Supermarket human resource practices and competition from mass merchandisers.

In recent years, much public concern has been raised about whether industrial restructuring has resulted in the creation of more "bad jobs" in the United States. Critics argue that employers have changed long-standing practices regarding the terms of employment and the way wages are set. The fear is that there are fewer jobs that offer a traditional long-term employment relationship and, at the same time, there are more low-skilled jobs with high rates of turnover and little opportunity for training and wage advancement. Empirical evidence suggests that for workers with less education and few skills, the opportunities for advancement through job ladders are dwindling (Bernhardt et al. 2001). The objective of this article is to use a new detailed data set to estimate the impact of restructuring on human resource (HR) practices in the retail food industry.

The retail food industry is, in many ways, an ideal industry for such a study. Although the retail sector of the economy has always had a relatively flat job hierarchy, supermarket jobs were once among the most highly paid and highly coveted retail jobs. However, the typical supermarket job is no longer a full-time, relatively well-paid position (often unionized), but rather a part-time job with irregular hours, low pay, and limited options for training or promotion (Hughes 1999). This shift has occurred at the same time that the industry has undergone dramatic product market restructuring as Wal-Mart and other mass merchandisers have entered the industry. Wal-Mart is now the largest food retailer in the United States with its share of the grocery market estimated to be close to 20%, having expanded from only ten supercenters in 1993 to over 1,866 supercenters by 2005.

In this article we analyze the relationship between growing competition from mass merchandisers like Wal-Mart and changes in HR practices within the industry. While case study evidence suggests that the proliferation of big box stores has had a substantial impact on the labor market, most empirical studies to date have focused on the changes in county-level employment and wages that occur after Wal-Mart entry (Basker 2005; Neumark, Zhang, and Ciccarella 2005). (1) There has been no large-scale data set available on both firms and workers that could be used to describe HR practices at the firm level. The data set used here allows analysis of changes in supermarket hiring, promotion, pay, and turnover policies at the establishment level in response to entry of mass merchandisers in the local market. We particularly focus on the role of firm exit, since such policies have been linked to firm performance and survival (Haltiwanger, Lane, and Spletzer 2006).

Background and Motivation

Measurement of shifts in HR practices of food retailers in response to changing product market competition is a challenge. Some guidance is provided by Lazear and Oyer (2004) who use measures of promotion, hiring, and wage setting to capture key aspects of HR practices--which they (and we) refer to as internal labor markets (ILMs). ILMs are generally characterized by long-term employment relationships, with most hiring done from within the firm for positions other than low-level "port-of entry" jobs. In firms with ILMs, wages are related to job characteristics and are relatively unresponsive to changes in the external labor market. Evidence supporting (though not proving) the existence of ILMs includes the persistence of firm wage differentials over time, upward mobility and returns to seniority within firms, and limited external hiring other than at ports of entry. As described by Groshen and Levine (1998), numerous theories have been developed to explain why firms create ILMs. These models focus on the importance of firm-specific human capital, incentives, and risk sharing as possible motivations. Fairris (2004) finds evidence that firms choose ILM practices to influence workforce quality, effort, and quit rate. The critical element is that firms adjust HR practices and may change their ILM status in response to competitive conditions.

While food stores are generally not known for innovative or high-performance HR practices (Ben-Ner, Kong, and Bosley 1999), there is some case study evidence of variation in HR practices across firms in this industry. In 2006, for example, Fortune magazine's list of the top 100 companies to work for included several supermarket chains, with Wegmans Food Markets and Whole Foods Markets ranked in the top twenty. For some stores facing increased competition, customer service is seen as an important edge, and long-term employment relationships may improve productivity and thus encourage the development of ILMs. Some food retailers have expanded the range of specialized services they offer, including more labor- and training-intensive services such as bakeries, delis, prepared food items, and other services (Warner 2005). Thus, the limited evidence available suggests that there is heterogeneity in the wage and ILM structure in the retail food industry, and that individual firms may respond only sluggishly to changes in the external market.

Data and Measurement

The data used in this article are drawn from the U.S. Census Bureau's Longitudinal Employer Household Dynamics (LEHD) database that matches workers with employers. This data base includes quarterly records of the earnings of almost all individuals from the unemployment insurance systems of most U.S. states starting in the 1990s. This study uses a subset of seven states (California, Idaho, Illinois, Maryland, North Carolina, Oregon, and Washington) that have sufficient years of longitudinal data. These data have been extensively described elsewhere (Abowd et al. 2006).

