Apprehensions by the U.S. border patrols dropped from 876,803 in 2007 to 556,032 in 2009. Because immigrants often send money home, we can use remittances from the United States to Mexico to infer whether the number of immigrants changed substantially during a recession. Figure 2 shows quarterly remittances to Mexico in millions of U.S. dollars as reported by Banco de México . The figure shows that remittances increased during the relatively mild 2001 recession but decreased substantially during the 2008–2009 Great Recession. These data again support the view that the number of Mexican immigrants to the United States fell during the Great Recession but not during the previous, milder recession. Moreover, Warren and Warren estimated that the net change of undocumented immigrants was negative during the Great Recession, which was related to a sharp decrease of new undocumented immigrants. The United States Department of Agriculture, Economic Research Service estimated number of full- and part-time agricultural workers fell from 1.032 million in 2007 to 1.003 million in 2008 and 1.020 million in 2009, before rising to 1.053 million in 2010.5 That is, the number of workers in 2008 was 3% to 5% lower than in the years before and after the Great Recession. Presumably the share of workers dropped by even more in seasonal agriculture, which employs most of the undocumented workers.Our agricultural workers data comes from the National Agricultural Workers Survey.The NAWS is a national, random sample of hired seasonal agricultural employees, who work primarily in seasonal crops.The NAWS is an employer-based survey. That is, it samples worksites rather than residences to overcome the difficulty of reaching migrant farm workers in unconventional living quarters.
These employers are chosen randomly within the U.S. Department of Agriculture’s 12 agricultural regions .Surveyors randomly select 2,500 employees of these growers to obtain a nationally representative sample of crop workers. Surveyors interview the more than 2,500 crop workers outside of work hours at their homes or at other locations selected by the respondent. The NAWS has a long,vertical growers visible history within farming communities, and the survey design incorporates questions aimed at data validation about legal status. Respondents receive a pledge of confidentiality and a nominal financial incentive for participation. As a result, only one to two percent of workers in the overall sample refuse to answer the legal status questions. The NAWS contains extensive information about a worker’s compensation, hours worked, and demographic characteristics such as legal status, education, family size and composition, and workers’ migration decisions. We dropped workers from the sample who were missing any relevant variable, 23% of the original survey sample. The NAWS is conducted in three cycles each year year to match the seasonal fluctuations in the agricultural workforce. Unfortunately, the public-use data, which we use, suppresses information about the cycle and aggregates the 12 regions into 6 regions. As a result, our data set consists of repeated annual cross sections of workers from 1989 through 2012. Column 1 of Table 1 presents national summary statistics for the variables used in our empirical analysis. Columns 2 and 3 provide data for California and for the rest of the country, because 37% of the sample works in California. Compared to workers in the rest of the country, Californian workers tend to have less education; have more farm experience; are more likely to be non-native, Hispanics; and are more likely to work in fruit and nut crops and less likely to work in horticulture.
After analyzing the effects of recessions on agricultural workers, we replicate the analysis for workers in construction, hotels, and restaurants, which also employ many immigrants.In March of each year, workers in the basic CPS sample are administered a supplemental questionnaire in which they are asked to report their income such as hourly wage rate and additional labor force activity such as hours worked in the previous week.8 Because information on immigration is available only since 1994, our sample period is 1994–2013. We include all workers who are 18 years and older.Three recessions occurred during our 1989–2012 sample period . The economy recovered quickly from the first of these recessions in 1990–1991. The second, 2001 recession was also relatively mild. However, the third recession, the 2008–2009 Great Recession, was much more severe and had longer-lasting economic and labor market effects than the first two. We analyze the effects of recessions on hourly earnings, the probability of receiving a bonus, and weekly hours of work of employed workers. For workers paid by time, hourly earnings are a worker’s hourly wage. For piece-rate workers, we use the workers’ reported average hourly earnings. The bonus dummy equals one for workers who receive a money bonus from an employer in addition to the wage, and zero otherwise. Weekly hours of work are the number of hours interviewees reported work at their current farm job in the previous week. The explanatory variables in all these equations are the same. The explanatory variables include all the usual demographic variables: age, years of education, years of farm experience, job tenure , gender, whether the workers is Hispanic, whether the worker was born in the United States, and whether the worker speaks English.The specification uses a legal status variable to capture the bifurcated labor markets for documented and undocumented workers. It also includes crop and regional dummies.
