Minimum Wage and Corporate Policy

Transcription

1 Minimum Wage and Corporate Policy Matthew Gustafson and Jason Kotter February 2018 Abstract Using cross-state and intertemporal variation in whether a state s minimum wage is bound by the federal minimum wage, we find that minimum wage increases lead to reduced corporate investment across a wide range of labor-intensive industries. The effect is largest for capital investments (i.e., capital expenditures and acquisitions of non-labor-intensive firms) and firms in minimum wage sensitive industries (i.e., restaurant, retail, and entertainment), and is accompanied by a reduction in both debt and equity financing. These findings suggest that a scale effect whereby minimum wage hikes reduce optimal production is the predominant channel through which minimum wage affects corporate investment. Matthew Gustafson (mtg15@psu.edu) and Jason Kotter (jasonkotter@psu.edu) are at the Smeal College of Business at the Pennsylvania State University. We thank Ashwini Agrawal, Jess Cornaggia, Cesare Fracassi, Dirk Jenter, Michael Reich, Matthew Serfling, Isaac Sorkin, Jeremy West, participants at the 2017 LBS Summer Finance Symposium, and seminar participants at Brigham Young University, Pennsylvania State University, the University of Utah, and Utah State University.

2 1. Introduction In 2016, California, New York, and Washington D.C. each passed laws that will raise their minimum wage rate to $15 per hour, more than double the current federal minimum wage of $7.25. The real minimum wage is now at a historical high in these states, and in the past two years over twenty other states have followed suit and increased their minimum wages. Yet, there is little evidence on whether and how minimum wages affect corporate policies. We motivate and empirically test two non-mutually exclusive hypotheses regarding how minimum wage increases affect corporate investment. The Substitution Hypothesis predicts that minimum wage increases will lead to increased investment as firms seek to become less reliant on labor. Stigler (1946) posits such an effect, which is consistent with existing evidence that minimum wage increases have little or no effect on the employment of minimum wage workers in the shortrun (Card and Krueger, 2015), but do lead to less minimum wage employment in the long-run (Meer and West, 2015; Sorkin 2015). By contrast, the Scale Hypothesis predicts that minimum wage increases will result in less investment because increased factor prices reduce optimal production. Traditional neoclassical, cost-of-adjustment, and q-theory models of investment support this idea, predicting a negative relation between input costs and capital investment. 1 To empirically identify which of these hypotheses dominates on average, we exploit intertemporal and cross-state variation in minimum wages. To control for the fact that states adjust minimum wages at non-random times, we identify exclusively off of state-level minimum wage changes that are mandated by the federal government. Specifically, we compare how corporate investment differentially changes surrounding federal minimum wage increases in states with a minimum wage equal to the federal minimum wage (i.e., bound states) compared to states with a higher minimum wage (i.e., unbound states). To the extent that federal minimum wage changes are determined by factors that are not specific to certain states, this mitigates the concern that states choose to adjust minimum wages at certain points in the economic cycle. Because both the Substitution and Scale Hypotheses predict the effect of minimum wage changes on investment to be increasing in firms exposure to minimum wage employees, we partition our sample on labor intensity. Difference-in-differences estimates suggest that labor-intensive firms in bound states respond to federal minimum wage increases by reducing both total investment (i.e., the sum of 1 See for example Jorgenson (1963), Hall and Jorgenson (1967), Lucas (1967, 1976), Treadway (1969, 1971), Lucas and Prescott (1971), Brainard and Tobin (1968), Tobin (1969). 1

3 capital, research and development (R&D), and mergers and acquisitions (M&A) expenditures), and capital expenditures compared to observably similar labor-intensive firms in unbound states. We find no statistically significant difference in the investment patterns of non-labor-intensive firms in bound versus unbound states surrounding federal minimum wage increases. To further control for unobserved time-varying differences between states whose minimum wage is bound by the federal wage and other states, we introduce non-labor-intensive firms as an additional control group in a triple differencing framework that includes state x year fixed effects to control for local economic conditions. Non-labor firms are an attractive control sample because they exhibit similar investment patterns to labor-intensive firms over the business cycle and they are exposed to the same local economic conditions. Yet, we predict and find evidence that their investment policies are unrelated to minimum wage changes. This triple differencing specification identifies the effect of minimum wage changes on the investment of labor-intensive firms by comparing the relative investment change of labor and non-labor-intensive firms in bound states surrounding changes in the federal minimum wage to the same relative change in unbound states. Importantly, descriptive evidence and robustness analyses suggest that the parallel trends assumption underlying this triple differing analysis is plausible. Across our various specifications, we estimate that a 10% increase in minimum wages results in labor-intensive firms (i.e., those in industry years with below median employee-to-asset ratios) reducing investment by around 5 to 10 percent relative to other firms. This broad based effect of minimum wage hikes on corporate investment is consistent with theories predicting that minimum wage changes will affect both skilled and unskilled employee wages (Grossman, 1983; Akerlof and Yellon, 1990). This result also suggests that the Scale Hypotheses is the predominant channel through which minimum wage increases affect corporate investment. In our next set of tests, we partition our labor-intensive sample into two subsamples, minimum wage sensitive industries and non-minimum wage sensitive industries. Ex-ante, it is unclear whether the effect of minimum wage changes on investment will be more or less negative in minimum wage sensitive industries compared to the overall effect on labor-intensive industries. The Scale Hypothesis predicts that the investment reduction will be magnified since these markets face the largest input cost increase, but the Substitution Hypothesis predicts that the most minimum wage sensitive industries will be the most incentivized to substitute capital for labor. 2

4 We define restaurant, retail, and entertainment industries as minimum wage sensitive, because according to the 2015 Current Population Survey these are the only industries in which over 10% of employees make minimum wage. Triple difference results indicate that the most minimum wage sensitive firms reduce their investment relative to non-labor-intensive firms following minimum wage increases. Consistent with the Scale Hypothesis, the magnitude of the effect is approximately twice our full sample estimate. We also find a significant, albeit smaller, reduction in the investment of other labor-intensive firms. Thus, studies focusing on only minimum wage sensitive industries or those comparing the outcomes of minimum wage employees to groups of higher wage employees may be understating the impact of minimum wage changes. Another prediction of the Scale Hypotheses is that minimum wage changes will have a more negative effect on capital investments, compared to investments that are more likely to substitute for low-wage labor, such as R&D expenditures. Consistent with this prediction, we find a negative relation between minimum wage changes and capital investment, but an insignificant (negative) relation between minimum wage changes and R&D spending. We also find a negative relation between minimum wage expenditures and acquisition expenses, but the magnitude of this effect is larger and only statistically significant for the acquisition of non-labor-intensive targets. Since acquisitions of non-labor-intensive targets are more similar to traditional capital investments, and less likely to be motivated by post-merger layoffs (see e.g., Dessaint, Golubov, and Volpin, 2017), these estimates are consistent with the predictions of the Scale Hypothesis. As a final test of the Scale Hypothesis, we examine the extent to which firms scale down their total business. Under this hypothesis, an increase in factor prices not only reduces investment, but also the optimal scale of production. We show that labor-intensive firms respond to increases in minimum wages by decreasing the size of the firm; a 10% reduction in the minimum wage leads to between a 2-4% decrease in total assets. The reduction in scale is split approximately evenly across the firm s capital structure; the increase in minimum wages leads firms to both use less debt and issue less equity. Our findings consistently support the Scale Hypothesis as the predominant channel through which minimum wage changes affect corporate investment. Not only are minimum wage changes negatively related to investment, but the negative relation is larger for minimum wage sensitive firms and investments that are less likely to substitute for labor, and accompanied by reduced financing activity. Although we know of no compelling alternative to the Scale Hypothesis that 3

5 intuitively explains this collection of results, we conduct several robustness analyses to mitigate the possibility that the timing of federal minimum wage increases, differences between firms in bound and unbound states, or measurement error in firm-level exposure to state minimum wage changes contributes to our findings. Two of the three sets of federal minimum wage changes overlap with recessionary periods. If these recessions differentially affect firms in bound states then this may lead to a violation of the main assumption underlying our triple difference analysis, which is that the relative investment of labor- and non-labor-intensive firms evolves similarly in bound and unbound states surrounding federal minimum wage changes. To examine this possibility we separately examine the investment response to federal minimum wage increases during recessionary and expansionary periods. We find evidence of significant investment declines in both settings that are of similar magnitude, making it unlikely that the overlap between federal minimum wage changes and recessions significantly affects our results. We also replicate our main results using a propensity score matched sample, in which the firms in bound and unbound states are similar along observable dimensions. We continue to find that labor-intensive firms reduce their investment following minimum wage increases. This makes it unlikely that our findings are due to differences between firms in bound and unbound states. Our results are also robust to limiting our analysis to firms located within 50 miles of each other but on opposite sides of the border of a bound and unbound state. This alleviates concerns that differences in local economic conditions bias our results, and confirms that our results are not driven by the concentration of bound states in the U.S. heartland. Finally, we conduct several robustness tests to ensure that mismeasurement in firms exposure to federal minimum wage changes do not bias our results. Our primary measure for minimum wage exposure is an indicator for whether or not a firm is headquartered in a bound state. A limitation to this measure is that is does not directly measure the distribution of employees, which is what determines a firm s exposure to state-level minimum wage changes. We first address this limitation by showing that the effect of minimum wage on corporate investment is concentrated among firms that are small relative to their state s size, and therefore likely to have their employees more concentrated in their headquarter state. Since firm size is correlated with many other firm characteristics, such as financial constraints (Whited and Wu, 2006), we show that our findings are robust to using an alternate measure of a firm s exposure to bound states, 4

6 which is based on the geographical dispersion measure in Garcia and Norli (2012). Finally, we proxy for the investment of minimum wage sensitive firms using state-industry-year level establishment data. Again, we find a significant negative relation between minimum wage hikes and establishment growth within bound states compared to unbound states. Although the inability to perfectly measure firms exposure to state-level minimum wage changes remains a limitation of our setting, the persistent negative relation between minimum wage changes and investment across our specifications makes it unlikely that this mismeasurement materially contributes to our results. Throughout our analysis we find little evidence that the Substitution Hypothesis is economically meaningful no form of investment is positively related to minimum wage changes and firms do not appear to reduce their employees-to-assets ratio. However, we cannot rule out that such substitutions happen in the long-run. This highlights one limitation to our triple differencing framework. Although it is well suited to estimate the qualitative short-run response of corporate investment to changes in state-level minimum wages, future research is needed to understand the extent to which our findings generalize to other settings and to precisely identify the magnitude of the effects we observe. In addition to being policy relevant, our study contributes to several large strands of academic literature. Our evidence that minimum wage increases affect corporate investment fills an important void in the growing literature investigating the effect of labor market conditions on corporate policy. 2 For instance, in contrast to the voluptuous literature examining how employee rights or firing frictions affect capital expenditures (Autor, Kerr, and Kugler, 2007; Fairhurst and Serfling, 2016), M&A activity (John, Knyazeva, and Knyazeva, 2015; Dessaint, Golubov, and Volpin, 2017; Chatt, Gustafson, and Welker, 2017), and innovation (Acharya, Baghai, Subramanian, 2014), there is little evidence on how minimum wages affect such activities. We also add to the vast literature examining the effects of minimum wage on employment and product markets. There is no conclusive evidence that minimum wages have an immediate effect on employment levels, however there does appear to be an effect on wage dispersion. 3 2 Much of this literature has focused on financial policies, rather than investment. For example, Matsa (2010) shows that firms use their financial policy to influence union negotiations, Agrawal and Matsa (2013) find that firms lower leverage in response to reduced unemployment insurance to reduce their wage bill, and Simintizi, Vig, and Volpin (2014) and Serfling (2016) show that firms reduce their leverage in response to increased firing frictions. 3 See for example Katz and Krueger (1992), Card and Krueger (1994), Neumark and Wascher (2000), Card and Krueger (2000), Dube, Lester, and Reich (2010), Giuliano (2013), Sorkin (2015), Meer and West (2015) for evidence 5

7 Perhaps because of this ongoing debate regarding the most fundamental effects of minimum wage, the list of other outcomes shown to be affected by minimum wage is relatively short, including price levels, firm profitability, and personal finance decisions. 4 The set of studies most related to our evidence on the effect of minimum wage on corporate investment are those examining restaurant entries and exits surrounding minimum wage changes. Aaronson, French, Sorkin, and To (2017) find that minimum wage increases are followed by more restaurant turnover (i.e., exit and entry), while Luca and Luca (2017) add texture to this result by showing that low quality restaurants are most likely to exit. Our findings generalize this literature on several important dimensions. First, we provide evidence that minimum wage increases affect public firms not only small, private establishments. Second, using a more comprehensive investment measure we show on net minimum wage increases lead to less corporate investment. Finally, we provide evidence that minimum wage increases affect the corporate policies of laborintensive firms, even if they are not heavily reliant on minimum wage labor. 2. Conceptual Framework Congress enacted the first federal minimum wage in the 1938 Federal Fair Labor Standards Act (FLSA). The minimum wage was set at $0.25 and the law applied to a relatively small subset of employees. Over time, Congress has both increased the minimum wage and expanded the universe of employees covered by the law. Since 1982, the federal government has raised minimum wage seven times through laws passed in 1989, 1996, and 2007, with the current rate of $7.25 becoming effective in The FLSA also allows states to set their own minimum wage rates, with the higher of the state and federal minimum wage applying to employees working in the state. The solid line in Figure 1 shows that many states began adopting minimum wage rates higher than the federal rate around Economic theory generates competing predictions regarding the effect of minimum wage increases on corporate investment. In traditional neoclassical, cost-of-adjustment, and q-theory models of investment, factor prices are an important determinant of investment (see e.g., Jorgenson, 1963; Hall and Jorgenson, 1967; Lucas, 1967, 1976; Treadway, 1969, 1971; Lucas and on employment levels and Dinardo, Fortin, and Lemieux (1996), Lee (1999), Macurdy (2015), David, Manning, Smith (2016) for examinations of wage dispersion. 4 See Aaronson (2001) and Aaronson and French (2007) for evidence on price levels, Draca, Machin, Van Reenen (2011) for evidence on profitability, and Tonin (2011) and Aaronson, Agarwal, and French (2012) for evidence on personal finance decisions. 5 See Internet Appendix Table A1 for a complete list of federal minimum wage changes. 6