For this study the LEHD data were matched with additional information on both firms and workers. Worker characteristics include date of birth, place of birth, race, and sex. Data from the 1997 and 2002 Economic Census include establishment characteristics such as payroll, sales, and product line. We also include controls for local economic conditions from Bureau of Economic Analysis data on per capita income, county population, and employment density.

Establishment-specific measures of concentration and competition are constructed for the retail food industry using the geocoded LEHD data. These measures of concentration and competition are created based on a 5-mile radius around the longitude and latitude of each establishment's location (see Davis et al. (2005) for more details). We calculate both sales-based four firm concentration ratios (CR-4) and Herfindahl indices on an establishment-specific basis. The CR-4 in this case represents the share of sales in a given region accounted for by the top four firms in that area (excluding the sales of the establishment itself). The Herfindahl index represents the sum of the squares of sales shares in each region. Measures of threat from outside the industry are derived in a similar fashion. The number, employment, sales, and payroll of mass merchandisers are calculated within each grocery-store specific 5-mile circle. A key innovation of this article is that the measures are establishment-specific and are not limited by arbitrary administratively defined geographic boundaries such as counties.

ILMs and HR Practices in Supermarkets

Following Lazear and Oyer (2004), we use measures of promotion, hiring, and wage setting to capture key aspects of HR practices of supermarkets. For promotion practices, we measure the proportion of workers hired into the second quintile that move to a higher quintile in five years and the wage growth of workers starting in the second quintile over the five-year time span. (2) Hiring patterns are captured by the churning rates (3) of all full-quarter workers in the establishment as well as by the proportion of accessions (new hires plus recalls) in the fourth and fifth earnings quintiles within the firm. Wage policies are measured by the mean and standard deviation of log real earnings for full quarter workers in the firm.

Given the high correlation of these measures across establishments (with the exception of worker wage growth) we employ cluster analysis to classify the supermarkets into two groups, which for convenience we call ILM and non-ILM. The clustering strategy uses nonhierarchical clustering based on the median value of the measures in each group. The measures include worker churning, mean earnings, the standard deviation of earnings, and the ratio of flow to full quarter workers. The clustering is done on pooled 1997 and 2002 data.

Table 1 (columns one and two) lists 2002 summary statistics for firms identified as ILM or non-ILM based on the cluster analysis. By construction, the ILM and non-ILM firms differ greatly across the variables included in the cluster analysis. The pattern in the two clusters is consistent with ILM theory. The ILM firms have significantly lower churning rates, higher average earnings, a higher standard deviation of earnings, and a higher share of full quarter workers relative to flow employment. The measures of promotion, hiring, and wage growth clearly illustrate the diversity of HR practices across supermarkets. In addition, the stores identified as ILM or non-ILM differ on other measures as well (not shown). Firms that are classified as ILM promote a larger portion of their workers into higher earnings quintiles, have stronger average within-firm earnings growth, and tend to promote from within rather than hire outside the firm to fill higher-earning positions.

The LEHD data enable creation of measures of the diversity of different aspects of HR practices at the establishment level. They do not, however, directly capture other measures commonly used to describe HR practices, notably training opportunities and incentive pay structures. In order to validate the LEHD-based summary index, we incorporated information from an external survey, the Supermarket Panel survey conducted by The Food Industry Center at the University of Minnesota. The Supermarket Panel Survey is conducted at the store level and typically completed by the store manager (King, Jacobson, and Seltzer 2002). We use the 2002 Supermarket Panel Survey to construct a HR practices index based on five store-level indicators: hours of training for new cashiers; hours of training for store managers, grocery department managers, and scanning coordinators; the proportion of full-time employees hired at the store; and two measures of incentive-based compensation and noncash compensation at the store. (4) These practices (more training, more full-time employment, and more incentive-based compensation) suggest the establishment is trying to develop firm-specific human capital and reduce turnover.

