WAL-MART AND MAINE: THE EFFECT ON EMPLOYMENT AND WAGES

by Brian A. Ketchum, Research Associate in the Antitrust Division, National Economic Research Associates,* and James W. Hughes, Associate Professor of Economics, Bates College

Introduction

Few events in rural America generate as much controversy as the opening of a Wal-Mart store. Local politicians often try to attract Wal-Mart with promises of tax abatements and road improvements, lured by the promise of jobs and tax revenues. Local businesses fear that they will be unable to compete effectively with Wal-Mart. Wal-Mart's tendency to locate on the outskirts of town, combined with its lower prices and drawing power raise the specter of downtown areas devoid of retail establishments. Workers fear that employment gains at Wal-Mart will be offset by larger job losses in other retail establishments. Furthermore, the jobs created at Wal-Mart are alleged to involve more part-time work, with lower wages and fewer benefits. Media stories relating how the establishment of a Wal-Mart destroyed the local retailing base and gutted downtown shopping areas in some southwestern towns feed these fears. While anecdotal stories abound, and are widely accepted, statistical evidence of the alleged destructive effects of Wal-Mart is surprisingly scant.

Wal-Mart entered the Maine market in 1992 with the opening of its Scarborough store. As of this writing, the Arkansas retailer has opened 19 stores in Maine, with still more to come. When Wal-Mart first opened in Maine, the Maine Business Research Report (MBRR) published an article purporting to show how Wal-Mart might affect the local retail environment.⁽¹⁾ The MBRR article was based on a prior study of rural retail markets in Iowa conducted by Professor Kenneth Stone of Iowa State University. The Stone study compared relative retail sales between Wal-Mart and non-Wal-Mart towns in Iowa one, three, and five years after the establishment of Wal-Mart. After five years, Stone found that retail sales in Wal-Mart towns increased by 6.5 percent, while similarly sized non-Wal-Mart towns experienced a decline in retail sales of 9.7 percent, both relative to state-wide averages. Stone attributes these trends to the presence or absence of a Wal-Mart store. Stone further claimed that a Wal-Mart store would not only capture all of the retail sales growth in a typical Iowa town, but would also take $10 million in sales from existing businesses.

While widely cited, we believe that such analyses are incomplete. Although the retail sales trends appear to be correlated with the establishment of a Wal-Mart, this methodology cannot rule out the possibility that other economic events, unrelated to the presence or absence of Wal-Mart, may be the underlying cause of the observed trends in retail sales. Suppose, for example, that Wal-Mart chooses to locate stores only in communities with expanding, rather than shrinking, economies. In comparing Wal-Mart and non-Wal-Mart communities, researchers would still observe growing sales in Wal-Mart communities and shrinking sales in non-Wal-Mart communities. However, it would be wrong to attribute these preexisting trends to the presence or absence of Wal-Mart.

A researcher could distinguish these two competing hypotheses by examining trends in retail sales for the years prior to the establishment of Wal-Mart. If similar trends of growth and decline were observed in the two sets of towns in the years both before and after the establishment of a Wal-Mart, this evidence would support the hypothesis that economic events other than Wal-Mart were responsible for the observed trends. In this case, the trends in retail sales observed following the establishment of Wal-Mart would be nothing more than the continuation of time trends in retail activity unrelated to Wal-Mart. Conversely, suppose that the trends in retail sales in the two sets of towns were similar prior to the arrival of Wal-Mart, only to diverge sharply thereafter. Such evidence would support the hypothesis that Wal-Mart stores were indeed responsible for the changing economic fortunes.

Another possibility is that the observed retail trends may reflect the effect of economic shocks that vary by town. For example, overall economic activity may vary between towns with the establishment of new plants and the closing of nonviable ones. Such shocks would affect the entire local economy, including retailing. If we examined only the retail sector, we would risk incorrectly attributing to Wal-Mart trends which actually result from changes in general economic activity. If trends in other sectors, such as manufacturing, services, agriculture, or construction, were similar to those in retailing, it would be difficult to attribute changes in these other sectors to the presence or absence of a Wal-Mart. If the trends were unique to the retail sector, it would strengthen the argument that the trends resulted from Wal-Mart strengthened.

The purpose of this paper is to isolate the effects of Wal-Mart on relative levels of employment and wages in Maine communities. As noted above, opponents of Wal-Mart claim that

• net retail employment gains are minimal in Wal-Mart communities, due to offsetting job losses in other retail establishments; and
• retail wages will be depressed in Wal-Mart communities due to the allegedly lower wages paid by the retailer.

Data and Methodology

The data set for this study consists of mean per capita employment (employment per 1,000 population) and mean wages for three major industry groups (retail, services, and manufacturing) in each of Maine's sixteen counties. The data set is constructed using data from the Maine Department of Labor for 1990 and 1994. The test group is comprised of those twelve counties having at least one Wal-Mart in operation by the end of 1994. The control group is the remaining four counties having no Wal-Marts in operation by the end of 1994.

Our methodology isolates the effects, if any, of having a Wal-Mart in the community by subtracting out the effects of time trends and county- or industry-specific events.⁽²⁾ First, we calculate mean retail employment per 1,000 population in Wal-Mart counties and non-Wal-Mart counties for 1990 and 1994. Changes in retail employment in Wal-Mart counties over this period are the result of either a) general changes in the retail sector; b) the establishment of a Wal-Mart between 1990 and 1994; or c) a combination of the two.

Changes in retail employment in non-Wal-Mart counties could only be the result of general changes in the retail sector. Thus taking the difference in the average employment change between Wal-Mart and non-Wal-Mart counties will remove the effects of general changes in the retail sector. This difference in retail employment across time and across county groups provides one measure of the change in retail employment due to the establishment of a Wal-Mart.

