Labor adjustment costs and asymmetric cost behavior: An extension

Abstract

The issue of asymmetric cost behavior has attracted significant interest in the managerial accounting literature. The literature has hypothesized that adjustment costs, particularly labor adjustment costs, play a significant and central role in driving empirically observed cost behavior patterns. Recent studies attempt to empirically test this hypothesis, albeit with distinct limitations. Using a new proxy for labor adjustment costs in a different population of firms, our study takes a fresh look at this hypothesis. We test the robustness of results documented in prior studies to help substantiate the credibility, reliability, and stability of prior findings.

Our proxy for labor adjustment costs captures the reliance on skilled labor across industries in a population of US public firms. Prior studies argue that skilled labor is associated with higher adjustment costs than unskilled labor due to greater hiring and firing costs associated with skilled labor. Based on the theoretical underpinnings of asymmetric cost behavior, we expect that a higher reliance on skilled labor will be associated with greater cost asymmetry. Our empirical results support this proposition. In additional subsample tests, we also find that the effect of labor adjustment costs on cost asymmetry is more pronounced when unemployment rates are low, for firms located in high Wrongful Discharge Laws (WDL) states, and for firms situated in low-hiring credit states. Together, these results provide compelling evidence that validates the consequential role of labor adjustment costs in determining asymmetric cost behavior.

Introduction

The issue of asymmetric cost behavior (also referred to as “cost stickiness”) has attracted considerable interest in the management accounting literature. It is observed that firms are less likely to decrease costs proportionately in response to a decrease in sales than they are to increase costs in response to proportional sales increases (i.e., costs are “sticky”) (Anderson et al., 2003). Asymmetric behavior of costs arises due to the use of managerial discretion to make resource adjustments in response to changes in demand. The early literature examining this phenomenon speculates that labor adjustment costs are one of the primary reasons for this observed asymmetry in costs (Anderson et al., 2003). We will refer to this proposition as the “labor adjustment cost hypothesis” of asymmetric cost behavior. The theoretical basis for this hypothesis is found in the labor economics literature, which represents labor as a quasi-fixed factor of production because firms incur costs when changing their demand for labor (“labor adjustment costs”). Labor adjustment costs are economically significant and involve costs of search (e.g., recruiter agency fees), selection and hiring (e.g., interviews), and training, as well as costs related to firing (e.g., lawsuits) and productivity losses (e.g., peer disruption) (e.g., Oi, 1962; Manning, 2006; Ghaly et al., 2017). Consequently, Anderson et al. (2003) argue that these labor adjustment costs may cause managers to hold onto labor resources when sales decline, thus generating cost asymmetry.

While the labor adjustment cost hypothesis underlies much of the sticky cost literature, conducting a formal test of this hypothesis has been challenging due to the inherent difficulty of identifying accurate proxies for labor adjustment costs (Banker et al., 2013). As a result, the existing literature has mostly relied on indirect proxies of employee adjustment costs as evidence of the adjustment costs hypothesis. For example, Anderson et al. (2003) model the variation in firms’ employee adjustment costs by using employee intensity (ratio of number of employees to sales revenue). However, the number of employees is a noisy measure for labor adjustment costs, since a firm can have a large number of unskilled workers associated with smaller adjustment costs levels.¹

To better explain cross-sectional variation in firms’ cost stickiness, recent studies utilize more direct proxies of labor adjustment costs. Banker et al. (2013) consider country-level employment protection legislation (EPL) as a proxy for adjustment costs associated with the legal impediments to firing workers in an international context. They examine the effect of country-level EPL on firm-level cost stickiness and find that cost stickiness is more pronounced in countries with higher levels of EPL. That is, employers faced with higher legal impediments due to downward adjustments of their firms’ workforce are more reluctant to fire workers in periods of sales decline. The interesting aspect of this study is that labor adjustment costs for the same type of labor input vary across observations. However, the Banker et al. (2013) study has some distinct limitations. First, the focus of this study is on institutional drivers of cost stickiness that are predicted to equally affect all employees within a particular country. As a result, while the study highlights the role played by institutional differences across countries on asymmetric cost behavior, it does not speak to the existence of differences in cost stickiness found across firms within the same country. Thus, their proxy of labor adjustment costs does not capture the intra-country variation in adjustment costs. Another limitation of Banker et al. (2013)’s setting is that the EPL measure is only available at the country-level for a small sample of 19 Organization of Economic Cooperation and Development (OECD) countries.² Additionally, given that the EPL proxy is a time-invariant country-specific measure, it is likely to be highly correlated with other country-level institutions and features, thus making it difficult to isolate and disentangle the effect of EPL in a cross-country setting (Isidro et al., 2016).

