2.1 – Historical Context

13Nowadays, virtually every government in the world has a system in place to honour their citizens for various achievements. For example, France has two national and four ministerial orders of merit, Japan awards 7 different types of state orders, and the British honours system includes 10 orders of chivalry. Since these awards carry no economic value they fall into the category of positional goods, a term introduced by Hirsch (1976) to indicate non-monetary products and services whose value is a function of their desirability by others. While the norm is that these awards are bestowed by the ruling body or government of the nation, they may also be bestowed by federated state governments, which is for example the case in Austria, Canada, and Germany. Germany is a federal republic consisting of 16 constituent states which all retain a measure of sovereignty. Shortly after their creation in 1949, both the Federal Republic of Germany and the German Democratic Republic introduced their own national systems of government awards for both civilian and military services alike. The state order of merit in the Federal Republic–the Bundesverdienstorden, introduced in 1951–was soon complemented by orders of merit introduced by the various constituent states. The latter are awarded by the respective Minister-President, the elected head of government of each state which corresponds to the governor of a U.S. state.

14Table 1 describes the state government award systems in the 16 federated states. Column (2) presents the year in which each state introduced their order of merit. Bavaria was the first state in the Federal Republic to introduce its own award in 1957. Lower Saxony did the same in 1961. In the 1970s and 1980s, Baden-Württemberg, Saarland, Rhineland-Palatinate, Berlin, Hesse, and North Rhine-Westphalia followed. With German reunification in 1990 the award system of the Democratic Republic was discontinued. All the new federal states created their own orders of merit during the 1990s (Saxony) and 2000s (Brandenburg, Mecklenburg-Vorpommern, Saxony-Anhalt, Thuringa). Also one of the old federal states, Schleswig-Holstein, introduced an order of merit in the 2000s. All but two German states have a state system of state government awards in place. The two exceptions are the old Hansaeatic states, Bremen and Hamburg, who by law reject any government awards as such wearable insignia would indicate a ranking of people in a way not compatible with the bourgeois spirit of the constitution of these states.

Table 1

Overview of the Federated State Orders of Merit in Germany

(1)State (2) Year (3) Award (4) Award criteria Panel A. States with state award before reunification Bavaria 1957 Bayerische Verdienstorden Recognition of outstanding contributions to the Free State of Bavaria and its people Lower Saxony 1961 Niedersächsischer Verdienstorden Outstanding contributions to the state Baden-Württemberg 1974 Verdienstorden des Landes Baden-Württemberg For outstanding contributions to the state in politics, society, culture, and economics Saarland 1974 Saarlandischer Verdienstorden Awarded for outstanding service to Saarland Rhineland-Palatinate 1981 Verdienstorden des Landes Rheinland-Pfalz For outstanding service to the state and people of Rhineland-Palatinate North Rhine-Westphalia 1986 Verdienstorden des Landes Nordrhein-Westfalen For extraordinary contributions to the people and state Berlin 1987 Verdienstorden des Landes Berlin Recognition of outstanding contributions to the State of Berlin Hesse 1989 Hessischer Verdienstorden For outstanding contributions to the state of Hesse Panel B. States that introduced state award after reunification Saxony 1997 Sächsischer Verdienstorden For outstanding contributions to the people and state of Saxony Thuringia 2000 Verdienstorden des Freistaats Thüringen Awarded as a sign of appreciation for outstanding services to the free state and its population Mecklenburg-Vorpommern 2001 Verdienstorden des Landes Mecklenburg-Vorpommern For exceptional performance over a long period of time, or an extraordinary individual performance for the benefit of the state Brandenburg 2003 Verdienstorden des Landes Brandenburg For extraordinary services to the State of Brandenburg and its people Saxony-Anhalt 2006 Verdienstorden des Landes Sachsen-Anhalt For exceptional performance over a longer period of time or an extraordinary individual performance for Saxony-Anhalt and its citizens Schleswig-Holstein 2008 Verdienstorden des Landes Schleswig-Holstein A special gesture of the state to recognize outstanding achievements Panel C. States that have not introduced state awards Bremen Na Na Na Hamburg Na Na Na

Overview of the Federated State Orders of Merit in Germany

15The awards all have in common that they are wearable both to formal attire, in the form of crosses worn around the neck or on the breast, and for everyday wear in the form of lapel pins. Another general – and for identification purposes, important – feature is that the award criteria, which are synthesized in the column (4), are virtually identical in that they all emphasize contributions to the respective state and its people.