8 Prescott, 1971; Brainard and Tobin, 1968; Tobin, 1969). These theories all predict a negative relation between minimum wages and capital investment to the extent that the minimum wage affects production costs. Such a relation is also consistent with Aaronson (2001), Dube, Naidu, and Reich (2007), and Aaronson, French, and Mcdonald (2008), which all provide evidence that a significant portion of the costs attributable to minimum wage are passed on to customers via higher prices. If there is any elasticity in demand, this behavior will trigger a scale effect whereby product market size shrinks, reducing optimal capital investment. The above arguments, which we refer to as the Scale Hypothesis, relate most directly to capital investment for production. Expanding the types and purposes of investment motivates the (non-mutually exclusive) Substitution Hypothesis, which predicts a positive relation between minimum wage increases and investment. Since firms can substitute capital for labor (see e.g., Arrow, Chenery, Minhas, and Solow, 1961), firms may respond to minimum wage increases by increasing investment, with the intention of substituting away from the now more expensive labor inputs in the long run. Stigler (1946) articulates this idea, arguing that the minimum wage may lead to increased investment if it (1) makes previously suboptimal production techniques optimal, or (2) shocks managers into adopting new technologies. This type of long-run substitution is broadly consistent with the evidence in the minimum wage literature. In a meta-analysis of 23 academic studies, Card and Krueger (2015) argue that there is little evidence that minimum wage increases have an immediate effect on employment, while Meer and West (2015) and Sorkin (2015) document a long-run reduction in the employment of minimum wage workers following minimum wage increases. To formalize these two non-mutually exclusive hypotheses, consider a firm in a perfectly competitive market that produces a single good y with two inputs (capital K and labor L). The firm maximizes ππ(kk, LL) = pppp(kk, LL) rrrr wwll (1) where p is the price of output, r is the rental price of capital, and w is the hourly wage. Assume that the firm s production function is increasing and concave, so that ff KK, ff LL > 0 and ff KKKK, ff LLLL < 0, where ff xx = and ff xxxx = 2 ff ( ). Since the production function is strictly concave, the firm maximizes profit by choosing levels of capital and labor such that the marginal product of each input equals its marginal cost. That is, the firm chooses K * and L * to satisfy 7

9 ppff KK = rr (2) ppff LL = ww (2 ) To generate predictions regarding the effect of wage changes on investment we take the total derivative of Eq. (2) with respect to w and solve for KK. KK = LL ff KKKK ff KKKK (3) Since diminishing returns to capital and downward sloping labor demand make both ff KKKK and LL negative, the sign of KK depends on the sign of the cross-partial derivative ff KKKK. 6 If ff KKKK is positive (i.e., capital and labor are gross complements), then KK < 0 and the Scale Hypothesis dominates the Substitution Hypothesis. Conversely, if ff KKKK is negative (i.e., capital and labor are gross substitutes) then KK > 0 and the Substitution Hypothesis dominates. Thus, the effect of wage changes on corporate investment depends on whether the investment is complimentary or substitutable with labor. Many types of investment are complementary to labor, such as a new production facility or a new retail store that depend on employees for the capital to be productive. Wage increases will result in these investments becoming less attractive. Other types of investment can substitute for labor because the investment reduces the marginal productivity of labor. For example, selfordering kiosks make the cashier at a fast food restaurant less productive. This stylized model suggests that the effect of minimum wage on investment is an empirical question, the answer to which depends on both the sensitivity of a firm s wage bill to minimum wage changes and the type of investment under consideration. 7 Figure 2 illustrates these empirical predictions separately for the Scale and Substitution Hypotheses. Under either hypothesis, investments that are complementary with low-wage labor will be more negatively (or less positively) affected by minimum wage increases, compared to investments that substitute for low-wage labor. Both hypotheses also predict that any effect of minimum wage changes on 6 Under our assumptions, Young s theorem shows that the cross-partial derivatives are equal, ff LLLL = ff KKKK. 7 In the Internet Appendix, we show that these conclusions hold in more complex settings, such as when firms can adjust prices in response to changes in minimum wage or simultaneously invest in two different types of capital. 8

10 investment will be more muted for firms that are less exposed to minimum wage employees (i.e. when LL / is smaller). 3. Sample and Descriptive Evidence To examine the effect of minimum wage changes on corporate investment we use a sample of (non-financial and non-utility) Compustat firms from 1987 to We begin our sample in 1987 because our identification strategy requires cross-sectional variation in state-level minimum wages, which first occurs with regularity in Since 1987 is 3 years before the first federal minimum wage change in our sample occurs, we balance our sample by ending in 2012, which is 3 years after the completion of the most recent federal minimum wage change. We have 10,645 firms across 25 years, amounting to 106,340 observations, although the effective sample size is somewhat smaller in some cases as fixed effects fully absorb some observations. Table 1 summarizes the characteristics of the firms in our sample; we winsorize all variables at the 1% and 99% levels. Seventy-two percent of firm-years are in states with their minimum wage equal to the federal minimum wage rate. On average, the minimum wage increases by approximately 3% per year. Defining investment as the sum of capital expenditures, R&D, and M&A, the average (median) firm invests at a rate of 17.2% (10.1%) of their total assets each year. Capital expenditures are the most common form of investment, closely followed by R&D. Less than 20% of investment is in the forms of M&A expenses. The average firm has liabilities of approximately 49% of total assets, holds approximately 19% of their assets in cash, and is profitable with annual net income equaling approximately 3% of beginning of period assets. See Appendix A for detailed definitions of all variables used throughout the analysis. 4. Identifying the Effect of Minimum Wage on Corporate Investment The active political debate surrounding minimum wages has spurred a significant body of academic research attempting to understand the effects of wage floors. Although variation in the timing of state minimum wage changes seems to present an attractive setting to estimate the economic effects of minimum wage, the analysis is complicated because states raise minimum wages at non-random times and the states that raise minimum wages differ from other states. Allegratto, Dube, Reich, and Zipperer (2013) show that states that increase minimum wages have different business cycle severity, inequality, and composition of the labor force. For instance, in our sample state-level minimum wage changes have a significant correlation of with a state s lagged unemployment rate, suggesting that minimum wage hikes are more common when 9

11 economic conditions are favorable. Since firms may factor state-level economic conditions into their investment decision, regressions of corporate investment on state-level minimum wage changes are difficult to interpret. Table A2 of the Internet Appendix illustrates this. In the absence of control variables, there is a highly significant positive relation between corporate investment and recent minimum wage increases, however the relation becomes statistically insignificant after the inclusion of firm and year fixed effects and turns negative (and insignificant) after controlling for firm- and state-level characteristics. One way to overcome the fact that local economic conditions contribute to state-level minimum wage changes is to use regional controls. In a seminal paper, Card and Krueger (1994) study the effects of an increase in New Jersey s minimum wage on fast-food restaurants along the New Jersey-Pennsylvania border. Since firms on either side of the state border face similar economic conditions, the effect of minimum wage can be identified. This approach has been widely adopted and expanded to more general settings (see, e.g. Dube, Lester and Reich, 2010; Allegretto, Dube and Reich, 2011; Magruder, 2013). However, it is difficult to apply this technique to the study of corporate decisions because firms do not concentrate their business activities along state borders. In the remainder of this section, we introduce a novel strategy to identify the effect of minimum wage changes on corporate investment. In Section 4.1, we motivate two controls groups that allow us to plausibly identify the effect of minimum wage changes on corporate investment. We then present preliminary results to motivate our primary identification strategy, which we discuss in Section Two Control Groups Bound Status and Exposure to Federal Minimum Wage Changes The first feature of our identification strategy exploits the fact that an increase in the federal minimum wage rate affects states with minimum wage rates equal to or less than the federal minimum wage (i.e., bound states) more directly than states with higher minimum wages (i.e., unbound states). 8 Thus, the approximately 1/3 of our sample that resides in unbound state-years is a natural control sample when examining the effect of federal minim wage changes on the investment of firms in bound states. By taking the decision to raise minimum wages out of the 8 Clemens and Wither (2016) use this idea to identify the effect of minimum wage changes on employee outcomes surrounding the federal minimum wage change. 10

12 control of the state, we begin to break the endogenous link between minimum wage changes and state economic conditions. Internet Appendix Table A3 provides a detailed breakdown of the annual percentage change in minimum wages for bound and unbound for each year since Bound states minimum wage increases by approximately 11% during the 7 years that the federal minimum wage increases and 0% in other years. On average, these discrete jumps in the minimum wage are matched by unbound states within 2 to 3 years. 9 Thus, a federal minimum wage increase represents a shock to the relative wages paid in bound versus unbound states, with the largest effect being in the year of the federal minimum wage increase. This setting lends itself to identifying the consequences of minimum wage changes, but only to the extent that we control for any differences in how the economies of bound and unbound states correlate with federal minimum wage increases. We start with a preliminary difference-indifferences specification examining how corporate investment differentially changes in bound versus unbound states surrounding federal minimum wage changes. Here, the coefficient of interest is β in the following equation. Inv it = β (Bound it-1 ΔMin Wage t ) + γ 1 (Bound it-1 ) + XX it 1 ΛΛ + SS it 1 ΩΩ + α i + τ tt + ϵ it (4) where Inv it is the total investment (i.e., the sum of capital, R&D, and M&A expenses) or capital expenditures of firm i at time t, scaled by lagged total assets. Bound it-1 indicates whether firm i is headquartered in a state with minimum wage equal to or less than the federal rate at the beginning of the year. ΔMin Wage t is the annual percentage in the nominal federal minimum wage for the year ending at the beginning of the calendar quarter before fiscal year end. XX it 1 is a vector of firm-level controls, including employees, liabilities, tangibility, ln(assets), profitability, market to book, and cash. To help control for differences between bound and unbound state-level economic conditions we also include SS it 1, which is a vector of controls for state-level economic conditions, including population, change in population, unemployment, change in unemployment, state-level average wage, and change in state-level average wage. Panel A of Table 2 estimates Equation 4 over the full sample. In Column 1, the dependent variable is total investment and the interaction between minimum wage changes and bound status 9 The minimum wage policies in unbound states vary considerably. Some unbound states index their minimum wage to inflation, others pre-schedule minimum wage rate increases years in advance, and others rarely adjust their minimum wage. 11

13 is negative, but statistically insignificant. The coefficient is of similar magnitude in Column 2 using capital expenditures as the dependent variable; however, it is significant at the 10% level. These findings are consistent with the Scale Hypothesis, which predicts that minimum wage increases will negatively impact corporate investment, and that this effect will be largest for capital investment. However, an important caveat to this analysis is that there may be uncontrolled for differences in the local economic conditions of bound states surrounding federal minimum wage increases, which compromise our ability to identify the causal effect of minimum wage changes on corporate investment Defining and Partitioning on Labor Intensity We next introduce a within-state control sample of firms that is subject to the same local economic conditions, but less sensitive to minimum wage increases, to control for the possibility that economic conditions differ in bound and unbound states. The discussion in Section 2 suggests that minimum wage changes will affect investment to the extent that minimum wage changes represent a factor cost shock. Thus, minimum wage changes are more likely to matter for laborintensive firms, especially those that employ a significant number of minimum or low wage employees. To the extent that the negative relation between recent minimum wage changes and corporate investment is causal (i.e., not driven by local economic conditions), we expect the relation to be concentrated among labor-intensive firms. Our primary measure of labor intensity, Labor, is a broad measure of how reliant an industry is on labor. The models of Grossman (1983) and Akerlof and Yellon (1990) support the use of a broad definition of minimum wage exposure by providing channels through which minimum wage increases will affect the wages of all workers, not just those being paid minimum wage. Specifically, the authors show that this type of spillover can occur if higher minimum wages either (1) create more demand for skilled labor, since the alternative has become more expensive, or (2) reduce the incentive for skilled employees to exert effort. After inflation adjusting total assets, we define a labor-intensive firm as one in a Fama- French 49 industry-year with a median employee-to-assets ratio that is above the median of all industry-years in our sample period. 10 Table A4 in the Internet Appendix lists the top and bottom 10 Fama-French industries based on the percent of years an industry is labor-intensive. Intuitively, 10 This definition allows for changes in the percentage of industries that are considered labor-intensive each year, which is important given the changing role of technology throughout our sample period. 12

14 industries that have been traditionally associated with minimum wage, such as retail and restaurants, make up a large proportion of the firms that we define as labor-intensive. Columns 1 and 2 in Table 2 Panel B show that the total investment and capital expenditures of labor-intensive firms in bound states significantly declines relative to labor-intensive firms in unbound states in the year following a federal minimum wage increase. Since not all labor-intensive industries are equally exposed to minimum wage workers, we introduce a second measure, Sensitive, which is an indicator for firms operating in the restaurant, retail, or entertainment industries (i.e., Fama-French 49 industries 7, 43, and 44), which employ the largest fraction of minimum wage workers. 11 Comparing the estimates in Columns 2 and 4 of Table 2 Panel B indicates that minimum wage sensitive firms exhibit almost twice the decline in capital expenditures following minimum wage changes, relative to the larger sample of laborintensive firms. The similarity between the coefficients in Columns 3 and 4 suggests that minimum wage sensitive firms do not significantly reduce other forms of investment. In contrast to the investment declines among labor-intensive and minimum wage sensitive firms following minimum wage increase, Panel C of Table 2 indicates no such effect after restricting the sample to non-labor intensive firms. Both estimates are statistically insignificant with t-statistics of 0.75 and -0.14, respectively. This evidence that investment, and in particular capital investment, declines most for the firms most reliant on minimum wage labor is consistent with the predominant effect of increasing the minimum wage on corporate investment being in the form of factor price increases. Again, these findings are consistent with the Scale Hypothesis discussed in Section Evidence on the Identifying Assumptions of Difference-in-Differences Estimates This difference-in-differences specification will identify the causal effect of minimum wage increases to the extent that firms in bound and unbound states are similar (and are evolving similarly) around the time that the federal government adjusts minimum wages. Table 3 provides some initial descriptive statistics on the plausibility of this assumption by presenting descriptive statistics partitioned by labor intensity and bound status. Conditional on labor intensity, firms in bound and unbound states are economically similar along most observable dimensions. This similarity is especially pronounced for labor-intensive and minimum wage sensitive firms, which 11 Specifically, the 2015 Current Population Survey shows that that the only industries in which over 10% of employees make minimum wage are Leisure and Hospitality and Retail Trade. 13