Observed changes in retail employment may also be the result of economic events outside of the retail sector, for example, changes in the employment base within a county. To account for such possibilities, we repeat the above exercise for our two control industries, manufacturing and services. Taking the difference between the change in employment in retailing and the change in employment in the control sectors will further isolate the employment change due solely to Wal-Mart by removing the effects of industry-specific events.

Finally, we evaluate the effect of Wal-Mart on wages by repeating the above exercise for average wages.

Results

The estimation results for employment and wages are found in Tables 1 and 2, respectively. In the retail trade sector, counties having Wal-Mart stores experienced an average increase of approximately one employee per 1,000 population (1.036) between 1990 and 1994. Non-Wal-Mart counties suffered a loss of slightly more than one-half of an employee per 1,000 population (-0.639) over the same time period. Neither change is statistically significant at the ten-percent level. The net retail employment change estimate indicates a relative increase of 1.676 (1.036 - (-0.639)) retail employees in Wal-Mart counties between 1990 and 1994. This change is also statistically insignificant, suggesting that the relative increase over the period cannot be accounted for by Wal-Mart's operations.

This result is not conclusive, however. If the change in employment in the retail trade is substantially different from the change in other sectors of the economy, the change in the relative performance of retailing and other industries might indeed be attributable to Wal-Mart. The net employment change across industry estimates compare the relative change in retail employment with the relative change in service employment (Table 1, Panel B) and manufacturing employment (Table 1, Panel C).

In the services industry, Wal-Mart counties gained approximately 10 retail employees per 1,000 population, compared to approximately seven employees per 1,000 population in non-Wal-Mart counties. The changes over time and across county groups were both statistically insignificant at the ten-percent level. The net employment change estimate between retail trade and service industries suggests a relative increase of less than one retail employee compared to the service sector. This estimate was statistically insignificant, meaning that we cannot reject the hypothesis that the differences in employment between the two industry sectors are the result of random variation, and not the result of Wal-Mart's operations.

Comparing retail employment with manufacturing employment, we find that Wal-Mart counties lost over six manufacturing employees per 1,000 population between 1990 and 1994, while non-Wal-Mart counties lost 25 employees. Both changes were statistically insignificant. The net employment change estimate showed a relative increase of over 18 manufacturing employees in Wal-Mart counties, but this estimate was also statistically insignificant. The net employment change across industries estimate indicates an increase in retail employment relative to manufacturing employment of approximately 17 employees, but this estimate was statistically insignificant. The lack of statistically significant results in comparing retail trade to both services and manufacturing suggests that the presence of Wal-Mart in a community cannot account for the relative gains in employment in Wal-Mart counties.

Turning to the results on wages in Table 2, we found that average weekly wages increased $28.02 in Wal-Mart counties, and by $19.78 in non-Wal-Mart counties over the period 1990 through 1994. The gain in Wal-Mart counties is statistically significant at the one-percent level. However, the relative gain in Wal-Mart counties relative to non-Wal-Mart counties of $8.24 was not statistically significant. The lack of significance for the relative gain implies that the gain in retail wages in Wal-Mart counties cannot be attributed solely to the presence of Wal-Mart.

As with employment, we compare wage changes in retailing with wage changes in the services and manufacturing sectors. Average weekly wages in the service sector increased in both Wal-Mart and non-Wal-Mart counties between 1990 and 1994. In Wal-Mart counties, wages rose by $37.97, while in non-Wal-Mart counties, the increase was $39.52. The increase in Wal-Mart counties is significant at the one-percent level, while the increase in non-Wal-Mart counties fails to achieve significance at the ten-percent level. The relative wage change estimate indicates a relative increase of $1.54 in non-Wal-Mart counties, although the change is statistically insignificant. The net wage change estimate between the retail and service sectors indicates a relative $9.78 increase in retail wage, but this relative gain is again statistically insignificant.

In the manufacturing sector, Wal-Mart counties experienced a weekly wage increase of $73.68, statistically significant at the one-percent level, while non-Wal-Mart counties showed a gain of $44.69. The increase in non-Wal-Mart counties is insignificant at the ten-percent level. Both the relative wage change estimate within manufacturing, and the estimate of relative wage changes between retailing and manufacturing are statistically insignificant.

These results suggest that the economies of Wal-Mart counties were growing faster overall than the economies of non-Wal-Mart counties. The relative growth in average weekly wages in Wal-Mart counties cannot be attributed to the presence or absence of a Wal-Mart.

Conclusions

By controlling for industry-specific shocks we have shown that the 12 counties where Wal-Mart is currently located have not experienced declines in either retail wages or employment between 1990 and 1994. Retailing, manufacturing, and services all experienced statistically significant wage gains in counties having a Wal-Mart store. Nevertheless, the differences in growth relative to non-Wal-Mart counties were not statistically significant, such that the relative wage growth in Wal-Mart counties could not be attributed to Wal-Mart's presence. This finding is contrary to the claims of Wal-Mart opponents.

Between 1990 and 1994, employment growth in the retail sector was quite small. Wal-Mart opponents claim that this is the result we should expect to see, as job growth at Wal-Mart is offset by employment shrinkage in other local businesses. However, after correcting for other time-, industry-, and county-specific economic shocks, we found no evidence that Wal-Mart was responsible for the lack of growth in retail employment. We conclude that these other economic shocks are responsible for the observed trends in retail employment.

The authors wish to thank Les Bray, a private consultant, for providing the data used in this report. Institutional affiliation is provided for information only. The views expressed in this article are those of the authors, and not necessarily those of their institutions.

2. The methodology employed here is based on the "difference-in-difference-in-difference" method used in J. Gruber, "The Incidence of Mandated Maternity Benefits," American Economic Review 84:3 (June 1994): 622-41.