Impact of the 1996 Summer Olympic Games on employment and wages in Georgia.

• Author: Hotchkiss, Julie L.

1. Introduction

   In September 1990, Atlanta won the bid to host the 1996 Summer Olympic Games. In spite of the approximately $2.5 billion price tag, the benefits derived from hosting the Olympic Games were expected to outweigh the costs. Positive media attention, construction of facilities and infrastructure, and employment increases were identified as the primary beneficial output of this massive endeavor (Humphreys and Plummer 1995; Newman 1999). While actual dollar inflows during the Olympics are relatively easy to identify, the "legacy" of the Olympics in terms of long-term benefits is more difficult to measure. In order to measure, for example, the employment legacy, it is important to isolate the increase in employment that would have taken place had the Olympics not come to Georgia. With that in mind, the purpose of this paper is to provide quantitative estimates of the impact of the 1996 Olympic Games on employment and wages in Georgia.

   Fundamentally, the demand for labor is a derived demand. Exogenous factors that affect the demand for labor include the price of other factor inputs, the demand for output, and the state of technology. Accordingly, one purpose of studying labor demand is to understand how exogenous changes in these variables affect employment and/or wage rates. The Olympic Games are expected to have had three exogenous effects on the labor market. First, there should have been a direct short-term effect on employment due to the direct spending by the Atlanta Committee for the Olympic Games (ACOG) on goods and services. Second, in conjunction with the Georgia Department of Technical and Adult Services, a Pastor Grant was obtained by ACOG to provide job training. This formal training, in addition to the experience obtained by the estimated 70,000 volunteers, should have impacted employment opportunities of workers. Third, investments in facilities and infrastructure, as well as migration resulting from positive publicity, are e xpected to have positively impacted employment and wages well beyond the Olympic event. If it can be shown that an exogenous shock to a labor market, such as that brought about by the Olympic Games, can improve the employment situation of workers, it may prompt urban policymakers to rely more on promoting development projects when tackling the issue of unemployment instead of relying on alternative strategies such as targeted wage subsidies. (1)

   The analysis presented here makes use of state-level unemployment insurance employment data (ES202 data) to measure the change in employment experienced by Olympic venue geographic areas and to compare that change with the employment change experienced by geographic areas in Georgia not affiliated with an Olympic venue and with geographic areas similar to venue areas but not in Georgia. Differences-in-differences (DD) statistical analyses will provide evidence that overall employment in venue and near-venue areas increased 17% more during and after the Olympic games than in nonvenue areas. We also show that this increase was not merely a metropolitan phenomenon; employment in the northern venue areas (the most heavily populated areas) increased 11% more during and after the Olympic games than did employment in other similar metropolitan areas in the south. In addition, a random-growth estimation procedure confirms that the employment difference measured post- versus pre-Olympics between venue-area and non-ven ue-area counties is not merely the result of systematic differences between the two types of counties.

   Not only is there evidence that the level of employment increased more in the venue and near-venue Georgia counties, but a modified DD analysis indicates that the rate of growth in employment was also positively impacted by the presence of the Olympics. We estimate a nearly 0.002 percentage-point-per-quarter increase in employment growth for venue-area counties relative to nonvenue-area counties post- versus pre-Olympics.

   Analysis of wages does not yield such clear-cut conclusions. While the DD analyses indicate that real per worker wages increased 7% more in venue area counties and that the rate of growth increased by nearly 0.001 percentage points per quarter, the random-growth estimator robustness check indicated that the amount of noise surrounding the wage series is too great to draw any definite conclusions.

2. Background and Data

   For the analyses, in this paper we identify counties in which Olympic venues were located as venue counties and counties adjacent to venue counties as near-venue counties. Together, these two groups of counties will be referred to as venue and near-venue (VNV) counties, and these counties will be the counties expected to be affected by the presence of the Olympic Games. The theory is that one would expect to observe employment and wage gains in areas geographically "close" to where Olympic events were held as opposed to those areas not close to Olympic events. Figure 1 depicts a map of Georgia with venue counties darkly shaded and near-venue counties lightly shaded. There are three main VNV county groups: North (including Atlanta and Athens), Savannah on the coast, and Columbus.

   Quarterly employment and wage data for each county in Georgia from 1985 through the third quarter of 2000 were obtained from administrative records made available by the Georgia Department of Labor. (2) Nominal (per-worker) wages were converted to real wages using the Consumer Price Index (CPI) (CPI = 100, based on average prices between 1982-1984).

   The industry mix in each county in each quarter was also calculated using the administrative data. These data reflect the percentages of employment distributed across industries at the two-digit Standard Industrial Classification (SIC) code in 1990. (3) Population levels for each county in 1990 were obtained from the Georgia Institute of Technology State Data and Research Center. (4) The industry mix and population levels in 1990 were included to control for county-level characteristics that might otherwise confound differences measured in county employment and wage levels across venue status.

   For an initial look at the potential employment and wage impact of the Olympics, Figures 2 and 3 present average employment indices and average per-worker real wage indices for VNV counties and for non-VNV counties. Each quarter plots the average employment (wage) in that quarter for each county category indexed by (i.e., divided by) employment (wage) in the first quarter of 1985. Figure 2 suggests, without controlling for any other characteristics such as population or industry mix, that employment grew at roughly the same rate across the two groups of counties prior to the early 1990s and that both groups of counties experienced similar employment declines during the recession of the early 1990s. In about the third quarter of 1992, there appears to be some divergence, with the gap opening even more somewhere between 1995 and 1996. Figure 3, plotted with a four-period moving average overlay to smooth seasonality, suggests that while wages do diverge between the two county categories, the divergence in wages is not as pronounced as the divergence in employment levels. The analysis that follows is designed to quantify the divergence that appears in Figures 2 and 3 and to determine if it is statistically significant after controlling for other county characteristics.

3. DD in Georgia

   A DD approach is undertaken to evaluate the employment and wage impact of the Olympic Games. (5) The idea behind the DD approach is to determine whether some statistic of interest (e.g., employment) changed more after some event for one group of observations than for another group of observations. The standard implementation focuses on differences in levels and includes dummy variables in a simple ordinary least-squares (OLS) regression indicating whether the period in question is pre- or postevent, whether the observation in question is for the affected group, and the interaction of these two indicators. Given what we observe in Figure 2, however, not only do the pre- and postevent levels of employment look different, but it appears that the rate at which employment was increasing changed as well. Consequently, a modified DD specification will be explored in addition to the standard one. Specifically, we will explore whether there was a...