2.2 – Data

16Due to restrictions in data availability I limit the study to the period after German reunification and focus on the states that have introduced state orders of merit since then. Besides the five former states of the German Democratic Republic – Sachsen, Thuringen, Mecklenburg-Vorpommern, Brandenburg, Sachsen-Anhalt – who all introduced state orders of merit after they ascended into the Federal Republic, the set of states includes also Schleswig-Holstein. I gather, by year and state, the names of all individuals who have been awarded a state order of these states from the webpages of the state governments. [2] Typically, the citation for an award is accompanied by a brief overview of the life of the awarded person. If not, I conduct google and Factiva searches for that person. In this manner I am able to single out the present or former CEOs from the list of awardees. I retain information on the company names and the time period of their tenure as CEOs. Next, I obtain financial data at firm level. My starting point is the Worldscope list for Germany (WSCOPEBD), available through Datastream. I retain only those firms that have their headquarters in one of the six states. As the firms of some of the honoured CEOs are not publicly traded I manually augment the list with information from annual reports available through Orbis, from company web pages, and from two printed databases: Medium Companies of Europe 1992/93 Volume 1 Medium Companies of the Continental European Community; Medium Companies of Europe 1993/94 Volume 1 Medium Companies of the Continental European Community. Table 2 provides descriptive statistics of the entire sample.

Table 2

Descriptive Statistics

(1) Awarded (2) Non-awarded (3) t-test of difference in means # CEO awarded 67 0 – Firm-year observations 132 2,342 – Operating margin (%) 7.1 6.9 0.68 ROA (%) 1.9 2.4 –0.75 ROE (%) 2.3 1.8 –0.59 # Employees 2,044 831 2.78*** Total assets (mEUR) 863 157 2.97***

Descriptive Statistics

This table presents in Column 1 descriptive statistics for firms where the CEO was awarded a state order of merit in one of the German federated states: Brandenburg, Mecklenburg-Vorpommern, Sachsen, Sachsen-Anhalt, Schleswig-Holstein, Thuringen. Column 2 presents descriptive statistics for firms where the CEO was not awarded a state order of merit by any of these states. All firms are head-quartered in one of the six states. Operating margin is measured as earnings before interest and taxes over sales; return on assets (ROA) is measured as net income over the book value of assets; return on equity (ROE) is measured as net income over the book value of equity. The sample runs from 1991 to 2015.

17The sample period runs from 1991 – the first year after Germany reunification took place on October 3, 1990 – to 2015. In total, 67 CEOs have been awarded one of the six state orders of merit instituted in this period. Of the total number of 2,474 firm-year observations, 132 (2,342) relate to observations where the CEO has been (has not been) awarded a state order. The relatively low number of awarded CEOs in the sample is expected since government awards by construction must be scarce in supply to fulfill their role as positional goods.

18Throughout the analysis, all continuous variables are winsorized at the 1% and 99% levels to minimize the impact of outliers. Corporate performance is expressed through three variables: Operating margin is measured as earnings before interest and taxes (EBIT) over sales; return on assets (ROA) measured as net income over the book value of assets; return on equity (ROE) measured as net income over the book value of equity. The means for operating margin, ROA, ROE, the number of employees, and total assets in column 1 (2) are the averages of all firm-year observations where the CEO has been (has not been) award a state order of merit. We note that there are no significant differences in terms of average performance between the two categories though firms where the CEO has been awarded are larger both in terms of the number of employees and in terms of total assets.