To create the index, each store was ranked as above or below the mean (calculated from the survey data) for each of the five measures. If the store was above the mean on at least three measures, it was coded as a high HR store. About one-third of the stores in the Supermarket Panel survey were ranked "high" according to this index. Using this index as a guide, we then categorized major supermarket chains into three groups: those at the high end of the HR scale, those at the low end, and those with high variability from one store location to another. While there was some subjectivity in this categorization, use of case study and industry knowledge helped to inform the process. (5)

Table 1 (columns three and four) shows the LEHD measures for firms categorized based on the Supermarket Panel as having high and low HR practices (those in the mid-range are excluded). These results show the expected consistency with the LEHD-based classification into ILM and non-ILM categories. The high and low HR stores show similar patterns as seen in columns one and two, though the differences between the two groups are generally smaller. In particular, the firms with a low score on the HR practices index as measured in the Supermarket Panel Survey have significantly higher churning rates, lower mean earnings, and less wage dispersion than do firms identified as having a higher HR index score.

Comparing the high and low rankings based on the Supermarket Panel to the grouping of stores based on the LEHD measures, we find considerable consistency despite using different data sources and measures of HR practices. In 2002, 77% of firms identified as having high HR practices based on the Supermarket survey were identified as ILM firms using the cluster analysis. Close to 60% of firms designated as having low HR practices were non-ILM firms based on the cluster analysis. These results increase our confidence that the LEHD measures of ILMs are capturing important differences in store HR policies and practices.

Supermarket Response to Competition from Mass Merchandisers

Faced with growing competition from nontraditional food sellers, supermarkets have typically responded with changes in marketing and pricing, yet they may also alter employment and compensation strategies in an effort to adapt. In this section we analyze the impact of competition from mass merchandisers on two outcomes: the likelihood of firm exit and the probability of changing HR practices (from ILM to non-ILM or vice versa).

Table 2 provides a descriptive look at the patterns of HR changes and firm exits over time for establishments that faced a significant competitive threat (defined as having two or more mass merchandisers within 5 miles) and those who did not. The findings suggest that individual establishments in the retail food industry are not rapidly adjusting HR practices despite changes in the external environment. Even in high-threat areas, of those firms that did not exit by 2002, most establishments do not change their ILM status: only about 11% switched ILM/non-ILM categorization. The primary response occurred on the entry/exit margin. In the high-threat areas 341 firms exited, of which 40% were ILM, and 269 entered, of which 43% were ILM. In the low-threat areas, only 24% of exiting establishments were ILM. Meanwhile, of the 180 firms that entered in low threat areas, 31% were ILM.

While both ILM and non-ILM firms are more likely to exit the industry than change HR practices, non-ILM firms pursue an exit rather than change strategy at a much higher rate than ILM firms, particularly in high-threat areas. Non-ILM firms are more than four times more likely to exit than change while ILM firms are just over twice as likely to exit as change in high-threat areas. In low-threat areas, by contrast, non-ILM firms are still more than four times more likely to exit than change while ILM firms are 1.5 times more likely to exit. Given that an establishment survives, the propensity to maintain the same HR policy over time is about the same for both ILM firms and non-ILM firms (near 80%).

Empirical Analysis and Findings

HR practices observed in the retail food industry are heterogeneous and persistent over time. In this section, we test how firm exit and firm ILM status are affected by product market competition while controlling for establishment characteristics. We first estimate a probit model for the probability that an establishment observed in 1997 exits by 2002. The key independent variables are the number of mass merchandisers within a 5-mile radius (measured in 1997)--which measures the competitive threat--and a dummy for whether the firm uses ILM HR practices. Thus we can compare the impacts of mass merchandisers on grocery store exit and the differential effect of mass merchandisers on ILM versus non-ILM grocery store establishments. Variables that control for other aspects of the competitiveness and size of the local product market include the 1997 four-firm concentration ratio of grocery stores, 1997 county per capita income, and 1997 county population. Establishment-level control variables include firm size and workforce composition (age, sex, education, and citizenship).

Table 3 presents the marginal effects from the probit regressions for firm exit. Focusing first on model one, stores with greater numbers of mass merchandisers within a 5-mile radius are more likely to exit while stores with ILM practices are less likely to exit. The large negative coefficient on ILM status is consistent with expectations, as ILM status is likely to be correlated with the age, multi-unit status, and overall economic performance of a grocery establishment. While the effect of mass merchandisers is relatively small, it is consistent with our expectations that competition from mass merchandisers adversely affects traditional grocery store survival.