15 we expect to be most affected by minimum wage changes. For example, labor-intensive firms in bound states invest at 14.9% of assets compared to 15.0% of assets in unbound states. The typical firm in the two groups is also very similar in terms of employees-to-assets ratio, size, liabilities, tangibility, market-to-book, profitability, and cash holdings. The differences between the nonlabor-intensive firms in bound and unbound states are more substantial. In particular, non-labor firms in unbound states have lower tangibility and profitability and hold higher levels of cash. Figure 3 shows that one reason for the similarity between the firms in bound and unbound states is that the bound status of many states changes throughout our sample period, meaning that some firms are both treatment and control firms at different times. California, Florida, Ohio, and New York comprise a large portion of our sample (see Internet Appendix Figure A1) and all change bound status at various points in our sample period. Moreover, these states cover the political spectrum and have different business environments. However, Figure 3 also indicates that the economies and political views of many bound states, represented by dark shading, are highly correlated. In particular, bound states are concentrated in the U.S. heartland, while states along the coasts are more likely to be unbound. This raises the possibility that the relative economic conditions of bound and unbound states may vary over time, perhaps in a manner that correlates with federal minimum wage increases. For instance, two of the three federal minimum wage increases occur in proximity to economic recessions. If recessions differentially affect bound states, the difference-in-differences analysis in Table 2 may be misattributing the differential effect of recessions on bound states to minimum wage changes. Figure 4 descriptively examines whether the investment of firms in bound and unbound evolves differently surrounding federal minimum wage increases. Each line represents firms of a given labor intensity and is derived from the following regression, which we estimate using a sample of firm-years from three years before until two years after each Federal minimum wage change. 12 YY iiiiii = β 1 (BBBBBBBBdd iiiiii YYYYYYrr 3 ) + β 2 (BBBBBBBBdd iiiiii YYYYYYrr 2 ) + β 3 (BBBBBBBBdd iiiiii YYYYYYrr 1 ) + β 4 (BBBBBBBBdd iiiiii YYYYYYrr 1 ) + β 5 (BBBBBBBBdd iiiiii YYYYYYrr 2 ) + XX it 1 ΛΛ + SS it 1 ΩΩ + αα iiii + ττ tt + εε iiii (5) 12 Longer windows are not possible because the first set of federal minimum wage changes ends in 1991 and the second begins in

16 where YY iiiiii is firm i s outcome at time t during event e. Each dot on a line plots a β from the above regression. For example, β1 corresponds to observations three years prior to a federal minimum wage change. A descriptive test for the assumptions underlying the difference-in-differences estimates in Table 2 is whether the lines in Figure 4 are flat prior to minimum wage increases (i.e., whether the trends in investment in bound and unbound states are parallel). The solid line suggests that this assumption is reasonable for the sample of labor-intensive firms, while the dashed lines cast doubt on this assumption for the sample of non-labor intensive firms and for the subsample of minimum wage sensitive firms. More generally, the dashed lines raise the possibility that state-level economic conditions differ in bound versus unbound states surrounding the passage of federal minimum wage changes. This casts some doubt on the central assumption of the difference-indifferences estimate, which is that firms in bound and unbound states are similar around the time that the federal government adjusts minimum wages Primary Identification Strategy: A Triple Differencing Approach Our primary identification strategy integrates the two control groups discussed in Section 4.1 firms in unbound states and non-labor intensive firms into a triple differencing specification. This approach augments the difference-in-differences specification with a control sample of non-labor-intensive firms. Importantly, Panel C of Table 2 suggests that minimum wage changes have little effect on the investment of non-labor intensive firms. This is not because labor and non-labor intensive firms generally invest differently. Figure 5 documents a strong positive association between the investments of the two types of firms. Average investment for both groups increases during the 1990s, spikes around the year 2000, and exhibits a less pronounced boom and bust cycle during the next decade. During our sample, the correlation between the two series is approximately Taken together, this evidence suggests that non-labor-intensive firms are a reasonable benchmark from which to identify the effect of minimum wage changes on the investment of labor-intensive firms their investment is generally influenced by similar economic forces, but less sensitive to minimum wage changes. The primary benefit to this triple differencing design, relative to the difference-indifferences design, is that we have a control sample of firms housed in the same state, and thus can directly control for state-level economic conditions surrounding federal minimum wage increases using state year fixed effects. Equation 6 formalizes this triple differencing 15

17 specification YY iiii = β BBBBBBBBdd iiiiii-1 ΔMMMMMM WWWWWWee iiii LLLLLLLLrr iiii + γγ 1 BBBBBBBBdd iiiiii-1 ΔMMMMMM WWWWWWee iiii + γγ 2 BBBBBBBBdd iiiiii-1 LLLLLLLLrr iiii + γγ 3 (ΔMMMMMM WWWWWWee iiii LLLLLLLLrr iiii ) + γγ 4 BBBBBBBBdd iiiiii-1 +γγ 5 LLLLLLLLrr iiii + γγ 5 ΔMMMMMM WWWWWWee iiii + XX iiii 1 ΛΛ + αα ii + ττ jjjj + εε iiii (6) BBBBBBBBdd iiiiii-1 is an indicator variable that is equal to one if state j has a state minimum wage less than or equal to the federal minimum wage at the end of the firms prior fiscal year. ΔMMMMMM WWWWWWee iiii is the percentage change in the nominal federal minimum wage over the firm s prior fiscal year, and LLLLLLLLrr iiii indicates that firm i is in a labor-intensive industry in year t. The coefficient on the triple interaction term, β, represents the differential effect of minimum wages on labor-intensive firms compared to non-labor-intensive firms across bound and unbound states. State year fixed effects absorb any unobserved economic changes in a given state-year, while firm fixed effects control for time invariant firm characteristics. 13 We do not expect the inclusion of firm-level control variables to meaningfully affect our estimate of β because β is the estimated response to an arguably exogenous change in minimum wage. Nevertheless, in most specifications we control for firm-level characteristics, measured in year t-1. Our firm controls include total assets, liabilities, asset tangibility, profitability, market-to-book, cash, and employees. We cluster standard errors at the state level. Since firms rarely change their headquarter state, this clustering deals with within firm correlation of errors, which Bertrand, Duflo, and Mullainathan (2004) argue is important in difference-in-differences estimation. This clustering also controls for any correlation in the error term that is attributable to any state or state-year economic shock. Results are similar clustering at the year or firm levels. This specification isolates the effect of minimum wage on corporate investment by comparing changes in investment after minimum wage increases between labor- and non-laborintensive firms across bound states (which are affected by federal minimum wage changes) and unbound states (which are not affected). Assuming that any differences between labor-intensive and non-labor-intensive firms do not change differentially in bound versus unbound states, in a manner that is correlated with the timing of federal minimum wage changes, this triple differencing framework will identify the effect of minimum wage changes on the differential investment of 13 Our Bound and Change in Min. Wage variables are measured as of each firm s fiscal year end, so including year x state fixed effects does not absorb the coefficient on Bound or the interaction of Bound x Δ Min Wage. Since these coefficients are identified off differences in fiscal years, it is not clear that they have any meaningful economic interpretation. 16

18 labor-intensive firms. This assumption is more intuitive than the assumption that firms in bound and unbound states evolve similarly surrounding federal minimum wage passages, which underlies the difference-in-differences estimation. Figure 4 provides descriptive evidence consistent with this assumption when comparing minimum wage sensitive firms and non-labor intensive firms. In both samples, firms in bound states experience a decline in investment relative to comparable firms in unbound states between three and two years before minimum wage changes, which reverses during the subsequent year. The pre-trends are less similar comparing the full sample of labor- and non-labor-intensive firms, although all three groups end up investing similarly in the year prior to federal minimum wage changes. In Section 6, we replicate our analysis using a matched sample in which (1) the labor- and non-labor-intensive firms in bound and unbound states are more similar and (2) the differential investment of firms in bound and unbound states exhibits similar trends within our labor- and nonlabor-intensive samples. The similarity of our estimates using the full and matched samples makes it unlikely that our findings are driven by trends in investment that are unrelated to minimum wage changes. Moreover, we do not know of any compelling reasons why our identifying assumptions would be violated. In particular, in Section 6 we discuss why and provide evidence suggesting it is unlikely that factors such as the financial crisis, the geographical concentration of bound states, or corporate lobbying affect our results through violations of our identifying assumptions. 5. Main Results: Minimum Wages and Corporate Investment In Section 2, we motivate two non-mutually exclusive hypotheses regarding the effect of minimum wage changes on investment. The Scale Hypothesis predicts that investment will decline as a firm s wage bill rises, while the Substitution Hypothesis predicts the opposite. Here, we use a triple differencing specification to provide evidence on the predominant channel through which minimum wage changes affect corporate investment. As discussed in Section 4.2, we identify the effect of minimum wages on investment by comparing how the relative investment of laborintensive and non-labor-intensive firms changes following federal minimum wage increases in bound states compared to unbound states. We begin with a broad measure of labor intensity, which defines (non-)labor-intensive firm-years as those in industry-years with (below) above median employees-to-assets ratios, and a comprehensive measure of corporate investment that includes capital, M&A, and R&D 17

19 expenditures. The significantly negative coefficients on the Bound x Δ Min. Wage x Labor triple interaction in Columns 1 and 2 of Table 4 corroborate the evidence in Figure 4 and Table 2 that minimum wage changes are followed by reduced investment for labor-intensive firms. Notably, the coefficient estimates are similar in Columns 1 and 2, suggesting that our specific choice of firm-level controls has little effect on the relation between arguably exogenous minimum wage changes and corporate investment. Comparing the coefficient in Column 2 to the average investment of 0.17 suggests that a 10% increase in minimum wage is predicted to reduce the investment of labor-intensive firms by approximately 8%. Our definition of bound status in Columns 1 and 2 is based on a firm s headquarters location, and is only appropriate if a firm s employees disproportionately reside in the state of their headquarters. In Columns 3 through 5, we take two steps to mitigate this measurement error. First, we partition the sample based on a firm s size relative to their home state. Column 3 (4) restricts the sample to the bottom (top) two quartiles of total assets relative to headquarter state population. We find that the negative relation between corporate investment and minimum wage is concentrated among firms that are small relative to their headquarter state. Scaling the coefficient of in Column 3 to the average investment of firms in the bottom two size quartiles, which is 19.9% of total assets, suggests that within this sample of smaller firms a 10% increase in minimum wage is predicted to reduce investment by just over 10%. Among larger firms, the estimated effect is a statistically insignificant 5.3%. These findings are consistent with large firms or firms in small states being less exposed to state-level minimum wage changes because their employees are more spread across state lines. However, this interpretation of the differential effects in Columns 3 and 4 is confounded if there is a correlation between firm size and omitted variables that influence the relation between minimum wage changes and corporate investment. In Column 5, we employ a second measure of firm-level exposure to states bound by the federal minimum wage by augmenting our bound indicator with the geographical dispersion measure used in Garcia and Norli (2012). As we discuss in Appendix A, Alt. Bound equals the percentage of state names mentioned in SEC filings that are bound states as of the filing date. This provides a continuous measure of exposure to minimum wage changes, assuming that firms primarily discuss states where they have employee operations. A limitation to the measure is that to maintain a consistent sample, we must extrapolate the Garcia and Norli (2012) sample to our 18

20 sample period and then replace missing observations with our Bound indicator. 14 We continue to find a significant negative relation between minimum wage increases and corporate investment using this alternate measure, which is similar in magnitude to the estimate in Column 3 using our small firm subsample. The evidence in Table 4 suggests that minimum wage increases have a statistically significant negative effect on the investment of labor-intensive firms, which is consistent with reduced scale being the predominant channel through which minimum wage changes affect corporate investment. These findings corroborate the difference-in-differences evidence in Table 2 under a different, more plausible set of identifying assumptions. Precisely identifying the magnitude of the effect is challenging. In particular, the magnitudes we estimate may not quantify the causal effect of minimum wage changes on labor-intensive firms to the extent that unbound states adjust their minimum wages in response to federal minimum wage increases, non-laborintensive firms are impacted by minimum wage changes, or firms in bound states adjust their behavior in anticipation of future federal minimum wage changes. Subject to these caveats, we estimate that a 10% increase in minimum wage results in the total investment of public firms declining by between 7% and 10% percent, with the magnitude being closer to 10% within the smaller half of firms, which are more directly affected by state-level minimum wage changes Partitioning on Minimum Wage Sensitivity The definition of labor intensity that we have used thus far is broad relative to those used in the minimum wage literature, which tend to focus on the few industries with the highest concentration of minimum wage employees. As we discuss in Section 4.1.2, one benefit to such a broad measure is that it is capable of capturing spillover effects whereby minimum wage increases may lead to increases in wages for a significant fraction of non-minimum wage employees (see e.g., Grossman, 1983 and Akerlof and Yellon, 1990). In this section, we examine whether the investment decline we observe following minimum wage increases is larger or smaller in the industries employing the most minimum wage workers. The Scale Hypothesis predicts that the investment reduction will be magnified within the most minimum wage sensitive industries. Alternatively, to the extent that these industries are particularly incentivized to substitute capital for labor following minimum wage hikes, the 14 Results are robust to adjusting the extrapolation method or dropping the approximately 15% of our sample for which the firm never appears in the Garcia and Norli (2012) dataset. 19