3.1 – Ex ante Effects of Government Awards

20The most useful feature of the evolution of state awards in Germany from a research standpoint is that the German states introduced the award systems at different times. Consequently, cross-sectional and time series variation in states’ introduction of award systems permit the use of strong econometric techniques applied to panel data sets. I employ these techniques to reduce the degree of omitted variables in my model of the determinants of corporate performance and employment. First, I use the introduction of orders of merit as identification to determine empirically whether CEOs change their behaviour once a state award system is introduced. Thus, I use the dates in column 1 of Table 1 to construct a yearly indicator variable equal to one for the firms that are headquartered in a state with an award system and zero for firms that are headquartered in a state where there is not yet an award system. To see if there is a general effect on all firms (in terms of profitability and employment) when a state award is introduced I estimate a fixed-effects model:

21

22where Y_t,i a is a vector of the following outcome variables during year t in firm i; operating margin measured as EBIT over sales; ROA measured as net income over the book value of assets; ROE measured as net income over the book value of equity; employment measured as the natural logarithm of the number of employees. The year-specific dummy variables a_t measure the common, economy wide shocks at time t. Thus, a_t controls for nation-wide shocks and trends that shape the dependent variables over time in Germany, such as business cycles, national changes in regulations and laws, long-term trends in dependent variables, etcetera. The firm-specific dummy variables b_i measure the firm-specific components. Thus, b_i controls for time-invariant differences in the dependent variables due to unexplained factors that differ across the firms i. The industry-specific dummy variables c_j control for differences across various industries j. By including these fixed effects in the model, all of the cross-sectional variation gets removed; coefficients are driven by changes in variables after a firm experiences the introduction of an award in its state. I_t,i is an indicator variable that takes the value of one for each year that the state in which a firm is headquartered has a state award system in place, and zero otherwise. Thus, the coefficient d measures the change in the outcome variables stemming from the introduction of awards. This is the key coefficient of interest since it gives us a difference-in-differences estimator that shows the changes in the dependent variable following the introduction of state government awards in the treatment state, controlling for changes in the dependent variable among the control group of states that have, at the time, not introduced a state award system. As such, the specification is a generalization of the difference-in-differences approach where the control group is constructed from the average of all firms in the sample, rather than from a different set of firms not ever experiencing any change in the existence of orders of merit.

23Finally, x is a control variable for firm size, measured as the natural logarithm of assets. Size is a particularly important control variable as it captures many firm characteristics such as executive compensation (Jensen and Murphy, 1990; Core et al., 1999; Baker and Hall, 2004; Gabaix and Landier, 2008; Gabaix et al., 2014), quality of managers (Himmelberg et al., 1999), capital structure (Titman and Wessels, 1988; Rajan and Zingales, 1995; Booth et al., 2001; Fama and French, 2002; Frank and Goyal, 2003) and it may be a proxy for the probability of default (Shumway, 2001) and the volatility of firm assets (Fama and French, 2002). Heteroscedastic robust standard errors are clustered by state to account for possibility of within-state interdependence in the residuals due to serial correlation (Bertrand et al., 2004).

24The results of model 1 are presented in Table 3. The first three columns present results for the profitability outcome variables and the last column presents results for the employment variable. We see that the difference-in-differences indicator coefficient is negative and statistically significant at the 10% level in each of the three profitability specifications. The point estimates are moderately large, indicating that profitability decrease, depending on specification used, by between 0.5 to 1.2 percentage points following the introduction of state orders of merit. By contrast, the difference-in-differences indicator coefficient for the fourth outcome variable, firm-level employment, is positive with a coefficient of 3.5 percentage points that is statistically significant at the 5% level.