The second model adds an interaction between number of mass merchandisers nearby and ILM status. The coefficient on this variable is positive and significant suggesting that grocery establishments with ILMs, relative to non-ILM grocery establishments, are more adversely affected by competition from mass merchandisers. The coefficient on mass merchandisers becomes insignificant while the coefficient on ILM status becomes larger. The size and significance of these first three coefficients remain constant as additional controls for local product market conditions and firm characteristics are added in the last two models, despite the relatively small sample size.

The overall results from the exit models are particularly informative given industry discussions about how to deal with emerging competition from mass merchandisers. Information from supermarket managers who faced early entry from mass merchandisers suggested a conservative strategy in adjusting labor practices. The Progressive Grocer noted that grocers who are "doing battle with supercenters tend to rely on tried-and-true weapons such as service and perishables" (Garry 1993). Competitive tactics such as expanding service departments, focusing on strong customer service, and putting more emphasis on the quality of perishables were mentioned by store managers as the best methods for competing against mass merchandisers and supercenters. These strategies rely on dedicated, long-term employees who are familiar with the longtime customers and have a commitment to service and are most effectively implemented with an ILM workforce.

A second issue is to identify factors that influence changes in the internal labor structure of food retailers, restricting attention to establishments who remain in business from 1997 through 2002. The results shown here are restricted to food retailers identified as having an ILM in 1997 and estimate the probability of switching from ILM to non-ILM status by 2002 using a probit model. Across all specifications presented in table 4, the number of mass merchandisers located in the same market area as a food retailer had no statistically discernible impact on switching behavior (similarly, the number of mass merchandisers had no statistically significant correlation with switching from non-ILM to ILM status in a separate regression not shown). Food retailers with an ILM structure were significantly less likely to switch to a non-ILM structure when operated in a region with a high concentration ratio than if it operated in a region with a low concentration ratio. This effect is apparent even when grocery store size is controlled for in the model, although the point estimate is reduced and statistical significance is at the 10% significance level. This result suggests that increases in the concentration ratio of grocery sales in a given region provide an insulating factor for food retailers to retain ILM-type HR practices. Overall, however, the competitive presence of mass merchandisers does not appear to influence food retailers into switching either to or away from an ILM structure.

Conclusion

Technology, changing consumer preferences, and competition from nontraditional food retailers like Wal-Mart have led to major changes in supermarket operations, pricing, and supply chain strategies over the past decade. Several recent empirical studies suggest that Wal-Mart's entry reduces employment and payroll in a county, yet the way an industry's labor market adjusts in response to such competitive shocks has not been clear. The evidence presented here suggests that there is considerable heterogeneity in HR practices across retail food establishments, and these practices are quite persistent even in the face of new external competition. Individual establishments do not appear to change HR strategies rapidly. Establishments with ILMs as a whole are less likely to exit, yet the probability of ILM establishments exiting increases when they are faced with increased competition from mass merchandisers. On the other hand, the results suggest that exits of non-ILM firms are unaffected by this increased competition. Our analysis finds that HR practices are persistent among food retailers and thus it is the entry and exit of firms rather than changes in strategies of existing firms which leads to aggregate changes in employment and payroll in local labor markets.

References

Abowd, J.M., B.E. Stephens, L. Vilhuber, F. Andersson, K.L. McKinney, M. Roemer, and S. Woodcock. 2005. "The LEHD Infrastructure Files and the Creation of the Quarterly Workforce Indicators." U.S. Bureau of the Census. Available at http://lehd.dsd.census.gov/led/library/techpapers/tp-2006-01.pdf

Ai, C., and E. Norton. 2003. "Interaction Terms in Logit and Probit Models." Economic Letters 80:123-9.

Basker, E. 2005. "Job Creation or Destruction? Labor-Market Effects of Wal-Mart Expansion." Review of Economics and Statistics 87(1):174-83.

Ben-Net, A., F. Kong, and S. Bosley. 1999. "Workplace Organization and Human Resource Practices: The Retail Food Industry." Working paper, The Food Industry Center, University of Minnesota.

Bernhardt, A., M. Morris, M. Handcock, and M. Scott. 2001. Divergent Paths: Economic Mobility in the New American Labor Market. New York: Russell Sage.