21 Substitution Hypothesis predicts that the investment reduction will be attenuated. To investigate this, we partition our labor-intensive measure based on how minimum wage sensitive the industry is. We continue to use the same triple difference framework and non-labor control group. We define minimum wage sensitive industries as the restaurant, retail, and entertainment industries and we define other labor-intensive industries as non-minimum wage sensitive. Panel A of Table 5 presents results comparing minimum wage sensitive firms to non-labor firms. Column 1 contains the full sample, while Columns 2 and 3 restrict the analysis to small and large firms relative to their headquarter state s size, respectively. The results indicate that the investment decline following minimum wage increases is larger for the most minimum wage sensitive industries. Column 2 shows that the effect is strongest within the sample of small firms and is approximately 66% larger than the corresponding full sample results in Column 3 of Table 4. Column 3 reveals no significant effect of minimum wage changes on the largest half of firms in minimum wage sensitive industries. These estimates suggest that small minimum wage sensitive firms respond to a 10% increase in minimum wage with a 12-15% cut in investment. This estimate might overstate the magnitude of the actual decline in production capacity for retail and restaurant firms. For these firms, one of the major forms of investment is property, which can be leased instead of bought. 15 In Table A5 of the Internet Appendix, we examine the extent to which firms substitute leases for investment by estimating the change in future operating lease commitments following minimum wage increases. 16 We find that approximately one-third of the investment decline is substituted for by operating leases. Consequently, our finding that minimum wage sensitive firms respond to a 10% minimum wage increase by reducing investment by 12-15% does not necessarily mean that these firms are delaying or cancelling 12-15% of their projects. Rather, they appear to be delaying or cancelling 8-10% of investment projects and financing another 4-5% with leases instead of capital investments. Notably, we find no such substitution among non-minimum wage sensitive industries, suggesting that the observed investment decline in these firms directly captures delayed or cancelled projects. To further clarify the magnitude of the investment decline that we observe in minimum wage sensitive industries, we present a back of the envelope estimate of the dollar change in wage 15 Sale-leasebacks are also a common tactic employed by restaurant and retail firms when faced with financial difficulties. Sears is one recent example. 16 We capitalize future lease commitments by discounting these commitments at the 10-year Treasury bond yield, following the procedure in Rauh and Sufi (2011). 20

22 expenses, investment, and leases for the median small restaurant in our sample, which has 564 employees, annual investments of $3.1 million, and net income of $211,000. Assuming that the firm s employees work full-time, a $0.73 increase in the minimum wage (i.e., 10% increase from current levels) wipes out the entire net income of the firm if more than 25% of employees earn minimum wage. The actual cost shock might be smaller if the restaurant is able to pass on some of the costs through higher prices or if it employs fewer minimum wage workers. 17 On the other hand, the shock might be larger if there are workers paid above the current but below the new minimum wage or to the extent that increases in the minimum wage affect the broader wage scale. 18 Our estimates imply that the restaurant responds to this $211,000 cost shock by cutting investment by about $471,000. For perspective, the entrepreneurial magazine Inc. estimates that the cost of opening a new restaurant location is between $ ,000, so the numbers are consistent with a small restaurant reacting to higher minimum wages by purchasing one fewer location. Panel B presents results from a similar analysis using the labor-intensive firms that are not minimum wage sensitive as the treatment group. The significantly negative coefficients of -0.13, -0.22, and in Columns 1, 2, and 4 are approximately half of the magnitude of the corresponding effect for minimum wage sensitive firms presented in Panel A. The smaller investment reduction within the non-minimum wage sensitive subsample further supports the Scale Hypothesis as the primary channel through which minimum wage changes affect corporate investment. In addition, the significant investment decline in non-minimum wage sensitive industries supports the economic importance of theories predicting that minimum wage increases affect more than just the wages of minimum wage employees (see e.g., Grossman, 1983 and Akerlof and Yellon, 1990). These findings suggest that although the effect of minimum wage changes is predictably larger in the most minimum wage sensitive industries, studies focusing on only minimum wage sensitive industries or those comparing the outcomes of minimum wage employees to groups of higher wage employees may be understating the impact of minimum wage changes. 17 In the last year of our sample (2012), the BLS estimated that 19% of leisure and hospitality workers were employed at or below the minimum wage. 18 When Seattle passed a higher city minimum wage, approximately 36% of restaurant workers were earning less than the new minimum (see Jardim et al, 2017). 21

23 5.2. Partitioning on Investment Type The evidence presented thus far suggests that the predominant effect of minimum wage increases is to reduce corporate investment, especially for the firms most reliant on minimum wage workers. These findings are consistent with the Scale Hypothesis. Here, we test another prediction of the Scale Hypothesis, which is that the effect of minimum wage increases on investment will be most negative for capital investments. By comparison, R&D expenses are less likely to complement minimum wage workers, and more likely to represent investments designed to develop new technologies, perhaps with the intention of a long-run substitution away from labor inputs. To the extent that we are identifying the causal effect of minimum wage changes on investment, our conceptual framework suggests that minimum wage changes should have a more negative effect on capital expenditures than R&D expenditures. In Table 6, we replicate our main tests after partitioning total investment into its components. Panel A contains the full sample, while Panel B restricts the labor-intensive firms to those most sensitive to minimum wage employees. Column 1 of Panels A and B indicates a significantly negative relation between minimum wage changes and capital expenditures, with t- statistics of and -3.58, respectively. By contrast, Column 2 provides no evidence of a significant relation between minimum wage changes and R&D expenses both t-statistics are less than 0.38 in magnitude and both point estimates are less than 13% the size of the coefficient in Column 1. In Panel B, a z-test for the difference between the coefficients in Columns 1 and 2, which assumes independence, rejects the null hypothesis that the coefficients are equal at the 5% level. Table A6 of the Internet Appendix shows that the main takeaways from Table 6 are similar using our alternate measure of bound status. Compared to our main specification, the relation between minimum wage changes and capital expenditures is somewhat weaker over the full sample and stronger when focusing on minimum wage sensitive firms. We continue to find no significant relation between minimum wage changes and R&D expenditures. These findings further support the Scale Hypothesis the predominant channel through which minimum wage changes affect corporate investment. The negative coefficients for both capital and R&D investments provide little support for the Substitution Hypothesis. We next examine the effect of minimum wage changes on M&A expenditures. Due to the wide variety of M&A motives, the effect of minimum wage changes on M&A activity is an empirical question. On the one hand, M&As are a common way for firms to acquire capital. To 22

24 the extent that this capital complements labor, we expect a negative relation between minimum wage changes and M&A activity. On the other hand, there may be a positive relation between minimum wage increases and M&A activity to the extent that the acquisition creates economies of scale by allowing the firm to layoff redundant workers. Dessaint, Golubov, and Volpin (2017) support this idea with evidence that labor restructuring is an important source of M&A value. Column 3 of Table 6 tests this prediction. Panel A shows that there is a significantly negative relation between minimum wage changes and the M&A expenditures of the average labor-intensive firm. The magnitude of the decline is similar to the reduction in capital expenditures reported in Column 1. Panel B also reveals a negative relation between minimum wage changes and M&A expenditures for firms in minimum wage sensitive industries. The estimate is statistically significant with a t-statistic of -2.1, although the magnitude is about half of the size of the estimated effect on capital expenditures. In Column 1 of Table 7, we replicate this analysis using M&A data from SDC. Here, the dependent variable, which we denote All SDC M&A, equals the total dollar value of M&A transactions reported in the SDC database, scaled by beginning of period total assets. On average, firms spend 2.2 percent of their lagged assets on M&As reported in the SDC database, which is approximately 70% of the total M&A expenditures reported in Compustat. Panels A and B both reveal a significantly negative relation between minimum wage changes and M&A expenditures, with point estimates that are similar in magnitude to those using the Compustat measure. One benefit to the SDC M&A data is that we can more precisely test the predictions in Section 2 by separately measuring M&As targeting labor-intensive and non-labor-intensive firms. M&As targeting labor-intensive targets are more likely to be motivated by post-merger restructuring (see e.g., Dessaint, Golubov, and Volpin, 2017). By contrast, mergers targeting nonlabor-intensive firms are less likely to be driven by the possibility of substituting away from labor inputs. The Scale Hypothesis predicts that we should see less negative effects of minimum wage on investment-types that substitute for labor. Consequently, we expect the negative effect of minimum wage changes on M&A activity to be concentrated in the subsample of non-laborintensive M&As, ceteris paribus. In Columns 2 and 3 of Table 7, we partition our SDC M&A measure based on whether or not the target is in a labor-intensive industry. The entire negative effect of minimum wage changes on M&A activity is concentrated within the sample M&As targeting non-labor firms. This is true 23

25 both in the full sample (Panel A) and our minimum wage sensitive sample (Panel B). Although these findings support the Scale Hypothesis, one caveat to this analysis is that we cannot rule out the possibility that the differential effect of minimum wage on labor M&A targets, compared to non-labor targets, is driven by some other M&A characteristic that is correlated with the target s labor intensity. To properly interpret the evidence in this section it is important to recognize that investment type varies by industry. For instance, non-labor intensive firms spend approximately twice as much relative to their assets on R&D compared to labor-intensive firms. This differential baseline investment can make difference-in-differences coefficients hard to interpret because the point estimate relates to a level change, which may not correspond to a percentage change (see e.g., Roberts and Whited, 2013). To ensure that any differences in the baseline rate of the various types of investment do not affect our inferences, Columns 1 through 3 of Tables A7 and A8 of the Internet Appendix replicate Tables 6 and 7 using logged dependent variables. The results between the two sets of tables are qualitatively similar, suggesting that our inferences are not materially affected by the fact that the baseline rate of each investment component varies across our treatment and control samples. Throughout Tables 6 and 7 is that there is little evidence of a positive relation between minimum wage changes and any form of investment. Put differently, the Scale Hypothesis dominates the Substitution Hypothesis across all investment types Further evidence that firms scale down The Scale Hypothesis predicts not only that firms react to minimum wage increases by cutting investment, but also by scaling down their operations. To test this, we use our main triple difference specification to examine the effect of minimum wage increases on firm size, measured as the natural logarithm of total assets. Panel A of Table 8 reports the results for all labor-intensive firms and for the subset of minimum wage sensitive firms, using both our main and our alternate measure of exposure to bound states. Across both samples and all specifications, we find a strong negative relation between minimum wage increases and firm size. The coefficient of in Column 1 of Panel A implies that a 10% increase in minimum wages leads labor-intensive firms to shrink their assets by about 2%; the magnitude is nearly twice as large for minimum wage sensitive firms (Column 2). Consistent with the Scale Hypothesis, firms react to higher minimum wages by scaling back the size of their business. 24

26 This change in business scale requires firms to adjust their debt and/or equity use. In Panels B and C we examine this, using net debt issuance (defined as the annual change in the total book value of debt, scaled by prior year total assets) and net equity issuance (the value of shares sold minus dividends and share repurchases) as dependent variables in our triple differencing specification. These broad measures of financing activity allow us to capture not only debt and equity issuance, but also principle repayments and payouts, all of which are related to the scale of a firm s operations. 19 Panels B and C show that minimum wage increases lead to a reduction in net debt and net equity financing for both labor-intensive and minimum wage sensitive firms. Summing the coefficients in Panels B and C suggests that the combined change in debt and equity financing activity is similar in magnitude to the decline in total assets, suggesting that we are accurately estimating the change in firm scale. Notably, the reductions in debt and equity financing are of similar magnitude. Consistent with this, we find little evidence that firms change their financial leverage following minimum wage increases (untabulated). Overall, Table 8 shows that firms respond to an increase in the minimum wage by scaling back their business operations. 6. Robustness Analyses The evidence in Section 5 is consistent with the Scale Hypothesis being the predominant channel through which minimum wage changes affect corporate investment. As predicted by the Scale Hypothesis, minimum wage changes are negatively related to subsequent investment and the negative relation is larger for (1) minimum wage sensitive firms, and (2) investments that are more likely to complement low-wage workers. To mitigate the unlikely possibility that this collection of results is due to violations in our identifying assumptions we extend our identification strategy along several dimensions. First, we examine whether firms investment response to minimum wage hikes depends on the current economic conditions. Next, we conduct a series of matched sample analyses to mitigate the possibility that differences between our treatment and control firms affect our inferences. Finally, we discuss additional strengths and limitations to our identification strategy When Do Minimum Wage Changes Matter? There is considerable variation in economic conditions surrounding the three sets of federal minimum wage increase in our sample period. The and minimum wage hikes overlapped with recessionary periods, while the hike was in the middle of an 19 Our debt issuance measure also has the advantage of capturing credit line use. 25

27 expansionary period. Here, we split the sample and separately examine the years surrounding the minimum wage hike to investigate whether the manner in which firms respond to minimum wage changes depends on the economic conditions at the time of the minimum wage hike. Columns 1 and 2 of Table 8 restrict the sample to the approximately 40% of observations during the expansionary period between 1992 and We report our main triple difference specification with both our original and alternative measure of bound status. The triple interaction coefficients in both specifications are somewhat larger than the corresponding coefficients using the full sample (see Table 4, Columns 2 and 5, respectively). Columns 3 and 4 show that dropping this expansionary period from the sample results in coefficients that are somewhat smaller than the full sample estimates. However, none of the estimates significantly differ across expansionary and recessionary periods. Thus, we cannot reject the null hypothesis that the effect of minimum wage changes on corporate investment is similar in expansionary and recessionary periods. This makes it unlikely that the timing of federal minimum wage changes within the business cycle is an important driver of our findings Matched Sample Analysis The key identifying assumption underlying our triple difference analysis is that any differences between labor-intensive and non-labor-intensive firms do not change differentially in bound versus unbound states, in a manner that is correlated with the timing of federal minimum wage changes. Under this assumption, the triple differencing specification we employ is well suited to estimate the qualitative corporate response to exogenous changes in minimum wages. In this section, we conduct several matched sample analyses to mitigate the possibility that differences between our treatment and control samples lead to a violation of this assumption Propensity Score Matching We begin with a propensity score matching procedure in which we match firms of a given labor intensity in bound states to observably similar firms of the same labor intensity in unbound states. Specifically, using nearest neighbor matching without replacement, we impose that each member of a match must be of the same labor intensity and then match each treated firm (i.e., firm in an unbound state) to the untreated firm (i.e., firm in a bound state) with the most similar fitted value as generated from the following equation: BBBBBBBBBB iiii = XX it 1 ΛΛ + Labor Intense iiii XX it 1 ΩΩ + ββ Labor Intense iiii + αα ee + ττ tt + εε iiii (7) 26