Table 3

Effects on Profitability and Employment from State Awards Introduction

(1) Operating margin (2) ROA (3) ROE (4) Employees State award indicator –0.012* –0.002* –0.005* 0.035** (0.005) (0.001) (0.002) (0.011) Size –0.004 0.008** –0.057 0.642*** (0.003) (0.003) (0.067) (0.050) Constant 0.052 –0.018 0.017* –1.007* (0.043) (0.062) (0.009) (0.444) Firm fixed effects Yes Yes Yes No Year fixed effects Yes Yes Yes Yes Industry fixed effects No No No Yes Observations 2,474 2,474 2,474 2,181 R-Squared within 0.155 0.167 0.168 0.215

Effects on Profitability and Employment from State Awards Introduction

This table presents estimates of the change in outcome variables following introduction of state award systems in the German federated states: Brandenburg, Mecklenburg-Vorpommern, Sachsen, Sachsen-Anhalt, Schleswig-Holstein, Thuringen. State award indicator is an indicator that takes the value of one for each year that the state in which a firm is headquartered has a state award system in place. Operating margin, ROA, and ROE are defined as in Table 2. Employees and Size are measured through the natural logarithm of the number of employees and total assets, respectively. The sample runs from 1991 to 2015. Robust standard errors in parenthesis are clustered by state. *** Significant at the 1% level; ** Significant at the 5% level; * Significant at the 10% level.

25Overall, the results of Table 3 suggest that, on an aggregate level, the introduction of state orders of merit decreases corporate profitability but increases employment. Thus, the mere introduction of awards can affect corporate behaviour. These aggregate results naturally lead to the question if the individual CEOs that will eventually be awarded behave differently than those CEOs who ultimately do not get the award? To investigate this question, I estimate the following fixed-effects model:

26

27Where, Y_t,i equals the same outcome variables as in regression (1), I_t,i is as before an indicator variable that takes the value of one for each year that the state in which a firm is headquartered has a state award system in place, and zero otherwise, x controls for firm size, and a_t, b_i, and c_j again represent year, firm, and industry fixed effects, respectively. L_t,i is an award indicator equal to one from the year of the introduction of the order of merit in the state where the firm is headquartered up to the year in which the incumbent CEO receives the honour. If a CEO is awarded after their retirement, the indicator variable L_t,i takes the value one for the years of the CEO’s tenure that overlap with the years in which there exists an order of merit in the state where the firm is headquartered. Thus, L_t,i is zero for all observations that relate to: A year in which no state award exists for the state in which a firm is incorporated; all years after the incumbent CEO was awarded; all incumbent CEOs that never receive an award. [3] Subsequently, g measures the relative ex ante behaviour in terms of profitability and number of employees of the firms whose CEOs will be awarded compared to the CEOs that are not awarded. As before, heteroscedastic robust standard errors are clustered by state. Table 4 presents the results.

Table 4

Ex ante Effects on Profitability and Employment from a State Award

(1) Operating margin (2) ROA (3) ROE (4) Employees Ex ante state award indicator –0.003 0.002 0.004 0.072** (0.023) (0.002) (0.005) (0.024) State award indicator –0.012* –0.002* –0.005* 0.033** (0.006) (0.001) (0.002) (0.011) Size –0.002 0.008*** –0.032 0.632*** (0.002) (0.002) (0.034) (0.047) Constant 0.008 0.002 0.013 –0.970* (0.016) (0.052) (0.011) (0.420) Firm fixed effects Yes Yes Yes No Year fixed effects Yes Yes Yes Yes Industry fixed effects No No No Yes Observations 2,474 2,474 2,474 2,181 R-Squared within 0.147 0.150 0.159 0.207