Davis, E.E., M. Freedman, J. Lane, B. McCall, N. Nestoriak, and T. Park. 2005. "Product Market Competition and Human Resource Practices: An Analysis of the Retail Food Sector." Working paper, The Food Industry Center, University of Minnesota.

Fairris, D. 2004. "Internal Labor Markets and Worker Quits." Industrial Relations 43:573-94. Garry, M. 1993. "Showdown! Standing up to Supercenters." Progressive Grocer 72(February):4450.

Groshen, E., and D. Levine. 1998. "The Rise and Decline (?) of U.S. Internal Labor Markets." Federal Reserve Bank of New York, Research Paper Number 9819.

Haltiwanger, J., J. Lane, and J. Spletzer. 2006. "Wage, Productivity and the Dynamic Interaction of Businesses and Workers." Labour Economics, in press.

Hausman, J., and E. Leibtag. 2005. "Consumer Benefits from Increased Competition in Shopping Outlets: Measuring the Effect of Wal-Mart," NBER Working Paper 11809.

Hughes, K. 1999. "Supermarket Employment: Good Jobs at Good Wages?" IEE Working paper W-11. Insititute on Education and the Economy, Columbia University.

King, R.P., E.M. Jacobson, and J.M. Setzer. 2002. The 2002 Supermarket Panel: Annual Report. The Food Industry Center, University of Minnesota.

Lazear, E., and P. Oyer. 2004. "Internal and External Labor Markets: A Personnel Economics Approach." Labour Economics 11:527-54.

Neumark, D., J. Zhang, and S. Ciccarella. 2005. "The Effects of Wal-Mart on Local Labor Markets." NBER working paper W11782.

Warner, M. 2005. "An Identity Crisis for Supermarkets: Stores are Losing Shoppers to Specialty Chains and Discount Giants." New York Times. October 6, 2005.

Elizabeth E. Davis is associate professor, University of Minnesota; Matthew Freedman is a Ph.D. candidate in Economics, University of Maryland-College Park, Julia Lane is Senior Vice President and Director, Economics, Labor and Population Department, NORC-University of Chicago: Brian McCall is professor, University of Minnesota; Nicole Nestoriak is Economist, Bureau of Economic Analysis; and Timothy Park is professor, University of Georgia.

Earlier versions of this article have benefited from helpful comments from Charles Brown, Erica Groshen, James Hertel, Jean Kinsey, Anne Russell, and Scott Scheuler. This research uses confidential data from the Census Bureau's Longitudinal Employer-Household Dynamics Program (LEHD), which is partially supported by the National Science Foundation Grant SES-9978093 to Cornell University (Cornell Institute for Social and Economic Research), the National Institute on Aging, and the Alfred P. Sloan Foundation. Disclaimer: The analysis presented here has undergone a more limited review than an official Census Bureau publication. The views expressed are attributable only to the authors and do not represent the views of the U.S. Census Bureau, Bureau of Economic Analysis, program sponsors or the data providers.

This article was presented in a principal paper session at the AAEA annual meeting (Long Beach, CA, July 2006). The articles in these sessions are not subjected to the journal's standard refereeing process.

(1) Other related research has focused on impacts on consumers. Hausman and Leibtag (2005), for example, find that the introduction of Wal-Mart has led to lower prices for a broad range of consumer goods.

(2) Much of the supermarket workforce is part-time and transitional by choice, with a smaller percentage of career retail food workers appearing in the data. We address this issue by focusing on full-quarter workers, as well as examining the promotion and wage growth of workers who have earnings that are in the second and higher quintiles of the firm wage distribution, rather than the bottom quintile. The sample includes all establishments classified as grocery stores (SIC 54111) that have at least fifteen full-quarter workers and thirty flow workers. Flow employment accounts for all workers employed by the firm at any time during the quarter, while full quarter employment measures all workers who were employed on either a part-time or full-time basis by the employer for the full quarter. The earnings measures reflect quarterly earnings without any adjustment for the number of hours worked during the quarter. There is no information on hours or weeks worked or the duration of employment within the quarter in the database.

(3) Churning is defined as accessions (new hires and recalls) plus separations (layoffs, quits, and firings) minus net job flows divided by employment.

