tis migt affect firms performance. One of te topics igligted by te report is te so-called employment-preserving role of codetermination (Capter 6, Section 25). In order to preserve its influence on te firm, labor seeks to maintain a ig sare in total input costs. Table 1 presents regression results tat sed ligt on te impact of equal representation (compared to one-tird representation) on wages and employment at te firm level. Tese previously unpublised results use te data set from Gorton and Scmid (2000). Tis data set comprises te 250 largest traded German corporations at te end of 1993, covering te period 1989-1993. About alf te companies in te data set are subject to equal representation, wile te remainder ave one-tird representation. [Table 1] Te regression results do not support te ypotesis tat te average wage at companies wit equal representation is iger tan at companies wit one-tird representation (Table 1, column 1). Tis is in line wit te separation of codetermination and collective wage bargaining documented in Kommission Mitbestimmung (1998, capter 7, section 6). Given te absence of a significant difference in average wage levels across codetermination regimes, te 42 percent difference in te wage bill-to-sales ratio indicated in our second regression (Table 1, column 2) terefore implies tat companies wit equal representation are overstaffed. Tis 42 percent difference in te wage bill corresponds to a 33 percent iger employee count wen normalized by sales (Table 1, column 3).
13 Migration of business activity away from codetermined firms A rational response by sareolders facing entrenced labor in codetermined firms in a repeated game would be to sift business activity gradually to corporate and oter business forms subject to less intrusive legislation. As Table 2 reports, precisely tis as occurred in Germany during recent decades. Te fraction of total private-sector employment accounted for by firms subject to twin codetermination bot sop-level and supervisory board-level codetermination stood at 30.5 percent in 1984, according to te Kommission Mitbestimmung. By te mid-1990s, tis fraction ad srunk to 24.5 percent. Conversely, te fraction of total private employment in te codetermination-free zone of te economy increased from 50.6 percent to 60.5 percent. A similar pattern played out across te economy as a wole, as panel B of Table 2 confirms. [Table 2] Cross-border acquisitions are a way of watering down te effects of codetermination and tereby increasing te value of te firm. As Kommission Mitbestimmung (1998, Capter 6, Section 23, Paragrap 7) reports, coordination problems among te labor representatives of te various subsidiaries and te top tier (te parent firm) witin groups of firms (concerns) substantially weaken te power of labor. Tus, German corporations may be attractive cross-border takeover targets if for no oter reason tan te potential to release latent value tat is pent up by codetermination-driven overstaffing. Te merger of Hoecst AG of Germany and Rône-Poulenc S.A. of France into Aventis S.A. in December 1999 is a case in point. Aventis S.A. is eadquartered in Strasbourg, France. Wile te German subsidiary, Hoecst AG, is still subject to equal representation, top-tier decision-making appens at Aventis S.A., wic is out of te reac of German codetermination laws.
14 A final strand of evidence regarding te deleterious effects of entrenced labor on economic performance comes from aggregate output statistics. Wile it migt appear eroic to assert a single cause of differences in growt rates across countries, it is noteworty tat growt rates of real GDP slowed substantially in Germany after te enactment of te 1976 Codetermination Act. Figure 2 sows tat, during te period 1960-1975, real GDP in Germany and te United States grew at nearly te same average annual rate. Germany grew during tis period at an annual rate of 3.46 percent, wile te United States grew at an annual rate of 3.44 percent. During te period 1976-1989 wic ends before te German reunification in 1990 te United States grew at an annual rate of 3.08 percent wile Germany grew at an average rate of only 2.00 percent. After a temporary pickup in growt in te wake of reunification, Germany slipped back onto a growt pat tat falls sort of U.S. growt by a wide margin. [Figure 2] Taken togeter, tese disparate pieces of evidence are consistent wit te ypoteses tat codetermination increases labor entrencment; entrencment induces sareolder responses; and employer adjustments to entrenced labor may exert a significant drag on economic growt. Tese adjustments include underinvestment and avoidance of corporate forms subject to restrictive codetermination laws, bot of wic are likely to reduce an economy s growt potential. CONCLUSIONS Te German corporate governance environment is unusual in several ways. Te system of codetermination is an important fact of German corporate life tat is bot well establised and likely to elicit responses from sareolders. Our simple two-period model of investment and employment igligts te potential for entrenced labor to exert meaningful effects on firm decision-making. In particular, codetermined firms can be expected on average to underinvest
15 relative to te levels tat oterwise would be observed, and one would expect tem to be overstaffed on average. We provide new empirical evidence tat demonstrates te second of tese predictions. Te end result of entrenced labor may be to reduce te long-run growt potential of te German economy.