Ex ante Effects on Profitability and Employment from a State Award

This table presents estimates of the change in outcome variables prior to the award to a CEO of a state order from one of the following German federated states: Brandenburg, Mecklenburg-Vorpommern, Sachsen, Sachsen-Anhalt, Schleswig-Holstein, Thuringen. Ex ante state award indicator is an award indicator equal to one for firms where the CEO will receive an award. It takes the value of one from the year of the introduction of the state award in the state where the firm is headquartered, up to the year in which the incumbent CEO receives the honour. State award indicator is an indicator that takes the value of one for each year that the state in which a firm is headquartered has a state award system in place. Operating margin, ROA, and ROE are defined as in Table 2. Employees and Size are measured through the natural logarithm of the number of employees and total assets, respectively. The sample runs from 1991 to 2015. Robust standard errors in parenthesis are clustered by state. *** Significant at the 1% level; ** Significant at the 5% level; * Significant at the 10% level.

28The only case for which the difference-in-differences estimator g is statistically significant is for the level of employees. The firms of CEOs who will be awarded a state award have on average a higher level of employees by 7.2 percentage points, after controlling for size and fixed effects. Thus, the firms that are ultimately awarded do not reduce their profits more than those CEOs who are not awarded but they do uphold a higher level of employees, which constitutes further evidence that a government may use awards as an incentive device to spur employment in the private sector.

3.2 – Ex post Effects of Government Awards

29The previous section has shown the superior ex ante performance of CEOs before they receive an award. The question then remains if the ex post behaviour changes once a CEO actually gets rewarded. To investigate this question, I estimate the following fixed-effects model:

30

31Where, Y_t,i equals the same outcome variables as in regressions (1) and (2), I_t,i is again an indicator variable that takes the value of one for each year that the state in which a firm is headquartered has a state award system in place, and zero otherwise, x controls for firm size, and a_t, b_i, and c_j respectively represent year, firm, and industry fixed effects. K_t,i is an indicator variable equal to one for each year that follows upon the award of a state order to the incumbent CEO until the tenure as CEO ends, or the sample period ends. K_t,i is zero for all other observations. Thus, h measures the ex post behaviour in terms of profitability and number of employees of the firms whose CEOs have been awarded. Heteroscedastic robust standard errors are clustered by state. The results are presented in Table 5.

Table 5

Ex post Effects on Profitability and Employment from a State Award

(1) Operating margin (2) ROA (3) ROE (4) Employees Ex post state award indicator –0.003 0.006 0.010 –0.012 (0.004) (0.005) (0.011) (0.015) State award indicator –0.011* –0.002* –0.004* 0.034** (0.005) (0.001) (0.002) (0.011) Size –0.002 0.009*** –0.023 0.490*** (0.002) (0.002) (0.041) (0.051) Constant 0.008 0.002 0.011 –1.050* (0.017) (0.050) (0.012) (0.465) Firm fixed effects Yes Yes Yes No Year fixed effects Yes Yes Yes Yes Industry fixed effects No No No Yes Observations 2,474 2,474 2,474 2,181 R-Squared within 0.123 0.133 0.166 0.203

Ex post Effects on Profitability and Employment from a State Award

This table presents estimates of the change in outcome variables following the award to a CEO of a state order from one of the following German federated states: Brandenburg, Mecklenburg-Vorpommern, Sachsen, Sachsen-Anhalt, Schleswig-Holstein, Thuringen. Ex post state award indicator is an award indicator equal to one for firms where the CEO has received an award. It takes the value of one from the year after the bestowal of the award up to the year in which the incumbent CEO leaves or the sample period ends. State award indicator is an indicator that takes the value of one for each year that the state in which a firm is headquartered has a state award system in place. Operating margin, ROA, and ROE are defined as in Table 2. Employees and Size are measured through the natural logarithm of the number of employees and total assets, respectively. The sample runs from 1991 to 2015. Robust standard errors in parenthesis are clustered by state. *** Significant at the 1% level; ** Significant at the 5% level; * Significant at the 10% level.