28 where BBBBBBBBBB iiii is firm i s bound status at time t, XX it 1 are all the firm characteristics that we control for throughout the analysis, and Labor Intense iiii is a firm s labor intensity at time t. In addition, we include Fama-French 49 industry and year fixed effects. In Table A9 in the Internet Appendix, we report the covariate balance of our propensity score matched sample. Many of the observable differences between the non-labor-intensive firms in bound and unbound states reported in Table 1 attenuate. For instance, the matching procedure reduces the two largest differences between the non-labor firms in bound and unbound states in Table 1, tangibility and cash holdings, by approximately 50%. More importantly, Figure 6 shows that this matching procedure results in a sample of firms where investment behavior evolves similarly before minimum wage changes, as evidenced by the parallel solid and dashed lines from year -3 to year -1. In Column 1 of Table 9, we replicate our main triple difference analysis using this matched sample. One consideration in this matched sample analysis is that fixed effects can absorb one member of a match, but not the other. To avoid this problem we use match-level fixed effects. We continue to find that firms significantly decrease investment following minimum wage increases. The magnitude of the effect is somewhat larger, but statistically indistinguishable, from the effect estimated over the full sample. The robustness of our full sample estimates to a matched sample makes it unlikely that our findings are due to differences in the types of firms housed in bound and unbound states. Nonetheless, our identification strategy still relies on the assumption that the differences between labor-intense and non-labor-intense firms do not change differentially surrounding federal minimum wage changes for firms in bound versus unbound states Matching on Geographical Proximity The last federal minimum wage change occurred during the 2008 financial crisis. There is evidence that the severity of the financial crisis varied somewhat by region (Clemens and Wither, 2016). 20 This will not directly affect our analysis since we include state x year fixed effects. However, the crisis could bias our inferences if it affected the relative condition of labor and nonlabor firms differently in bound and unbound states. The evidence in Table 8 that the effect of minimum wage on investment is similar in expansions and recessions casts doubt on this specific 20 The evidence actually suggests that the housing crisis was most severe in unbound states which should bias us against finding a decline in investment and debt use in bound states. Perhaps unsurprisingly, our results are robust to excluding the crisis. 27

29 story. However, a more general story whereby the relative condition of labor and non-labor firms evolves differently in bound and unbound states remains a possibility, especially given the geographical concentration of bound states in the U.S. heartland. We next address this possibility using a geographical matching procedure. We first geocode the address of the firm s headquarters using the street-level geocoding datasets provided by SAS Maps Online. This procedure returns the latitude and longitude of the street that the company is located on. If we are not able to match the firm s street-level address, we geocode based on the centroid of the zip code (plus four additional digits when available) of the headquarters address. We use the latitude and longitude of the firm s headquarters to calculate the distance between each firm in our sample. We then match each treated firm-year (i.e., firm-year in an unbound state) to all untreated firm-years (i.e., firm-years in bound states) in the same Fama- French 49 industry that are within a set radius of the treated firm s headquarters. We then estimate our main triple difference specification using match-level fixed effects and weighting each matched firm-year by the inverse of the total number of firm-years within the match. This ensures that each match is equally represented in our sample. Conceptually, this framework identifies the effect of minimum wage by comparing firms located along opposite sides of a state border. Since all of the firms are within close proximity of each other, we expect the local economic climate and business environment to be nearly identical. As a result, any differences between the firm s investment behavior are most likely driven by the fact that firms on one side of the border are affected by the minimum wage increase and firms on the other side are not. We report the results in Columns 2-4 of Table 10 for firms located within 200 miles, 150 miles, and 100 miles of each other, respectively. We find a negative relation between minimum wage and investment across all three radiuses, with a coefficient ranging between and In unreported results, we re-estimate our geographical matched sample analysis using only the nearest match firm (rather than all potential matches within the radius) and find similar results. The similarity between these matched sample results and our full sample analyses make it unlikely that our findings are driven by differential economic conditions in unbound and bound states surrounding federal minimum wage changes. 28

30 6.3. Other Considerations Lobbying We also consider the extent to which firm lobbying might introduce differences between labor and non-labor firms in bound and unbound states that vary with federal minimum wage changes. Firms spend a significant amount of money lobbying against minimum wage increases; there are two ways that this lobbying effort could impact our identification. First, a state s bound status might be determined by state-level lobbying. We find no evidence that the probability of a state switching its bound status is a function of firm-level characteristics, even if these characteristics are partitioned by labor intensity (untabulated). This suggests that the success of state-level lobbying against minimum wage is not correlated with firm conditions, which means that state-level lobbying is unlikely to impact our estimates of the effect of minimum wage on investment and financing policy. To the extent that lobbying confounds our estimates, it must be related to federal-level lobbying. Specifically, it must be that the effectiveness of federal-level lobbying is correlated with differences in the investment policies of firms in bound and unbound states. Moreover, these differences must be unique to labor-intensive firms, since we control for state-level economic conditions with state x year fixed effects. We believe that this complicated lobbying story is unlikely for at least four reasons. First, most of the federal-level lobbying is done by trade organizations that represent firms across large geographic areas, making it less likely that their lobbying efforts vary with geographic differences specific to bound vs. unbound states. Second, laws to increase the federal minimum wage have been proposed in every single Congress during our sample period. This suggests that the impetus for the law change does not vary with local economic conditions. Third, anecdotally federal minimum wage increases have been enacted in a bi-partisan way with support across both bound and unbound states. 21 This support from politicians in both bound and unbound states reduces the possibility that lobbying efforts from one type of state are the driving force behind minimum wage changes. Finally, descriptive evidence does not 21 For example, when debating the most recent federal minimum wage change in 2007, Representative Charles Rangel noted, This is a bi-partisan bill providing critical momentum for the bi-partisan effort to raise the minimum wage. Consistent with this, the bill was sponsored by Representative Jim McCrery (R-LA) and Representative George Miller (D-CA). California was an unbound state, but Louisiana was a bound state. Several other representatives from partially bound states (states with minimum wage rates higher than the then current minimum wage, but lower than the newly proposed wage) also played key roles in the legislation. The bill passed the house by a vote of 360 to 45, suggesting that this was indeed a bi-partisan effort supported by representative from all states. 29

31 support the most likely remaining story, which is that when labor-intensive firms in bound states are in poor condition (i.e., when investment is likely to be low) they have less money to lobby against minimum wage increases. In contrast, we find that on average labor-intense firms in bound states invest at a somewhat higher rate than labor-intense firms in unbound states prior to federal minimum wage increases Measuring Minimum Wage Exposure As we have discussed throughout the analysis, precisely measuring a firm s exposure to federal minimum wage changes is an empirical challenge. Our primary measure, Bound, captures this concept with error by setting exposure equal to one for firm s headquartered in a bound stateyear and zero otherwise. This measure will contain measurement error to the extent that all of a firm s employees are spread across state lines. Thus far, we have done two things to mitigate the possibility that this measurement error drives our results. First, we have replicated our main tests on the sample of firms with below median ratios of firm size to state size. Consistent with small firms or firms in big states housing more of their employees within their headquarter state, the effect of minimum wage changes on investment is strongest among in this subsample. A limitation to this analysis is that there may be factors that are correlated with both firm size and the sensitivity of investment to minimum wage changes, which could possibly explain the increased effect of minimum wage changes on investment. For instance, Whited and Wu (2006) provides evidence that small firms are more likely to be financially constrained. Second, we show that our findings are robust to using an alternative measure of bound status, Alt. Bound, which is based on state mentions in SEC filings. A benefit to this measure is that it does not assume that all employees reside in their firm s headquarter state. Two limitations are that the measure is not available throughout our entire sample period and does not directly measure employee location. Although the ability to perfectly measure a firm s exposure to federal minimum wage changes remains a limitation of our analysis, the consistency of our findings across various specifications suggests that noise in our measurement of firm exposure to bound states does not confound our estimates. Here, we conduct a final set of tests that use industry-level establishment data as an alternative measure of investment. 22 We define establishment growth at the state-industry-year 22 We thank Meer and West (2015) for making the data they obtain from the Bureau of Labor Statistics available. 30

32 level as the change in the number of establishments during the year scaled by the beginning of year number of establishments. The benefit to this analysis is that it captures changes in activity within a state, irrespective of whether or not a firm operates across state lines. Two other important differences between this and our previous analyses are that establishment growth is a more focused measure of investment than in our previous tests and this analysis is not restricted to public firms. We continue to label the restaurant, retail, and entertainment industries as minimum wage sensitive. Here, these industries correspond to NAICS industries 44, 45, 71, and 72. Because establishment growth is an especially good measure of investment for minimum wage sensitive industries, we begin by restricting the sample to minimum wage sensitive industries and estimating a difference-in-differences analysis comparing establishment growth rates in minimum wage sensitive industries in bound versus unbound states surrounding federal minimum wage changes. We include state and industry x year fixed effects, along with controls for state economic conditions. The negatively significant Bound X Δ Min. Wage interaction in Column 1 of Table 10 indicates that the establishment growth rate in bound states (compared to unbound states) is significantly smaller following federal minimum wage increases. The magnitude of the effect suggests that establishment growth falls by 1% in the year surrounding a federal minimum wage increase. In Column 2, we obtain an almost identical estimate after introducing non-minimum wage sensitive industries as an additional control group in a triple differencing specification. However, these triple differencing results should be interpreted with caution since establishment growth may not be a representative measure of investment for many non-minimum wage sensitive industries. 7. Conclusion The minimum wage is one of the most frequently revised socio-economic policies in the United States. In January of 2018, 18 states increased their minimum wages. Although an enormous amount of research has been devoted to understanding the labor market effects of minimum wages, almost no work has been done on how minimum wage rates affect first order corporate decisions. We fill this void by providing evidence that minimum wage changes do affect the policies of large public corporations, and not just those in the most minimum wage sensitive industries. Building on existing theory and a stylized model, we motivate two non-mutually exclusive hypotheses regarding the effect of minimum wage increases on corporate investment. The Scale 31

33 Hypothesis predicts a negative relation between minimum wage increases and investment since higher input costs lead to smaller market size, while the Substitution Hypothesis predicts that firms will increase investment in response to higher minimum wage as they attempt to substitute away from labor inputs. To determine which of these hypotheses dominates, we introduce a novel empirical strategy to identify the effect of minimum wage changes on corporate investment. Our findings consistently suggest that the Scale Hypothesis is the predominant channel through which minimum wage changes affect corporate investment. Not only does the investment of labor-intensive firms significantly decline relative to other firms following minimum wage increases, but consistent with the predictions of the Scale Hypothesis this decline is (1) largest for firms that rely heavily on minimum wage labor, (2) concentrated in capital investments (not in investments likely to substitute for labor), and (3) accompanied by a reduction in firm size. Existing policy discussions revolve around raising the minimum wage to $15 per hour. Currently, around 42% of U.S. workers currently make less than this, which implies that the effects that we document in this paper could affect a substantial number of firms. Consequently, our paper suggests that policy makers should carefully consider the effect of minimum wages not only on employees, but also on firms. Our empirical framework is well-suited to identify the short-run effects of minimum wage increases on corporate policy, however future research is needed to conclusively identify longer-run effects. 32

34 References: Aaronson, D. (2001). Price pass-through and the minimum wage. Review of Economics and statistics, 83(1), Aaronson, D., Agarwal, S., & French, E. (2012). The spending and debt response to minimum wage hikes. The American Economic Review, 102(7), Aaronson, D., & French, E. (2007). Product market evidence on the employment effects of the minimum wage. Journal of Labor Economics, 25(1), Aaronson, D., French, E., & MacDonald, J. (2008). The minimum wage, restaurant prices, and labor market structure. Journal of Human Resources, 43(3), Aaronson, D., French, E., Sorkin, I., & To, t. (2017). Industry dynamics and the minimum wage: A putty-clay approach. International Economic Review, forthcoming. Acharya, V. V., Baghai, R. P., & Subramanian, K. V. (2014). Wrongful discharge laws and innovation. Review of Financial Studies, 27(1), Agrawal, A. K., & Matsa, D. A. (2013). Labor unemployment risk and corporate financing decisions. Journal of Financial Economics, 108(2), Akerlof, G. A., & Yellen, J. L. (1990). The fair wage-effort hypothesis and unemployment. The Quarterly Journal of Economics, 105(2), Allegretto, Sylvia, Arindrajit Dube and Michael Reich (2011). Do Minimum Wages Really Reduce Teen Employment? Accounting for Heterogeneity and Selectivity in State Panel Data. Industrial Relations 50, 2: Allegretto, S. A., Dube, A., Reich, M., & Zipperer, B. (2013). Credible research designs for minimum wage studies. Unpublished Working Paper. Arrow, K. J., Chenery, H. B., Minhas, B. S., & Solow, R. M. (1961). Capital-labor substitution and economic efficiency. The Review of Economics and Statistics, Autor, D., Kerr, W., & Kugler, A. (2007). Does employment protection reduce productivity? Evidence from U.S. states. The Economic Journal 117, F189-F217. Bertrand, M., Duflo, E., and Mullainathan, S., (2004). How much should we trust differences-indifferences estimates? Quarterly Journal of Economics 119, Brainard, W. C., & Tobin, J. (1968). Pitfalls in financial model building. The American Economic Review, 58(2), Card, D., & Krueger, A. B. (1994). Minimum Wages and employment: A case study of the fastfood industry in New Jersey and Pennsylvania. The American Economic Review, 84(4), Card, D., & Krueger, A. B. (1995). Time-series minimum-wage studies: a meta-analysis. The American Economic Review, 85(2), Card, D., & Krueger, A. B. (2015). Myth and measurement: the new economics of the minimum wage. Princeton University Press. Chatt, R., Gustafson, M., & Welker, A. (2017). Firing frictions and the mergers and acquisitions market. SSRN Working Paper. 33