The closest attempt at capturing the variation in labor adjustment costs across firms within the same country is made in Dierynck et al. (2012). They construct a proxy based on the composition of the labor force (“blue collar” versus “white collar”), using a small sample of Belgian private firms with the objective of documenting the observed cost behavior when managers face benchmark-beating incentives. Dierynck et al. (2012) proxy for capturing the variation in labor adjustment costs is based on Belgian legal requirements that result in ex-ante higher costs related to adjusting the number of white-collar employees compared to the costs of adjusting the number of blue-collar employees.

The advantage of the Dierynck et al. (2012) approach is that Belgian law creates clear exogenous variation in the adjustment costs driven by the requirements of employment laws in the country. The disadvantage is that Belgian legal requirements with respect to blue-collar versus white-collar workers only apply to a setting where such legal requirements exist, making it difficult to generalize their results to other labor markets that do not share similar regulations. Another feature of the Dierynck et al. (2012) study is that they utilize a firm-level proxy of adjustment costs based on the composition of labor within a firm. A concern with using a firm-level proxy for managerial decision-making in the current period is that the composition of labor may be a partial outcome of prior managerial decisions made in response to past shocks, which may continue to have persistent and direct effects on costs in the current period (Banker et al., 2013).

Thus, prior studies that examine the labor adjustment cost hypothesis have limitations due to incomplete proxies of labor adjustment costs and/or due to the limited generalizability of the findings. The principal objective of our study is to extend the findings of these studies by examining a new measure of labor adjustment costs in a different population of firms in order to reinforce the robustness and credibility of the evidence in support of this consequential hypothesis (Clemens, 2017; Christensen and Miguel, 2018). We exploit employee data from the Occupational Employment Statistics (OES) program of the Bureau of Labor Statistics to more effectively capture the variation in average skill level of the workforce across industries. The U.S. Department of Labor O*NET program classifies occupations based on skill level.³ The skill level is based on an occupation’s requirements related to education, work experience, and the level of employee training. We use this data to construct a Labor Skills Index (LSI) that has been employed in the finance literature by Belo et al. (2017) and Ghaly et al. (2017) to capture firms’ reliance on skilled labor. The LSI Index measures the weighted average skill level of all occupations within an industry, and provides a proxy for an industry’s reliance on skilled labor (Ghaly et al., 2017).

A key advantage of the LSI Index is that it captures skill levels of different occupations which results in a finer and continuous measure of labor skills, instead of a simple dichotomous classification of blue-collar and white-collar workers, as used in Dierynck et al. (2012). Additionally, it enables us to investigate the differences in cost stickiness found across firms within the U.S. labor market by utilizing a large sample of U.S. firms in various industries. By focusing on the variation in employees’ skill levels within a single country, we are able to examine differences in labor-related adjustment costs that are independent of the effects of a particular country’s institutions. Finally, our industry-level measure captures reliance on skilled labor, which is mostly exogenous with respect to a particular firm’s resource adjustment decisions. Notwithstanding its distinct advantages, a limitation of the LSI measure, as used in our context, is that while the theoretical construct calls for capturing firm-level reliance in skilled labor, our LSI measure can only be captured at the industry-level, since the occupational skills data derived from BLS is only available at the industry-level.⁴

We utilize a sample that consists of 60,183 firm-year observations of U.S. listed firms from the period 1999 to 2016 to examine the labor adjustment cost hypothesis using the LSI Index as a proxy for labor-related adjustment costs. Our results indicate that firms that rely on high-skilled labor experience greater cost asymmetry. The impact of LSI on cost stickiness is economically significant: A one standard deviation increase in the Index is associated with an increase in cost stickiness of 4.2 basis points.⁵