32We see that the difference-in-difference indicator variable h is insignificant for all of the four outcome variables, which is evidence that a state award does not trigger an ex post change in CEO behaviour. This result is in contrast with the findings of Malmendier and Tate (2009) that CEOs significantly underperform after receiving major business awards. Why then do we observe this difference in ex post behaviour between government awards and business press awards? Though more research would be needed, it is possible that the answer lies in the unique features attached to state awards compared to other awards. A government award can namely be revoked if it is later deemed that the honoured CEO has brought the award in disrepute. For example, in the United Kingdom there are numerous cases of CEOs who have been stripped of their knighthoods after their companies have failed (Hillman, 2015). It is possible that the revocability of government awards spurs the CEO to perform well – or at least not to underperform – also after the honour has been bestowed. Essentially the state award remains an incentive also after it has been bestowed.

4.1 – Dynamics of Government Award Introductions on Profitability and Employment

33Table 3 demonstrated that firm performance (firm level employment) decreased (increased) somewhat following episodes of orders of merit introductions. In this section I will address two related questions. The first is whether the documented impact was temporary or long-term. Second, I will explore the dynamics of changes in the outcome variables before and after the state awards introductions to ensure that the declining profitability and increasing levels of employment did not precede the creation of state orders of merit.

34To clarify the timing of the impact of the introductions I examine the dynamics of the relationship between the years of introduction and outcome variables by augmenting regression (1) with a series of time dummy variables. The purpose is to trace out the year-by-year effects of state awards introduction. The regression follows:

35

36where Y_t,i equals the four outcome variables from regression (1) during year t in firm i. I_i^-n equals one for firms in the n^th year before a state award is introduced in the state where the firm is headquartered and I_i⁺ⁿ equals one for firms in the n^th year after introduction of an order of merit in the state where the firm is headquartered. Furthermore, I_i^–3 equals one for all years that are three or more years before awards introduction and I_i⁺³ equals one for all years that are three or more years after awards introduction. For all other cases, including the year of introduction, the I’s equal zero. I thereby isolate the dynamic effect of introductions on the four outcome variables relative to the year of introduction. The vectors a_t, b_i, and c_j are vectors of year, firm, and industry dummy variables, respectively. Standard errors are clustered by state.

37Figure 1, plots a (b) [c] {d} show the year by year estimates of awards introduction on Operating margin (ROA) [ROE] {Employees} and the 95% confidence intervals. The figure illustrates two key points. First, changes in the various outcome variables did not precede awards introduction. Of the 12 pre-introduction year estimates only one is statistically significant, ROE for the third year before the introduction. Second, the impact becomes significant at 5% shortly after the awards introduction for all four outcome variables, and stays significant for several years. Though effects are statistically significant already the first year after the introductions, they appear to increase gradually year-by-year to the third post-introduction year.

Figure 1

Dynamics of State Award Introductions and Outcome Variables

Dynamics of State Award Introductions and Outcome Variables

This figure plots the year by year estimates (solid lines) and 95% confidence intervals (dotted lines) obtained from regression (5) on the dynamics of awards introduction with respect to: Operating margin in plot (a); ROA in plot (b); ROE in plot (c); Employees in plot (d).

38In sum, neither changes in firm profitability nor changes in firm level employment precede awards introduction. By contrast, profitability is reduced and employment is increased following these introductions.

4.2 – Placebo Tests

39All of the results presented in Tables 3 to 5 are, of course, also subject to the criticism that I may have omitted an important variable linked to one of the outcome variables. While the fixed effects approach is not very vulnerable to this problem, my approach could be biased if some of the six states in my sample experienced radical policy changes around the same time that the state introduced the order of merit. I use placebo tests to mitigate concerns that the results are spuriously caused by other events coinciding with the introduction of awards and the years of awarding the CEOs. In such tests we should expect to find no effects of the various award indicator variables used. Table 6 depicts the result from tests where the events studied in Tables 3 to 5 are assigned to placebo dates occurring prior and after the actual event years.