(4) There are nine indicators of incentive based and noncash compensation in the survey, including the issuance of annual bonuses, individual performance incentive pay, incentive pay based on product or category performance, an employee stock ownership plan, individual health insurance, family health insurance, disability insurance, a company funded pension plan, and a 401(k) plan. From the survey, we count the number of indicators that are typically part of the compensation of both full-time personnel and part-time personnel. This is consistent with the definition of an ILM in which benefits accrue to jobs and not individuals within the firm.

(5) We considered other store-level organizational factors, such as membership in a self-distributing chain and unionization, but in the end did not include these in the HR index. Close to 50% of self-distributing stores (those in which stores and distribution centers are under common ownership) are high HR establishments while only 38% of wholesaler supplied stores pursue this strategy. In contrast, unionization did not seem closely related to the HR measure. The proportion of high HR stores is very similar across union and nonunion stores in the survey. Table 1. Summary Statistics on Employment and Pay Practices of Supermarkets Employment or Pay Low HR High HR Measure, 2002 Non-ILM ILM Index Index Churning rate 0.20 0.11 * 0.14 0.10 Log mean earnings 8.27 8.79 * 8.50 8.70 * Log SD of earnings 7.86 8.53 * 8.28 8.47 * Flow-to-full-quarter

employment ratio 1.62 1.35 * 1.36 1.37 Sample size 1,781 1,242 82 188 Note: The first two columns cluster stores into non-ILM and ILM categories based on LEHD measures. The last two columns group stores into low and high HR practices based on external survey data. * Difference between ILM (high HR index) and non-ILM (low HR index) is statistically significant at the 5%, level. Table 2. Changes in ILM Status of Food Stores in Areas with High or Low Threat of Competition from Mass Merchandisers

Number of Did Not Change Competition Level Establishments ILM Status by and HR Status in 1997 2002% High-threat location

Non-ILM 462 44.6

ILM 537 62.9 Low-threat location

Non-ILM 511 47.4

ILM 354 66.4

Changed ILM Exited by Competition Level Status by 2002 and HR Status 2002% (%) High-threat location

Non-ILM 11.5 43.9

ILM 11.3 25.7 Low-threat location

Non-ILM 9.2 43.4

ILM 13.5 20.1 Table 3. Probability of Firm Exit: Probit Model Estimation Results Marginal Effects Model (1) Model (2) Number of mass merchandisers nearby 0.0132 *** 0.0043

(0.0048) (0.0062) ILM -0.2080 *** -0.2586 ***

(0.0214) (0.0306) ILM x number of mass merchandisers 0.0168 *

(0.0090) Four-firm concentration ratio Controls for firm size and workforce

composition and state dummy variables No No Observations 1,864 1,864 Marginal Effects Model (3) Model (4) Number of mass merchandisers nearby 0.0057 0.0100

(0.0062) (0.0076) ILM -0.2594 *** -0.2172 ***

(0.0307) (0.0365) ILM x number of mass merchandisers 0.0169 * 0.0166 *

(0.0090) (0.0093) Four-firm concentration ratio 0.0792 * 0.0465

(0.0449) (0.0479) Controls for firm size and workforce

composition and state dummy variables No Yes Observations 1,864 1,864 Note: Standard errors are in parentheses. Marginal effects calculated including the Ai-Norton (21103) correction for the magnitudes and standard errors of the interaction effects. Single (*), double (**), and triple (***) asterisks denote significance at 10%, 5%, and 1% levels, respectively. Table 4. Probability of Changing HR Practices: Probit Model of Switching from ILM to Non-ILM Marginal Effects Model (1) Model (2) Number of mass merchandisers nearby 0.0035 0.0009

(0.0067) (0.0066) Four-firm concentration ratio -0.1768 ***

(0.0636) Log per capita income Controls for firm size, workforce No No

composition, and state dummy variables Observations 668 668 Marginal Effects Model (3) Model(4) Number of mass merchandisers nearby 0.0076 -0.02

(0.0067) (0.0065) Four-firm concentration ratio -0.2126 *** -0.0867 *

(0.0630) (0.0513) Log per capita income -0.2730 *** -0.2259 ***

(0.0664) (0.0659) Controls for firm size, workforce

composition, and state dummy variables No Yes Observations 668 668 Note: Standard errors are in parentheses Models also include controls for change in ownership and log of county population. Single (*), double (**). and triple (***) asterisks denote significance at 10%, 5%, and I% levels, respectively.