A complete understanding of the effects of exchange rates on economic activity requires an econometric analysis that accounts for these three different forms of exposure and for their evolution overtime.4 Indeed, we find that this decomposition of channels is important for observing the evolving pattern of industry wage responsiveness to exchange rates.
The second important difference between the current paper and the previous work relates to differences in the data used. Although our sample is broader in scope by including all the 2-digit manufacturing sectors this comes at the cost of a higher level of industry aggregation. Revenga (1992) uses data from 3 and 4-digit industries. Our lower estimates for employment elasticities can possibly be explained by this aggregation difference. If most of the employment reallocation gets down among 4-digit industries within a single 2-digit industry, then one should expect very low employment elasticities to exchange rates at the 2-digit level while there is substantial employment turnover within that industry. Low net employment fluctuations in the 2digit industry could be consistent with high rates of job creation and destruction within the industry.
Davis and Haltiwanger (1997) provide some evidence in this direction in their documentation of the substantial allocative employment effects of oil price shocks. Using information at the plant level from census data, they show that there is job churning in response to oil shocks and the amount of this churning differs across the 4-digit versus 2-digit level of industry aggregation of employment data. At the 2-digit level, within manufacturing sector adjustment accounts for 67 percent of the total job re-allocation. At the 4-digit level, however, within sector reallocation is only 45 percent of the total.
These findings suggest that the more disaggregated the data, the more likely that net employment will appear to respond to industry shocks. Which level of aggregation is appropriate should depend on the question being asked. If one is mainly interested in the amount of employment and jobs, the aggregated numbers are more appropriate. If one is interested in the quality and composition of the jobs, different approaches to disaggregation could be applied.
These issues of aggregation may not mask the significant wage effects of dollar movements. Even when there is job destruction and hiring within a sector, worker earnings can still adjust substantially. Kletzer (1998, Table 6) provides insightful data relevant to this point: a matrix of post-job displacement employment by sector between 1979 and 1994 shows that displaced workers are most likely to find new positions within their old sector, and this reemployment is almost always (on average) associated with mean wage declines.
In the next section we develop a simple model of dynamic labor market equilibrium that explicitly incorporates the channels for exchange rate effects. Through the demand side of the labor market, we show that industry features – including trade orientation, costs of adjustment, and competitive structure — can influence the level and timing of labor market adjustments to exchange rates.