Table 6

Placebo Tests

(1) Operating margin (2) ROA (3) ROE (4) Employees Panel A Placebo state award indicator –0.011 0.002 –0.002 0.017 (0.009) (0.003) (0.002) (0.023) Size –0.005 0.007** –0.055 0.636*** (0.006) (0.003) (0.070) (0.060) Constant 0.050 –0.019 0.016* –1.000* (0.044) (0.066) (0.008) (0.448) Firm fixed effects Yes Yes Yes No Year fixed effects Yes Yes Yes Yes Industry fixed effects No No No Yes Observations 2,474 2,474 2,474 2,181 R-Squared within 0.158 0.166 0.168 0.221 Panel B Placebo ex ante state award indicator –0.003 0.004 0.004 0.023 (0.025) (0.005) (0.004) (0.032) Placebo state award indicator –0.012 0.002 –0.003 0.017 (0.010) (0.003) (0.003) (0.022) Size –0.002 0.008*** –0.031 0.610*** (0.002) (0.002) (0.038) (0.089) Constant 0.008 0.003 0.013 –1.022* (0.016) (0.052) (0.012) (0.450) Firm fixed effects Yes Yes Yes No Year fixed effects Yes Yes Yes Yes Industry fixed effects No No No Yes Observations 2,474 2,474 2,474 2,181 R-Squared within 0.152 0.167 0.165 0.216 Panel C Placebo ex post state award indicator –0.003 0.005 0.009 –0.016 (0.003) (0.004) (0.011) (0.014) Placebo state award indicator –0.011 –0.002 –0.002 0.018 (0.009) (0.003) (0.002) (0.023) Size –0.003 0.010*** –0.028 0.619*** (0.002) (0.002) (0.034) (0.052) Constant 0.009 0.004 0.011 –0.998* (0.019) (0.051) (0.012) (0.443) Firm fixed effects Yes Yes Yes No Year fixed effects Yes Yes Yes Yes Industry fixed effects No No No Yes Observations 2,474 2,474 2,474 2,181 R-Squared within 0.150 0.171 0.160 0.218

Placebo Tests

Panel A (Panel B) [Panel C] presents placebo estimates of the change in outcome variables obtained in Table 3 (Table 4) [Table 5]. Placebo state award indicator equals one for the firms that are headquartered in a state with an award system two years before the actual introduction occurs. Placebo ex ante state award indicator is equal to one from two years before the introduction of the state award up to two years before the year in which the incumbent CEO receives the honour. Placebo ex post state award indicator equals one for all years that follow two years after the award of a state order to the incumbent CEO. Operating margin, ROA, and ROE are defined as in Table 2. Employees and Size are measured through the natural logarithm of the number of employees and total assets, respectively. The sample runs from 1991 to 2015. Robust standard errors in parenthesis are clustered by state. *** Significant at the 1% level; ** Significant at the 5% level; * Significant at the 10% level.

40Panel A presents the results of regression (1) but where the award indicator, I_t,i, now takes the value of one already from two years before the state in which a firm is headquartered has a state award system in place. Panel B shows the results from regression (2) where the award indicator L_t,i is equal to one from two years before the introduction of the state award up to two years before the year in which the incumbent CEO receives the honour. Panel C presents the results of regression (3) where the indicator variable K_t,i now equals one for all years that are more than two years after the award of a state order to the incumbent CEO.

41Across the panels I find that the indicator variables for performance and employment are all statistically insignificant at conventional levels, which suggests that the previously presented generalized difference-in-differences estimates do reflect the impact of state awards, and not simply differential trends in the various treatment and control groups caused by spurious events.