To substantiate our conclusion that the relation between LSI and cost asymmetry is being driven by labor adjustment costs, we exploit characteristics of the economic and regulatory environment that cause differences in labor adjustment costs in three additional tests that incorporate exogenous variation in labor adjustment costs. First, we split our sample based on periods of high and low unemployment. In periods of high unemployment, it is easier for firms to find both high-skilled and low-skilled workers. Labor adjustment costs are expected to be lower for all workers during such periods. However, in periods of low unemployment, the tightness in labor markets leads to an increase in search and hiring costs, which rise to a greater extent for skilled labor than for unskilled labor (Muehlemann and Leiser, 2018). This phenomenon is borne out in anecdotal evidence as well. For instance, during the recent tightness in the labor market in the United States, the Federal Reserve Bank notes in its periodic survey, known as the Beige Book (2018), that firms have been experiencing greater difficulty in hiring workers, particularly skilled ones. As a consequence, we expect that the difference in adjustment costs between high-skilled and low-skilled workers will be greater in periods of low unemployment. Firms relying on skilled labor will be more reluctant to fire workers in sales downturns, leading to greater cost asymmetry for these firms. We find results that are consistent with this reasoning. We document that the link between firms’ reliance on skilled labor and cost asymmetry is more pronounced when unemployment rates are low.

Second, following Serfling (2016) and Ghaly et al. (2017), we exploit cross-state variation in state-level wrongful discharge laws (WDLs) to investigate the effect of exogenous variation of firing costs on the relationship between LSI and cost stickiness. Since firms located in states with stronger employment protection laws experience higher costs associated with dismissing employees, these firms are likely to have greater labor adjustment costs, which leads to more cost stickiness. WDLs generally apply to full-time and non-unionized employees, who are more skilled overall (Ghaly et al., 2017). Skilled workers have a higher probability of litigating against their employers, which increases ex-ante litigation risk and legal proceeding costs for firms that rely on these skilled workers (Ghaly et al., 2017). As a result, when a firm faces stringent worker protection laws, the difference in adjustment costs between high-skilled and low-skilled labor is likely to be higher. Consistent with this assumption, we predict and find that the effect of reliance on skilled labor on cost stickiness is stronger in the presence of WDL.

Third, we exploit exogenous variation in firms’ hiring costs by considering whether firms’ headquarters are located in states that support hiring credits. Hiring-credit programs facilitate the search process and reduce hiring costs with the intention of creating quality jobs with high earnings potential, thus ex-ante reducing labor adjustment costs for more highly skilled workers (Ghaly et al., 2017). As a consequence, the difference in adjustment costs between skilled and unskilled workers will be lower for firms located in the high hiring-credit states as compared to those situated in low hiring-credit states. Hence, we predict that cost stickiness for firms relying on high skilled workers will be more pronounced in low hiring-credit states. Our results are consistent with this prediction.

Our results are robust to a battery of sensitivity tests. We control for firm and industry fixed effects to address endogeneity concerns that may cause potential firm and industry heterogeneity bias. In addition, we consider the possibility that mergers and acquisitions (M&A) may be driving some of our results. To rule out this possibility, we conduct our tests excluding firm years with M&A activities and obtain similar results.⁶ We follow Balakrishnan et al. (2014) by using two alternative specifications of the Anderson et al. (2003) model (hereafter, referred to as the ABJ model) to verify that the presence of fixed costs does not bias our results. We also consider an alternative firm-level measure of cost stickiness developed in Weiss (2010) and obtain consistent results. To rule out the possibility that intangible assets may be driving the results, we control for R&D expenditures, and our results continue to hold. Finally, our main empirical tests are based on demonstrating the impact of labor adjustment costs on stickiness in firms’ operating costs. However, since the labor adjustment cost hypothesis is more directly related to firms’ employee-related costs, we conduct a sensitivity test by replacing operating costs with employee costs for a smaller subsample of firms that disclose employee costs. The results show that reliance on high-skilled labor induces greater cost asymmetry in terms of employee costs as well.

Our study thus provides compelling corroborating evidence of the labor adjustment costs hypothesis that underlies much of the cost stickiness literature. Utilizing a large sample of U.S. firms from a variety of industries, we extend the results in prior literature that emphasize the role of reliance on high skilled labor in firms’ resource allocation decisions (e.g., Dierynck et al., 2012). Since the labor adjustment hypothesis serves as a foundation for the theory of asymmetric cost behavior, we believe our study makes an important contribution to our understanding of cost behavior by reinforcing the robustness and credibility of the evidence in support of this consequential hypothesis.

Our paper is structured as follows. Section 2 presents the development of our research design. Section 3 provides our sample selection, and section 4 reports main empirical results. Section 5 shows additional analyses. Section 6 concludes.