35 Clemens, J., & Wither, M. (2016). The Minimum Wage and the Great Recession: Evidence of Effects on the Employment and Income Trajectories of Low-Skilled Workers. NBER Working Paper. David, H., Manning, A., & Smith, C. L. (2016). The contribution of the minimum wage to US wage inequality over three decades: a reassessment. American Economic Journal: Applied Economics, 8(1), Dessaint, O., Golubov, A., & Volpin, P. F. (2017). Employment protection and takeovers. Journal of Financial Economics, 125, DiNardo, J., Fortin, N. M., & Lemieux, T. (1996). Labor Market Institutions and the Distribution of Wages, : A Semiparametric Approach. Econometrica, 64(5), Draca, M., Machin, S., & Van Reenen, J. (2011). Minimum wages and firm profitability. American economic journal: applied economics, 3(1), Dube, A., Lester, T. W., & Reich, M. (2010). Minimum wage effects across state borders: Estimates using contiguous counties. The review of economics and statistics, 92(4), Dube, A., Naidu, S., & Reich, M. (2007). The economic effects of a citywide minimum wage. ILR Review, 60(4), Fairhurst, D., & Serfling, M. (2016). Employment protection, investment, and firm growth. SSRN Working Paper. Garcia, D., & Norli, Ø. (2012). Geographic dispersion and stock returns. Journal of Financial Economics, 106, Giuliano, L. (2013). Minimum wage effects on employment, substitution, and the teenage labor supply: Evidence from personnel data. Journal of Labor Economics, 31(1), Grossman, J. B. (1983). The impact of the minimum wage on other wages. Journal of Human Resources, Hall, R.E., & Jorgenson, D.W. (1967). Tax policy and investment behavior. American Economic Review, 57, Jardim, E., Long, M.C., Plotnick, R., van Inwegen, E., Vigdor, J., & Wething, H. (2017). Minimum Wage Increases, Wages, and Low-Wage Employment: Evidence from Seattle. Working Paper #23532, National Bureau of Economic Research. John, K., Knyazeva, A., & Knyazeva, D. (2015). Employee rights and acquisitions. Journal of Financial Economics, 118(1), Jorgenson, D.W. (1963). Capital theory and investment behavior. The American Economic Review, 53(2), Katz, L. F., & Krueger, A. B. (1992). The effect of the minimum wage on the fast-food industry. Industrial & Labor Relations Review, 46(1), Lee, D. S. (1999). Wage inequality in the United States during the 1980s: Rising dispersion or falling minimum wage?. Quarterly Journal of Economics, Luca, D., & Luca, M. (2017). Survival of the fittest: The impact of the minimum wage on firm exit. Harvard Business School Working Paper #

36 Lucas, R. E. (1967). Optimal investment policy and the flexible accelerator. International Economic Review, 8(1), Lucas, R. E. (1976). Econometric policy evaluation: A critique. In Carnegie-Rochester Conference Series on Public Policy (Vol. 1, pp ). North-Holland. Lucas Jr, R. E., & Prescott, E. C. (1971). Investment under uncertainty. Econometrica, 39, MaCurdy, T. (2015). How effective is the minimum wage at supporting the poor? Journal of Political Economy, 123(2), Magruder, J. (2013). Can Minimum Wages Cause a Big Push? Evidence from Indonesia. Journal of Development Economics 100, 1: Matsa, D. A. (2010). Capital structure as a strategic variable: Evidence from collective bargaining. The Journal of Finance, 65(3), Meer, J., & West, J. (2015). Effects of the minimum wage on employment dynamics. Journal of Human Resources. Neumark, D., & Wascher, W. (2000). Minimum wages and employment: A case study of the fastfood industry in New Jersey and Pennsylvania: Comment. The American Economic Review, 90(5), Rauh, J.D. and Sufi, A. (2011). Explaining corporate capital structure: Product markets, leases, and asset similarity. Review of Finance, 16(1), Roberts, M. & Whited, T. (2013). Endogeneity in empirical corporate finance. In George Constantinides, Milton Harris, and Rene Stulz, eds. Handbook of the Economics of Finance, Vol. 2, Serfling, M. (2016). Firing costs and capital structure decisions. The Journal of Finance. Simintzi, E., Vig, V., & Volpin, P. (2014). Labor protection and leverage. Review of Financial Studies, 28(2), Sorkin, I. (2015). Are there long-run effects of the minimum wage?. Review of economic dynamics, 18(2), Stigler, G. (1946). The economics of minimum wage legislation. The American Economic Review, 36, Tobin, J. (1969). A general equilibrium approach to monetary theory. Journal of Money, Credit and Banking, 1(1), Tonin, M. (2011). Minimum wage and tax evasion: Theory and evidence. Journal of Public Economics, 95(11), Treadway, A. B. (1969). On rational entrepreneurial behaviour and the demand for investment. The Review of Economic Studies, 36(2), Treadway, A. B. (1971). The rational multivariate flexible accelerator. Econometrica, 39, Whited, T. M., & Wu, G. (2006). Financial constraints risk. The Review of Financial Studies, 19,

37 Appendix A: Variable Definitions This Appendix provides definitions and sources for the variables used in our paper. Panel A defines the dependent variables used throughout the analysis, while Panel B defines the explanatory variables. In our empirical analysis, all explanatory variables, except for the annual minimum wage change, are measured one year prior to the end date over with the dependent variable is computed. Panel A: Dependent Variables Variable Name Total Investment CAPEX M&A R&D All SDC M&A Labor M&A Non-Labor M&A Ln (Assets) Debt Issuance Equity Issuance Establishment Growth Rate Definition (Sources) The sum of annual capital expenditure (CAPX), mergers and acquisitions expenses (AQC), and research and development expenses (XRD) scaled by beginning-of-period total assets (AT). (Compustat) Annual capital expenditure (CAPX) scaled by beginning-of-period total assets (AT). (Compustat) Annual mergers and acquisitions expenses (AQC) scaled by beginning-of-period total assets (AT). (Compustat) Annual research and development expenses (XRD) scaled by beginning-of-period total assets (AT). (Compustat) The total dollar value of M&As in Thomson s SDC M&A database in a fiscal year scaled by beginning of year total assets. Excludes M&As labeled as spinoffs, recapitalizations, self-tender or exchange offers, repurchases, privatizations, acquisitions of remaining interest, and leveraged buyouts. (SDC, Compustat) The total dollar value of M&As in Thomson s SDC M&A database with targets in a labor-intensive industry (i.e., a Fama-French 49 industry year with above median employees-to-assets). (SDC, Compustat and Ken French s website) The total dollar value of M&As in Thomson s SDC M&A database with targets in a non-labor-intensive industry (i.e., a Fama-French 49 industry year with below median employees-to-assets). (SDC, Compustat and Ken French s website) The natural logarithm of end-of-period total assets (AT). (Compustat) Inflation adjusted annual change in total debt (DLTT+DLC) scaled by beginning-of-period total assets (AT). (Compustat) Net equity issuance (SSTK-PRSTKC) minus dividends (DVC) scaled by beginning-of-period total assets (AT). (Compustat) Establishment growth is measured at the state-industry-year level and is defined as the change in the number of establishments during the year scaled by the beginning of year number of establishments. (Data as used in Meer and West (2015), which they obtain from the Bureau of Labor Statistics) 36

38 Panel B: Explanatory Variables Variable Name Δ Min. Wage Bound Alt. Bound Labor Definition (Sources) Annual percentage in the nominal federal minimum wage for the year ending at the beginning of the calendar quarter before fiscal year end. (U.S. Department of Labor) An indicator for a firm year that begins with the state minimum wage being equal to or less than the federal minimum wage. (U.S. Department of Labor) The percentage of state mentions in 10-K filings that are of states with minimum wage less than or equal to the federal minimum wage, according to the dataset used in Garcia and Norli (2012) that runs from 1995 through To maintain a consistent sample and because the lack of electronic SEC filings prevents us from computing this measure prior to 1995, firm-years prior to 1995 (after 2008) are extrapolated from the earliest (latest) observation in the dataset. If a firm does not appear in Garcia and Norli (2012) s dataset then Alt. Bound equals Bound. (U.S. Department of Labor, Garcia and Norli, 2012) An indicator for a firm in a Fama-French 49 industry year with above median employees-to-assets (Compustat and Ken French s website). Sensitive An indicator for a firm that is in a minimum wage sensitive industry (FF 49 industries 7, 43, and 44). Note that non-sensitive firms are defined as laborintense firms not in a sensitive industry. (Current Population Survey) Employees Total Liabilities Tangibility Ln(Assets) Profitability MtB Cash Population Δ Population Unemployment Δ Unemployment Average Wage Δ Average Wage Employees (EMP) divided by total assets (AT). (Compustat) Total liabilities (LT) divided by total assets (AT). (Compustat) Net property, plant, and equipment (PPENT) divided by total assets (AT). (Compustat) The natural log of one plus total assets (AT) in 1983 $ millions. (Compustat) Operating income before depreciation (OIBDP) divided by total assets (AT). (Compustat) The value of debt (DLTT+DLC) plus the market value of equity ( PRCC_F *CSHO), divided by total assets (AT). (Compustat) Cash and equivalents (che) divided by cash total assets (at). (Compustat) The total number of residents in a state. (BLS website) The annual percentage change in the total number of residents in a state. (BLS website) The state-level unemployment rate. (BLS website) The annual percentage change in the state-level unemployment rate. (BLS website) The state-level average wage. (BLS website) The annual percentage change in the state-level average wage. (BLS website) 37

39 Figure 1. Minimum Wage Changes Over Time. This chart shows the quarterly evolution of minimum wages in the United States from 1975 to The dotted line plots the average nominal state-level minimum wage for the subset of states that have a minimum wage higher than the federal minimum wage. The dashed line shows the nominal federal minimum wage. The solid line (right axis) shows the number of states that have a state minimum wage higher than the federal minimum wage. Source: U.S. Department of Labor. 38

40 Panel A: Scale Hypothesis Firm s Minimum Wage Exposure Low High Type of Investment Labor Comp. Labor Subs Panel B: Substitution Hypothesis Firm s Minimum Wage Exposure Low High Type of Investment Labor Comp. Labor Subs Figure 2. Empirical Predictions. This Figure presents empirical predictions for the two non-mutually exclusive hypotheses regarding the effect of minimum wage on corporate investment. Panel A describes the predictions of the Scale Hypothesis, while Panel B does the same for the Substitution Hypothesis. Each panel is split into four quadrants, based on whether a firm has high or low sensitivity to minimum wage workers and whether the investment is a substitute or complement with labor. The magnitudes entered into each quadrant are defined relative to the other quadrants within each panel (i.e., +++ > ++, but the relation between + and - is unknown). 39

41 Figure 3. Bound Vs. Unbound States. For each state, this figure shows the percent of quarters from 1987 to 2012 for which the state is bound by the federal minimum wage; i.e. the state-level minimum wage is less than or equal to the federal minimum wage or the state has no state minimum wage law. Source: U.S. Department of Labor. 40

42 Figure 4. Descriptive evidence on the effect of minimum wage on firm investment in event time. This figure presents average investment in event time for firms of different labor intensity and minimum wage sensitivity. The sample is restricted to observations that are within three years before or two years after each of the three federal minimum wage increases (i.e., events). Each of the three events consists of consecutive federal minimum wage increases. Thus, federal minimum wages increase between year -1 and year 1 (year 0 is omitted) and between year 1 and year 2. To be included in the sample a firm must exist for at least four of the five years in the event window. Each point represents the differential investment of firms in bound versus unbound states for firms of a given labor intensity or minimum wage sensitivity, after controlling for year x quarter and event firm fixed effects as well as all control variables used in our difference-in-differences analysis (i.e., Employees, Liabilities, Tangibility, Ln(Assets), Profitability, MtB, and Cash, population, change in population, unemployment, change in unemployment, state-level average wage, change in state-level average wage). The solid line includes firms in industry years with above median employee-to-assets ratios, which we denote as labor intensive. The short-dash line contains non-labor intensive firms by this same definition. Finally, the long-dashed line includes firms in the most minimum wage sensitive industries (i.e., restaurant, retail, and entertainment). 41

43 Figure 5. Average Annual Investment over Time for Labor- and Non-Labor-Intensive Firms. This figure presents average investment, defined as the sum of capital, M&A, and R&D expenditures scaled by beginning of period total assets, for each year of our sample period. The solid line plots the average investment of labor-intensive firms, while the dashed line plots the same average for non-labor-intensive firms. We define labor-intensive firms as those in industry-years with above median employee-to-assets ratios. 42

44 Figure 6. Matched Sample Trends in Labor and Non-Labor Corporate Investment surrounding Federal Minimum Wage Changes. This figure presents average investment in event time for firms of different labor intensity. The sample is restricted to observations that are within three years before or two years after each of the three federal minimum wage increases (i.e., events). Each of the three events consists of consecutive federal minimum wage increases. Thus, federal minimum wages increase between year -1 and year 1 (year 0 is omitted) and between year 1 and year 2. The sample is also restricted to our matched sample, which matches all (non-)labor-intensive firm in a bound state to an observably similar firm of the same labor intensity in an unbound state using the propensity score matching procedure described in Equation 7. Each point represents the differential investment of firms in bound versus unbound states for firms of a given labor intensity, after controlling for year x quarter and event firm fixed effects as well as all control variables used in our difference-in-differences analysis (i.e., Employees, Liabilities, Tangibility, Ln(Assets), Profitability, MtB, and Cash, population, change in population, unemployment, change in unemployment, state-level average wage, change in state-level average wage). The solid line includes firms in industry years with above median employee-to-assets ratios, which we denote as labor intensive. The long-dash line contains non-labor intensive firms by this same definition. 43

45 Table 1: Sample Summary Statistics This table presents descriptive statistics for the firm-level characteristics used in our regression analysis. Definitions for the variables shown in this table are found in Appendix A. Mean Median Std. Dev. Min Max Bound Δ Min. Wage Total Investment CAPEX M&A R&D Employees Total Liabilities Tangibility Ln(Assets) Profitability Ln(MtB) Cash Observations 106,340 44