16. 1 Let β be te regression coefficient of a 0/1 variable, ten te cange in te dependent variable as a result of a switc of tis indicator variable from zero to one amounts to: e β 1. For details see Halvorsen and Palmquist (1980). Based on te regression coefficients presented by FitzRoy and Kraft (1993, Table 2), te aforementioned decrease of 19.7% is tus calculated as follows: e 0.13 0.06 e.
17 REFERENCES FitzRoy, Felix R., and Kornelius Kraft (1993) Economic Effects of Codetermination. Scandinavian Journal of Economics 95, 365-75. Franks, Julian, and Colin Mayer (2000), Ownersip and Control of German Corporations, Review of Financial Studies, fortcoming. Gorton, Gary, and Frank Scmid (2000) Class Struggle inside te Firm: A Study of German Codetermination. NBER Working Paper No. 7945, <ttp://papers.nber.org/papers/w7945>. Halvorsen, R. and R. Palmquist (1980) Te Interpretation of Dummy Variables in Semilogaritmic Equations, American Economic Review 70, 474-475. Jensen, Micael C., and William H. Meckling (1979) Rigts and Production Functions: An Application to Labor-Managed Firms and Codetermination, Journal of Business 52, 469-506. Kommission Mitbestimmung (1998) Mitbestimmung und neue Unternemenskulturen Bilanz und Perspektiven. Güterslo (Germany): Bertelsmann Stiftung. Myers, Stuart C. (2000) Outside Equity, Journal of Finance 55, 1005-37. United Nations (1990) International Standard Industrial Classification of All Economic Activities. Statistical Papers Series M, No. 4, Rev. 3. New York: United Nations.
18 Table 1 Employment, compensation, and equal representation. Te data set comprises te largest 250 traded German corporations by assets as of end of fiscal year 1993. Observations are from te period 1989-1993. Subsidiaries are not consolidated. Te panel is unbalanced due to missing observations. For details on te data set see Gorton and Scmid (2000). Te wage bill is in 1991 Deutsce marks and includes pension contributions. Sales are in 1991 Deutsce marks. Numbers of employees generally are as of end of fiscal year; for a few companies it is fiscal-year averages. Equal representation: Equal to 1 if tere is equal representation on te supervisory board, 0 oterwise. Firm size: (Log of) stock market capitalization (as of te end of te calendar year tat ends before te respective fiscal year). Insiders: Fraction of equity control rigts eld by management, oter employees, or families. Banks: Fraction of equity control rigts eld by domestic banks. Government: Fraction of equity control rigts eld by domestic government entities. Largest sareolder: Maximum fraction of equity control rigts eld by a single sareolder. ISIC: Industry affiliation based on International Standard Industrial Classification (United Nations, 1990) were category D (manufacturing) serves as te numeraire industry. Sample years are represented by indicator variables were 1993 serves as te numeraire year. Standard errors are corrected following Newey-West (1987); t-statistics significance levels (in two-tailed tests): * denotes 10% level, ** denotes 5% level, and *** denotes 1% level. Te effect of equal representation is te product of te regression coefficient of te interaction term Equal representation Firm size (if significant), and te median value of firm size in te sub-sample of companies wit equal representation. Dependent Variable (1) Log Ratio of Wage Bill to Number of Employees (2) Log Ratio of Wage Bill to Sales (3) Log Ratio of Number of Employees to Sales Explanatory Variable Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic Equal representation Firm size 3.772 10-3 1.030 2.039 10-2 4.495*** 1.592 10-2 3.361*** Firm