We have seven main explanatory variables: dummies for each of the three recessions, the recession dummies interacted with the legal status dummy , and regional unemployment rates for workers in all sectors of the economy. We use separate dummies for each recession to allow for differential effects across the recession . The interaction terms capture whether employers treat undocumented workers differently than legal workers during a recession. We include the unemployment rate because it peaks after the end of each recession . We do not report the unemployment rate interacted with the undocumented dummy because we cannot reject that its coefficient is zero in any equation. We treat all these variables as exogenous to the compensation and weekly hours of individual agricultural workers. We start by examining the effects of recessions on NAWS workers’ hourly earnings. Column 1 of Table 2 presents regression estimates for the ln hourly earnings equation. The coefficients on the demographic variables have the expected signs and are generally statistically significantly different from zero at the 5% level. Undocumented workers’ hourly earnings are 2.1% less than those of documented workers. Females earn 6.4% less than males. Hispanics earn 4.9% less than non Hispanics. Unlike most previous studies, we find a statistically significant effect of education. English speakers earn 3.9% more than non-English speakers. The coefficients on the recession dummies reflect the effect of the recession on documented workers. Documented workers’ hourly earnings rose 1.8% during the 1990–1991 recession, 4.2% during the 2001 recession, and 6.9% during the Great Recession. We draw two conclusions about the effects of recessions on documented workers. First, the hourly earning effect of the Great Recession was larger than that of the relatively minor recessions, which is consistent with literature on business cycles and the farm labor market in the 1970s . Second, in all recessions, documented workers’ wages rose, which suggests that recessions cause the hired-agricultural-worker supply curve to shift leftward relatively more than did the demand curve. The sum of the coefficients on the recession dummy and its interaction with the undocumented dummy captures the effect of a recession on undocumented workers. The 1990–1991 recession did not have a statistically significant effect on undocumented workers.
Hourly earnings for undocumented workers rose by 3.4% during the 2001 recession and 1.9% during the Great Recession. In contrast to the pattern for documented workers, the undocumented workers’ earnings rose by less during the Great Recession than during the 2001 recession. Thus, not only do undocumented workers earn less than documented workers do in general, but their hourly earnings rise less during recession than do the earnings of documented workers. That is, the wage gap between documented and undocumented workers widens during recessions. In addition to hourly earnings, 28% of the workers in our sample receive bonus payments , which supplement relatively low wage payments . These deferred payments play a similar function to that of efficiency wages in other sectors . We use a binary indicator equal to one if a worker receives a money bonus. Column 2 of Table 2 shows the results of a regression using a linear probability model . For documented workers, the probability of receiving a bonus did not rise during the two relatively minor recessions but increased by 5.8 percentage points during the Great Recession. Thus, the Great Recession not only raised documented workers’ hourly earnings, but it raised the probability that they received a bonus substantially. For undocumented workers, the probability of receiving a bonus fell by 2.9 percentage points during the 1990–1991 recession and rose by 9 percentage points during the Great Recession. Again, this result is consistent with the theory that the Great Recession caused a large supply side shock. Thus,vertical grow for both documented and undocumented workers, the Great Recession had a larger, positive effect on the probability of receiving a bonus than did earlier recessions. The unemployment rate has a statistically significant effect on the probability of receiving a bonus payment. A one percentage point increase in the unemployment rate raised the probability of receiving a bonus by 0.9 percentage points.Because our data set includes information about only employed workers, we cannot directly observe the effect of a recession on total employment. However, we can examine the effect on workers’ weekly hours. When employers have difficulty recruiting workers, they have employees work more hours per week to compensate for an unusually small workforce. For documented workers, weekly hours fell by 2.2 hours during the 1990–1991 recession, but rose by 1.1 hours during the 2001 recession, and 2.3 hours during the Great Recession. For undocumented workers, weekly hours were not statistically significantly affected during the two relatively minor recessions, but rose by 2.6 hours during the Great Recession—more than for documented workers. An increase in the overall unemployment rate by 1 percentage point lowered the weekly hours by 0.3 hours. Thus, an increase in the overall unemployment rate lowered weekly hours, but weekly hours rose during relatively large recessions.We conducted five robustness checks. First, we estimated all three equations separately for documented and undocumented workers. That is, we allowed all the coefficients to vary between these two groups instead of only the recession dummies.
The coefficients on our seven key recession variables were virtually unchanged . Second, we estimated all three regressions eliminating all newcomers , about 3,300 people or 7.5% of the sample, to see if compositional changes in the workforce during recessions are driving our results. However, the coefficients were virtually unchanged . Third, we estimated all three regressions leaving out the unemployment rate. Doing so had negligible effects on the other recession variable coefficients . Fourth, we excluded the crop dummies, in case they are endogenous. The recession variable coefficients were unaffected .Do recessions have different effects in agriculture than in other sectors of the economy that employ many undocumented immigrants, such as construction, hotels, and restaurants? To answer this question, we constructed a comparable data set based on the March Current Population Survey for 1994–2013. We can look at the effects from only two recessions, 2001 and the Great Recession, because the CPS does include certain key variables prior to 1994. It also lacks a variable on bonus payments. In contrast to the NAWS, the CPS data does not record whether an immigrant is undocumented. Therefore, we focus on immigrants in general and form interaction terms between immigrant status and the recession dummies. Otherwise, we use as similar a set of demographic variables as possible. Table 4 presents the regression results for the ln wage and weekly hours in the three sectors. In none of these three sectors did either recession affect the wages of non-immigrants or of immigrants. Presumably, wages are sticky in these sectors, partially due to union and other contracts and minimum wage laws. The unemployment rate had a statistically significant effect only in the construction sector, and that positive effect is small, as in the agricultural sector. The 2001 recession did not affect the weekly hours in these sectors.