46 Table 2: Minimum Wage and Corporate Investment, Difference-in-Differences Estimates This table presents OLS difference-in-differences estimates of the relation between minimum wage changes and total investment and CAPEX. Total investment is defined as the sum of capital, M&A, and R&D expenditures scaled by beginning of period total assets. CAPEX is defined similarly, except that it excludes M&A and R&D expenditures. The explanatory variable of interest is the interaction between bound state-years (i.e., those with state-level minimum wages equal to the federal minimum wage) and federal minimum wage changes. Panel A includes the full sample. Panel B restricts the sample to labor-intensive firms (Columns 1 and 2) and Minimum wage sensitive firms (Columns 3 and 4). Panel C restricts the sample to non-labor intensive firms. (Non-)Labor-intensive firms are those in industryyears in the top (bottom) half of median employees-to-asset ratios. Minimum wage sensitive firms are those in the retail, restaurant, and entertainment industries. All columns include controls for firm characteristics (Employees, Liabilities, Tangibility, Ln(Assets), Profitability, MtB, and Cash) and state economic conditions (population, change in population, unemployment, change in unemployment, state-level average wage, change in state-level average wage) as well as firm and year fixed effects. Appendix A defines all control variables, t-statistics based on standard errors that are clustered by state are reported in parentheses below the coefficients, and *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. Panel A: Full Sample Total CAPEX Investment (1) (2) Bound X Δ Min. Wage (-0.70) * (-1.77) Bound (0.80) 0.003** (2.16) Firm Controls YES YES State Economic Conditions YES YES Firm Fixed Effects YES YES Year Fixed Effects YES YES Adj. R-squared Observations 106, ,340 Panel B: Labor-Intensive and Minimum Wage Sensitive Firms Labor-Intensive Firms Min. Wage Sensitive Firms Total CAPEX Total CAPEX Investment Investment (1) (2) (3) (4) Bound X Δ Min. Wage ** (-2.47) ** (-2.16) (-0.73) ** (-2.20) Bound (1.17) (0.69) (0.06) (-0.22) Firm Controls YES YES YES YES State Economic Conditions YES YES YES YES Firm Fixed Effects YES YES YES YES Year Fixed Effects YES YES YES YES Adj. R-squared Observations 52,890 52,890 11,191 11,191 45

47 Panel C: Non-Labor-Intensive Firms Total CAPEX Investment (1) (2) Bound X Δ Min. Wage (0.75) (-0.14) Bound (0.10) 0.003** (2.38) Firm Controls YES YES State Economic Conditions YES YES Firm Fixed Effects YES YES Year Fixed Effects YES YES Adj. R-squared Observations 52,686 52,686 46

48 Table 3: Average Firm Characteristics, Partitioned by Bound and Labor Intensity This table presents the mean for each firm-level characteristics used in our regression analysis, partitioned by pabor intensity and bound status. Columns 1 and 2 present averages for labor-intensive firms in unbound and bound states, respectively. Columns 3 and 4 present firms in minimum wage sensitive industries, while Columns 5 and 6 present averages for non-labor intensive firms. States are defined as bound if the state-level minimum wage is less than or equal to the federal minimum wage or if the state does not have a state minimum wage. We classify firms into labor and non-labor based on the employees-to-assets ratio. For each industry year, we calculate the median employees-toassets ratio. We then calculate the median across all industry-year medians; firms are classified as labor-intense if their employees-to-assets ratio is greater than this sample median. We define minimum wage sensitive industries as the restaurant, retail, and entertainment industries. Definitions for the variables shown in this table are found in Appendix A. Labor-Intensive Min. Wage Sens. Non-Labor-Intensive Unbound Bound Unbound Bound Unbound Bound (1) (2) (1) (2) (5) (6) Total Investment Employees Total Liabilities Tangibility Ln(Assets) Profitability MtB Cash Observations 10,760 42,480 2,480 8,711 19,040 34,060 47

49 Table 4: Labor Intensity, Minimum Wage Changes, and Total Investment This table presents OLS estimates where the dependent variable is total investment, defined as the sum of capital, M&A, and R&D expenditures scaled by beginning of period total assets. The explanatory variable of interest is the triple interaction between a bound state, a federal minimum wage change, and a labor-intensive firm, defined as a firm in an industry-year with an above median employees-to-assets ratio. In Columns 1 through 4 Bound is an indicator for a firm headquartered in a state-year with minimum wage equal to the federal minimum wage. In Column 5 we use an alternate bound measure (Alt. Bound), based on the geographical dispersion measure in Garcia and Norli (2012), which counts the percentage of state mentions in a firms 10-K filings that refer to a bound state. Δ Min. Wage is the annual percentage change in federal minimum wage ending one quarter before the end of the fiscal year over which investment is measured. All columns include firm and state-year fixed effects, while Columns 2 through 5 also include controls for firm characteristics. Column 3 (4) restricts the sample to firm-years in the bottom (top) two quartiles of the ratio of the firm s total assets to the headquarter state s size. Appendix A defines all control variables, t-statistics based on standard errors that are clustered by state are reported in parentheses below the coefficients, and *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. Dependent Variable: Total Investment All Sized Firms Small Firms Large Firms Alt. Bound (1) (2) (3) (4) (5) Bound X Δ Min. Wage X Labor ** (-2.22) *** (-2.85) *** (-3.51) (-1.23) ** (-2.31) Bound X Δ Min. Wage (0.13) (0.92) 0.141*** (2.95) (-0.30) 0.135* (1.90) Bound X Labor 0.012*** (2.89) 0.015*** (3.72) 0.027*** (5.52) (0.58) 0.015*** (2.68) Δ Min. Wage X Labor (1.02) 0.063** (2.18) 0.124*** (2.95) (0.41) 0.091* (1.90) Labor (-0.86) * (-1.77) *** (-2.71) (0.76) (-0.84) Bound 0.008* (1.98) (0.19) * (-1.70) 0.014** (2.19) (-0.32) Employees 1.146*** (8.56) 1.264*** (8.35) 0.755*** (3.78) 1.144*** (8.63) Total Liabilities *** (-4.98) ** (-2.21) *** (-13.89) *** (-5.00) Tangibility 0.062*** (5.65) 0.038*** (3.12) 0.069*** (4.04) 0.062*** (5.64) Ln(Assets) *** (-22.49) *** (-19.06) *** (-15.73) *** (-22.46) Profitability (-0.46) * (-1.72) 0.074*** (6.57) (-0.46) Ln(MtB) 0.027*** (34.12) 0.026*** (31.45) 0.020*** (9.32) 0.027*** (34.19) Cash 0.050*** (5.42) 0.039*** (4.08) 0.104*** (8.27) 0.050*** (5.43) Firm Fixed Effects YES YES YES YES YES State Year Fixed Effects YES YES YES YES YES Adj. R-squared Observations 106, ,340 52,549 52, ,340 48

50 Table 5: Minimum Wage Sensitivity, Minimum Wage Changes, and Total Investment The ordinary least squares dependent variable is investment, defined as the sum of capital, M&A, and R&D expenditures scaled by beginning of period total assets. In Panel A, the sample contains only Sensitive firms, defined as those in the most minimum wage sensitive industries (restaurant, retail, and entertainment) and non-labor-intensive firms (those in industry-years with below median employees-to-assets ratios). Here, the explanatory variable of interest is the triple interaction between a bound state, a federal minimum wage change, and a minimum wage sensitive firm. In Columns 1 through 3 Bound is an indicator for a firm headquartered in a state-year with minimum wage equal to the federal minimum wage. In Column 4 we use an alternate bound measure (Alt. Bound), based on the geographical dispersion measure in Garcia and Norli (2012), which counts the percentage of state mentions in a firms 10-K filings that refer to a bound state. Δ Min. Wage is the annual percentage change in federal minimum wage ending one quarter before the end of the fiscal year over which investment is measured. In Panel B, the sample excludes Sensitive firms, defined as those in the most minimum wage sensitive industries (restaurant, retail, and entertainment). Here, the explanatory variable of interest is the triple interaction between a bound state, a federal minimum wage change, and a labor-intensive firm. All columns include firm and state-year fixed effects, the same controls for firm characteristics presented in Table 4, and the complete set of effects and interactions that comprise the explanatory variable of interest. Column 2 (3) restricts the sample to firm-years in the bottom (top) two quartiles of the ratio of the firm s total assets to the headquarter state s size. Appendix A defines all control variables, t-statistics based on standard errors that are clustered by state are reported in parentheses below the coefficients, and *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. Panel A: Minimum Wage Sensitive Firms, Compared to Non-Labor Firms Dependent Variable: Total Investment All Sized Firms Small Firms Large Firms Alt. Bound (1) (2) (3) (4) Bound X Δ Min. Wage X Sensitive *** (-3.21) *** (-3.00) (-1.02) *** (-3.21) Firm Controls YES YES YES YES Firm Fixed Effects YES YES YES YES State Year Fixed Effects YES YES YES YES Adj. R-squared Observations 63,347 29,854 32,714 63,347 Panel B: Non-Minimum Wage Sensitive Labor Firms, Compared to Non-Labor Firms Dependent Variable: Total Investment All Sized Firms Small Firms Large Firms Alt. Bound (1) (2) (3) (4) Bound X Δ Min. Wage X Labor ** (-2.41) *** (-3.03) (-1.07) ** (-2.01) Firm Controls YES YES YES YES Firm Fixed Effects YES YES YES YES State Year Fixed Effects YES YES YES YES Adj. R-squared Observations 95,140 48,251 45,947 95,140 49

51 Table 6: Partitioning on Investment Type The ordinary least squares dependent variable is capital (Column 1), R&D (Column 2), or M&A (Column 3), or expenditures scaled by beginning of period total assets. Panel A includes the full sample and the explanatory variable of interest is the triple interaction between a bound state, a federal minimum wage change, and a labor-intensive firm (i.e., a firm in an industry-year with above median employees-to-assets ratios). Bound is an indicator for a firm headquartered in a state-year with minimum wage equal to the federal minimum wage and Δ Min. Wage is the annual percentage change in federal minimum wage ending one quarter before the end of the fiscal year over which investment is measured. In Panel B, the sample contains only Sensitive firms, defined as those in the most minimum wage sensitive industries (restaurant, retail, and entertainment) and non-labor-intensive firms (those in industry-years with below median employees-to-assets ratios). Here, the explanatory variable of interest is the triple interaction between a bound state, a federal minimum wage change, and a minimum wage sensitive firm (i.e., a firm in the restaurant, retail, or entertainment industries). All columns include firm and state-year fixed effects, the same controls for firm characteristics presented in Table 4, and the complete set of effects and interactions that comprise the explanatory variable of interest. Appendix A defines all control variables, t-statistics based on standard errors that are clustered by state are reported in parentheses below the coefficients, and *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. Panel A: All Labor Firms, Compared to Non-Labor Firms Dependent Variable: CAPEX R&D M&A (1) (2) (3) Bound X Δ Min. Wage X Labor * (-1.74) (-0.23) * (-1.76) Firm Controls YES YES YES Firm Fixed Effects YES YES YES State Year Fixed Effects YES YES YES Adj. R-squared Observations 106, , ,340 Panel B: Minimum Wage Sensitive Firms, Compared to Non-Labor Firms Dependent Variable: CAPEX R&D M&A (1) (2) (3) Bound X Δ Min. Wage X Sensitive *** (-3.58) (-0.38) ** (-2.05) Firm Controls YES YES YES Firm Fixed Effects YES YES YES State Year Fixed Effects YES YES YES Adj. R-squared Observations 63,347 63,347 63,347 50

52 Table 7: Decomposing the Effect of Minimum Wage Changes on M&A Activity The ordinary least squares dependent variable is SDC M&A (Column 1), Labor M&A (Column 2), or Non-Labor M&A (Column 3) expenditures scaled by beginning of period total assets. Labor M&As are those with targets in labor-intensive industries (i.e., a firm in an industry-year with above median employees-to-assets ratios). All other M&As are Non-Labor M&As. As we discuss in Appendix A, all dependent variables are defined using data from Thomson s SDC M&A database. Panel A includes the full sample and the explanatory variable of interest is the triple interaction between a bound state, a federal minimum wage change, and a labor-intensive firm. Bound is an indicator for a firm headquartered in a state-year with minimum wage equal to the federal minimum wage and Δ Min. Wage is the annual percentage change in federal minimum wage ending one quarter before the end of the fiscal year over which investment is measured. In Panel B, the sample contains only Sensitive firms, defined as those in the most minimum wage sensitive industries (restaurant, retail, and entertainment) and non-labor-intensive firms (those in industry-years with below median employees-to-assets ratios). Here, the explanatory variable of interest is the triple interaction between a bound state, a federal minimum wage change, and a minimum wage sensitive firm (i.e., a firm in the restaurant, retail, or entertainment industries). All columns include firm and state-year fixed effects, the same controls for firm characteristics presented in Table 4, and the complete set of effects and interactions that comprise the explanatory variable of interest. Appendix A defines all control variables, t-statistics based on standard errors that are clustered by state are reported in parentheses below the coefficients, and *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. Panel A: All Labor Firms, Compared to Non-Labor Firms Dependent Variable: All SDC M&A Labor M&A Non-Labor M&A (1) (2) (3) Bound X Δ Min. Wage X Labor ** (-2.16) (0.15) *** (-4.15) Firm Controls YES YES YES Firm Fixed Effects YES YES YES State Year Fixed Effects YES YES YES Adj. R-squared Observations 106, , ,340 Panel B: Minimum Wage Sensitive Firms, Compared to Non-Labor Firms Dependent Variable: All SDC M&A Labor M&A Non-Labor M&A (1) (2) (3) Bound X Δ Min. Wage X Sensitive ** (-2.21) (-0.65) *** (-3.56) Firm Controls YES YES YES Firm Fixed Effects YES YES YES State Year Fixed Effects YES YES YES Adj. R-squared Observations 63,347 63,347 63,347 51