size 1.078 10-1 4.543*** 1.157 10-2 0.295-4.527 10-2 -1.089 Insiders 1.765 10-2 0.146 1.782 10-1 1.238 1.837 10-1 1.289 Banks 4.919 10-2 0.223 7.639 10-1 2.135** 6.686 10-1 2.875*** Government -1.574 10-2 -0.208-4.738 10-2 -0.315-1.091 10-1 -0.652 Largest sareolder 6.925 10-2 0.598 5.646 10-2 0.356-8.581 10-2 -0.498 ISIC A -3.456 10-2 -0.480 1.889 10-1 1.758* 1.842 10-1 1.915* ISIC C 1.159 10-1 1.366 7.226 10-1 4.402*** 6.234 10-1 3.317*** ISIC E 1.658 10-1 2.659*** -3.968 10-1 -3.939*** -5.684 10-1 -5.241*** ISIC F -1.591 10-1 -1.273 6.362 10-3 0.073 6.965 10-2 0.596 ISIC G -2.425 10-1 -2.180** -6.724 10-1 -5.131*** -3.145 10-1 -2.631*** ISIC H -4.533 10-1 -5.605*** 6.518 10-1 4.944*** 1.114 9.690*** ISIC I 4.554 10-2 0.788-7.177 10-2 -0.317-1.556 10-1 -0.703 ISIC K 3.260 10-1 3.451*** -1.669 10-1 -0.406-4.593 10-1 -1.250 ISIC N -2.314 10-1 -2.891*** 5.998 10-1 5.274*** 7.200 10-1 6.438*** 1992-9.095 10-2 -2.179** -1.346 10-2 -0.176 5.580 10-2 0.729 1991-2.382 10-1 -3.712*** 3.038 10-2 0.407 1.710 10-1 2.319** 1990-3.004 10-1 -4.181*** -8.384 10-3 -0.084 1.992 10-1 1.986** 1989-3.251 10-1 -4.667*** -5.871 10-2 -0.682 1.461 10-1 1.705* Constant 2.308 5.091*** -1.939-2.580*** -5.112-6.402*** F-statistic 5.740*** 7.987*** 7.568*** R 2 adj. 0.093 0.139 0.134 Effect of equal representation --- 0.420 0.328 Number of observations 858 798 783
19 Table 2 Fraction of employees by codetermination type. Employees fall into te category of so-called twin codetermination if teir employers ave bot works councils (codetermination on te sop-floor level) and equal representation on te supervisory board. Equal representation on te supervisory board may result from te 1951 Montan Codetermination Act or te 1976 Codetermination Act. Employees are assigned to te so-called single codetermination regime if teir employers ave works councils but are not subject to equal representation on te supervisory board. Companies witout equal representation on te supervisory board may be subject to 1/3 codetermination according to te 1976 Codetermination Act or ave no labor representation on te supervisory board. Employees are assigned to a regime of no codetermination if teir employers ave neiter works councils nor equal representation on te supervisory board. Te public sector generally as representation on te sop-floor level, but supervisory boards do not exist. Media companies and many nonprofit organizations are exempt from codetermination due to te constitutional freedoms of expression and fait. Te terms twin and single codetermination were taken from Kommission Mitbestimmung (1998), wo compiled te numbers. Kommission Mitbestimmung calculated te numbers from data tat originate from multiple sources and years, wic explains te coice of te reference year 1994/96. Percentage Fractions of Employees by Codetermination Type Panel A: Private Sector Codetermination Type 1984 1994/96 Twin Codetermination 30.5 24.5 Single Codetermination 18.9 15.0 No Codetermination 50.6 60.5 Total (Percent) 100 100 Panel B: Wole Economy (Private, Public, and Nonprofit Sectors) Codetermination Type 1984 1994/96 Twin Codetermination 22.2 18.2 Single Codetermination 40.8 36.9 No Codetermination 37.0 44.9 Total (Percent) 100 100
20 Figure 1 Timeline of Firm Decision-Making t 0 t 1 t 2 Project launc Expansion? Partial liquidation? Liquidation
21 Figure 2 Real Growt: Germany and te United States in Comparison 1000 100 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 Germany United States Real GDP: annual observations; United States: GDP in billions of cained 1996 US$; Germany: GDP at constant prices (1995=100); rebased to 100 in 1960; Source: IMF.