4.3 – Accounting for Heterogeneity Across States

42The models described in regressions (1) to (3) have a number of benefits in terms of controlling for differences across the physical location of the firms. First, the firm and industry fixed effects control for time-invariant differences in the outcome variables due to unexplained factors that differ across firms and industries. Second, the time fixed effects control for the German business cycle. However, it is possible that the business cycles differ across the different states. Five of the six states in my sample ascended to the Federal republic after the German Democratic Republic was dissolved. Compared to the sixth state, Schleswig-Holstein, the economy of these states has been characterized by higher unemployment and larger state subsidies for investments following the reunification (Statistische Ämter des Bundes und der Länder, 2015). Regional business cycles may thus introduce a bias into the results in Tables 3 to 5. I therefore re-estimate the analysis using an augmented version of the models in regressions (1) to (3) allowing the time effects, which are essentially business cycle effects, to vary across the states. In other words, I replace the year fixed effects in those regressions, a_t, with year-state fixed effects, which controls for business cycle effects in a state at time t. The benefit of this approach is that it reduces the likelihood that the estimates will be biased by a correlation between state cycles and award introductions. The cost comes in terms of a large number of lost degrees of freedom. The results are presented in Table 7.

Table 7

Heterogeneity Across States

(1) Operating margin (2) ROA (3) ROE (4) Employees Panel A State award indicator –0.012* –0.001 –0.004* 0.030** (0.006) (0.001) (0.002) (0.010) Size –0.004 0.007** –0.060 0.666*** (0.004) (0.003) (0.060) (0.052) Constant 0.050 –0.021 0.022** –1.000* (0.040) (0.060) (0.008) (0.437) Firm fixed effects Yes Yes Yes No Year-state fixed effects Yes Yes Yes Yes Industry fixed effects No No No Yes Observations 2,474 2,474 2,474 2,181 R-Squared within 0.212 0.199 0.201 0.261 Panel B Ex ante state award indicator –0.002 0.002 0.004 0.069** (0.022) (0.003) (0.005) (0.023) State award indicator –0.012* 0.000 –0.004* 0.029** (0.006) (0.001) (0.002) (0.009) Size –0.003 0.007*** –0.030 0.500*** (0.002) (0.002) (0.030) (0.055) Constant 0.010 0.002 0.014 –0.997 (0.014) (0.050) (0.010) (0.601) Firm fixed effects Yes Yes Yes No Year-state fixed effects Yes Yes Yes Yes Industry fixed effects No No No Yes Observations 2,474 2,474 2,474 2,181 R-Squared within 0.208 0.203 0.194 0.266 Panel C Ex post state award indicator –0.002 0.006 0.011 –0.012 (0.002) (0.006) (0.011) (0.015) State award indicator –0.011* 0.000 –0.005* 0.030** (0.006) (0.000) (0.002) (0.010) Size 0.001 0.008*** –0.022 0.488*** (0.001) (0.002) (0.040) (0.052) Constant 0.006 0.002 0.012 –1.001 (0.010) (0.048) (0.012) (0.755) Firm fixed effects Yes Yes Yes No Year-state fixed effects Yes Yes Yes Yes Industry fixed effects No No No Yes Observations 2,474 2,474 2,474 2,181 R-Squared within 0.205 0.203 0.200 0.259

Heterogeneity Across States

This table presents augmented estimates of the change in outcome variables obtained in Tables 3 to 5 where the year fixed effects are replaced by year-state fixed effects. State award indicator (Ex ante state award indicator) [Ex post state award indicator] are defined as in Table 3 (Table 4) [Table 5]. Operating margin, ROA, and ROE are defined as in Table 2. Employees and Size are measured through the natural logarithm of the number of employees and total assets, respectively. The sample runs from 1991 to 2015. Robust standard errors in parenthesis are clustered by state. *** Significant at the 1% level; ** Significant at the 5% level; * Significant at the 10% level.

43We see from Table 7 that the replacement of year fixed effects to year-state fixed effects leads to one significant change. In Panel A, ROA, which was significant at the 10% level in Table 3 is no longer significant. All other difference-in-differences coefficients retain their significance, although with, generally, a slightly lower economic magnitude. Overall, the results of Table 7 suggest that the effects of government awards on corporate performance and employment cannot be explained away by the existence of regional business cycles.