53 Table 8: Labor Intensity, Minimum Wage Changes, and Capital Structure The ordinary least squares dependent variable is the natural logarithm of total assets (Panel A), debt issuance, defined as the difference between current and lagged book value of debt (Panel B), and equity issuance (Panel C). Debt and equity issuance are scaled by beginning of period total assets. Columns 1 and 2 include the full sample, while Columns 3 and 4 are limited to the subsample of Sensitive firms, defined as those in the most minimum wage sensitive industries (restaurant, retail, and entertainment) and non-labor-intensive firms (those in industry-years with below median employees-to-assets ratios). The explanatory variable of interest is the triple interaction between a bound state, a federal minimum wage change, and a labor-intensive (or sensitive) firm. In columns 1 and 3, we measure bound exposure based on the state of the firm s headquarters, while in Columns 2 and 4 we measure exposure based on the geographical dispersion measure in Garcia and Norli (2012), which counts the percentage of state mentions in a firms 10-K filings that refer to a bound state. Δ Min. Wage is the annual percentage change in federal minimum wage ending one quarter before the end of the fiscal year over which investment is measured. All columns include firm and stateyear fixed effects, the same controls for firm characteristics presented in Table 4, and the complete set of effects and interactions that comprise the explanatory variable of interest. Appendix A defines all control variables, t-statistics based on standard errors that are clustered by state are reported in parentheses below the coefficients, and *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. Panel A: Dependent Variable is Ln(Total Assets) All Labor Firms Min. Wage Sensitive Firms Bound Alt. Bound Bound Alt. Bound (1) (2) (3) (4) Bound X Δ Min. Wage X Labor ** (-2.35) *** (-3.20) *** (-3.11) *** (-3.28) Firm Controls YES YES YES YES Firm Fixed Effects YES YES YES YES State Year Fixed Effects YES YES YES YES Adj. R-squared Observations 106, ,340 63,347 63,347 Panel B: Dependent Variable is Debt Issuance All Labor Firms Min. Wage Sensitive Firms Bound Alt. Bound Bound Alt. Bound (1) (2) (3) (4) Bound X Δ Min. Wage X Labor *** (-4.57) *** (-2.73) *** (-3.01) (-1.64) Firm Controls YES YES YES YES Firm Fixed Effects YES YES YES YES State Year Fixed Effects YES YES YES YES Adj. R-squared Observations 106, ,340 63,347 63,347 52

54 Panel C: Dependent Variable is Equity Issuance All Labor Firms Min. Wage Sensitive Firms Bound Alt. Bound Bound Alt. Bound (1) (2) (3) (4) Bound X Δ Min. Wage X Labor (-0.03) ** (-2.18) ** (-2.06) ** (-2.16) Firm Controls YES YES YES YES Firm Fixed Effects YES YES YES YES State Year Fixed Effects YES YES YES YES Adj. R-squared Observations 106, ,340 63,347 63,347 53

55 Table 9: Economic Conditions and the Effect of Minimum Wage Changes on Investment This table presents OLS estimates where the dependent variable is total investment, defined as the sum of capital, M&A, and R&D expenditures scaled by beginning of period total assets. Columns 1-2 (3-4) restrict the sample to firm-years between 1992 and 2000 (prior to 1992 and after 2000). The explanatory variable of interest is the triple interaction between a bound state, a federal minimum wage change, and a labor-intensive firm, defined as a firm in an industry-year with an above median employees-to-assets ratio. In Columns 1 and 3 Bound is an indicator for a firm headquartered in a state-year with minimum wage equal to the federal minimum wage. In Columns 2 and 4 we use an alternate bound measure (Alt. Bound), based on the geographical dispersion measure in Garcia and Norli (2012), which counts the percentage of state mentions in a firms 10-K filings that refer to a bound state. Δ Min. Wage is the annual percentage change in federal minimum wage ending one quarter before the end of the fiscal year over which investment is measured. All columns include firm and state-year fixed effects, and controls for firm characteristics. Appendix A defines all control variables, t-statistics based on standard errors that are clustered by state are reported in parentheses below the coefficients, and *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. Dependent Variable: Total Investment Expansion Expansion Recession Recession (1) (1) (2) (2) Bound X Δ Min. Wage X Labor (-1.34) * (-1.78) Alt. Bound X Δ Min. Wage X Labor ** (-2.29) * (-1.87) Firm Controls YES YES YES YES Firm Fixed Effects YES YES YES YES State Year Fixed Effects YES YES YES YES Adj. R-squared Observations 41,557 41,557 63,752 63,752 54

56 Table 10: Matched Analyses This table replicates our main triple difference analysis using a matched sample, with total investment as the dependent variable. Firm-years are classified as labor-intense if their industry-year employees-to-assets ratio is greater than the sample median. States are defined as bound if the state-level minimum wage is less than or equal to the federal minimum wage or if the state does not have a state minimum wage. In column 1, we present estimates using a propensity score matched sample. To construct our matched sample we using nearest neighbor matching without replacement, imposing that each member of a match must be of the same labor intensity. We then match each treated firm (i.e., firm in an unbound state) to the untreated firm (i.e., firm in bound state) with the most similar fitted value as generated from Equation 7. We report the covariate balance in Table A9 in the Internet Appendix. Columns 2 through 4 present results using three different matched samples based on the geographical proximity of firm headquarters. Specifically, we match each firm-year located in an unbound state to all firm-years located in a bound state that are within a set geographic radius and that operate in the same Fama-French 49 industry. Because some firms have more matches within the radius than others, we weight each match by the inverse of the total number of matched firms so that each match is equally weighted in the regression. Columns 2, 3, and 4 restrict the sample to matches in which the headquarters are located within 200, 150, and 100 miles from each other, respectively. All columns include the set of firm- and state-level control variables used in Table 2, as well as matched pair fixed effects and the complete set of effects and interactions that comprise the triple interaction. Appendix A defines all control variables, t-statistics based on standard errors that are clustered by state are reported in parentheses below the coefficients, and *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. Bound Propensity Within 200 Miles Within 150 Miles Within 100 Miles (1) (2) (3) (4) Bound X Δ Min. Wage X Labor ** (-2.35) *** (-3.45) *** (-4.72) *** (-5.37) Firm Controls YES YES YES YES State-level Controls YES YES YES YES Match Fixed Effects YES YES YES YES Adj. R-squared Number of Matches 25,756 13,520 10,263 8,232 55

57 Table 11: Minimum Wage Sensitivity and Industry Establishment Growth This table presents OLS estimates where the dependent variable is the establishment growth rate in a given industryyear (i.e., (year-end establishments beginning of year establishments) / beginning of year establishments). Column 1 restricts the sample to minimum wage sensitive industries, which we define as firms in the restaurant, retail, and entertainment industries (i.e., two-digit NAICS classifications 44, 45, 71, and 72). Here, the explanatory variable of interest is the interaction between bound state-years (i.e., those with minimum wages equal to the federal minimum wage) and federal minimum wage changes. Column 2 adds the other two-digit NAICS industries as a control group in a triple differencing framework. The explanatory variable of interest is the triple interaction between a bound state, a federal minimum wage change, and a minimum wage sensitive industry. Bound is an indicator for a state-year with minimum wage equal to the federal minimum wage, and Δ Min. Wage is the annual percentage change in federal minimum wage ending one quarter before the end of the fiscal year over which investment is measured. All columns include industry year and either state fixed effects and economic controls (Column 1) or state-year fixed effects (Column 2). T-statistics based on standard errors that are clustered by state are reported in parentheses below the coefficients, and *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively. Dependent Variable: Establishment Growth Rate Diff-in-Diff Triple Diff (1) (2) Bound X Δ Min. Wage X Sensitive *** (-5.35) Bound X Δ Min. Wage *** (-3.92) Bound X Sensitive (0.13) Bound 0.011* (1.98) State Fixed Effects YES NO State Economic Controls YES NO State x Year Fixed Effects NO YES Industry x Year Fixed Effects YES YES Adj. R-squared Observations 3,519 23,403 56

58 Internet Appendix to Minimum Wage and Corporate Policy Matthew Gustafson and Jason Kotter 57

59 1. Extensions to Conceptual framework The stylized model in Section 2 of the paper suggests that the effect of minimum wage on investment is an empirical question, the answer to which depends on both the sensitivity of a firm s wage bill to minimum wage changes and the type of investment under consideration. Here, we show that these conclusions hold in two settings that are more complex. First, we allow firms to pass on some of the costs of minimum wage in the form of higher prices. Second, we derive the conditions under which a firm that simultaneously invests in two different types of investment changes its investment mix in response to wage increases. When firms can adjust prices after a shock to minimum wage, price is a decreasing function of output. The total derivative of the first order condition becomes: KK = LL ppff KKKK + ff KK ff LL ppff KKKK + ff KK ff KK (IA 1) As in the main paper, diminishing returns to capital and downward sloping labor demand make both ff KKKK and LL negative, and ff KK and ff LL are both positive because they are normal inputs in the production process. As long as the demand curve for the product is not upward sloping, 0, the denominator of Eq (IA 1) will be negative. Consequently, the sign of KK is determined by the numerator of Eq. (IA 1). If ff KKKK is negative (i.e., capital and labor are gross substitutes) then the numerator is negative and the same result that we obtain in the main paper holds: the Substitution Hypothesis dominates and an increase in wages leads to an increase in investment. In contrast, if capital and labor are gross complements ( ff KKKK > 0), then the relation between capital investment and wages is indeterminate: an increase in wages can lead investment to increase, decrease, or stay the same. While we cannot determine the sign in general, we gain insight by examining the conditions when investment increases or decreases. If ppff KKKK < ff KK ff LL, then KK > 0 and the Substitution Hypothesis dominates even if capital complements labor. This will be more likely if capital and labor are only weakly complementary or if product demand is very inelastic. In contrast, the Scale Hypothesis dominates if capital and labor are especially complementary or product market demand is sufficiently elastic. This discussion illustrates that both the Scale and 58

60 Substitutions Hypotheses remain possible even after we allow firms to pass on some of the costs of minimum wage in the form of higher prices. Next, we consider the case where a firm simultaneously invests in two types of capital: capital that is complementary to minimum wage labor and capital that substitutes for minimum wage labor. To do this, we assume a nested constant elasticity of substitution (CES) production function. The CES production function is a generalized function that includes Cobb-Douglas as a special case and the nested structure allows the elasticities of substitution to both vary and be asymmetric across the three inputs (Sato, 1967). Formally, consider a firm that produces output y such that y = α(γl ν + (1 γ)k ν ) ρ\ν + (1 α)r ρ 1\ρ (IA 2) with 0 < {αα, γγ} < 1; {ρ, ν} 1 where K is production capital, L is labor, and R is technology capital. We are interested in the response of investment in K and R to an increase in the cost of labor, which is reflected in the elasticity of substitution, defined as the percentage change in demand for that investment-type for a percentage change increase in the wage. If this elasticity is greater than one, then the inputs are gross substitutes and an increase in minimum wage leads to higher levels of investment. In contrast, if the elasticity is less than one, the inputs are gross complements and an increase in minimum wage causes a reduction in investment. Given Eq. (IA 2), the elasticity of substitution between capital and labor is σσ LLLL = 1 (1 νν). Empirical evidence suggests that σσ LLLL < 1, or that capital and labor are gross complements. Similarly, define the elasticity of substitution between the combined capital/labor input and technology as σσ {LL,KK}RR = 1 (1 ρρ). We assume that capital is more laborcomplementary than technology, so that σσ LLLL < σσ {LL,KK}RR. This corresponds with the idea underlying our framework in Section 2.3 that some investments are more labor-complementary, while other investments are more likely to substitute for labor. σσ LLLL and σσ {LL,KK}RR are both Hicks-McFadden elasticities (upon which the CES model is built), which do not allow the firm to substitute across all inputs. This is an important limitation, since an increase in wages might lead firms to substitute out of labor into or out of both types of investment simultaneously. Consequently, we need an elasticity measure that allows the firm to respond to increases in wages by potentially adjusting labor, capital, and technology. The Morishima (1967) 59

61 elasticity is a generalization of the Hicks-McFadden elasticity that allows for this simultaneous substitution between all of the inputs of production. An additional benefit of the Morishima measure is that it allows for asymmetric elasticities between any given pair of inputs, which makes it particularly suited to describe nested CES production functions. Anderson and Moroney (1993) derive the Morishima elasticities for a generalized nested production function. Applying their derivation to the nested CES model described in Eq. (1), we find that: MM LLLL = σσ LLLL (IA 3) MM LLLL = γγσσ {LL,KK}RR + (1 γγ)σσ LLLL (IA 4) where MM iiii is the Morishima elasticity between goods i and j for a percentage change in the price of good i. Recall that γγ (0,1) and that we have assumed that capital is more complementary to labor than technology, so that σσ LLLL < σσ {LL,KK}RR. Consequently, Eq. (IA 4) is a convex combination and so MM LLLL < MM LLLL which means that an increase in wages leads firms to cut investment in capital relative to investment in technology. This is analogous to the results presented in Section 2; the model predicts that the largest cuts investment are for capital-types that are most complementary to labor. To further interpret Eq. (IA 4), it is necessary to make one additional assumption. Consistent with our discussion in Section 2, assume that technology and the combined output of labor and capital are gross substitutes, σσ {LL,KK}RR > 1. Given this, Eq. (IA 4) shows that an increase in wages can lead firms to increase investments in technology (MM LLLL > 1), decrease investments (MM LLLL < 1), or hold technology investments constant (MM LLLL = 1). Investment in technology is more likely to increase when technology is a better substitute for the labor/capital input and when labor is a more important input in production (γγ is bigger). Conversely, investment in technology is more likely to fall when labor and capital are more strongly complimentary and when capital is a more important input in production. This result stands in contrast to the general framework presented in the paper where a minimum wage increase always causes investment in labor-substitutable capital to rise, and is a good illustration of the importance of allowing the firm to simultaneously choose investment in all inputs. As the price of labor rises, the firm does tend to substitute technology for labor, particularly when labor is an important part of production. However, at the same time, the increase in wages increases the relative price of capital (since it is complementary to labor), which causes the firm 60

62 to reduce capital investment. This reduction in capital makes it harder to substitute technology for labor, since the technology now needs to replace the outputs of both labor and capital. This reduces the incentives to invest in technology. Combined, the net effect on technology investment is ambiguous. Nevertheless, this more complex model supports the idea that the effect of minimum wage on corporate investment is an empirical question, the answer to which depends (at least in part) on how reliant a firm is on minimum wage labor and the type of investment under consideration. 61

63 References: Anderson, R. K., and Moroney, J. R. (1993). Morishima elasticities of substitution with nested production functions. Economics Letters, 42, Morishima, M. (1967). A few suggestions on the theory of elasticity. Keizai Hyoron (Economic Review), 16, Sato, K. (1967). A two-level constant-elasticity-of-substitution production function. The Review of Economic Studies, 34,

64 Figure A1. Sample Composition by State. This chart shows the number of firm years in our regression sample for each state. 63