Skip to navigation
Skip to main content
Changes in the U.S. and Midwestern Farm Labor Markets: NAWS and USDA Data -- Wallace E. Huffman
Markets: NAWS and USDA Data*
By Wallace E. Huffman
Department of Economics
Iowa State University
The objective of this paper is to provide an assessment of alternative regularly collected farm labor data at the U.S. and regional level. The focus is on the USDA-NASS data constructed from the Quarterly Agricultural Labor Survey (QALS), the USDL-Aguirre International data from the National Agricultural Workers Survey (NAWS), and the Bureau of the Census data from the Current Population Survey (CPS). The oldest of these surveys is QUALS, the CPS has been around since the 1970s, and the NAWS started in 1989 as a result of attempts to provide better information on the socio-demographic characteristics of SAS workers, which were
*Paper prepared for, “NAWS at 10: A Research Seminar,” University of California, Davis, CA, October 7, 2000.
believed to be difficult to reach with traditional survey methods. First, the paper summarizes the methodology underlying each survey, including a discussion of advantages and disadvantages or limitations, summarizes changes in the hired farm labor market for the U.S. and Midwest Region during the 1990s, and ends with some conclusions and recommendations on the combination of data that is most useful for understanding farm labor in the 21st Century.
II. Review of Existing Regular Farm Labor Surveys
This section provides an overview of each of the methodology underlying main existing farm labor surveys: the Quarterly Agricultural Workers Survey (QALS) conducted by NASS and the USDA, the National Agricultural Workers Survey (NAWS) conducted by the U.S. Department of Labor, and the Current Population Survey (CPS) conducted by the Bureau of Census, U.S. Department of Commerce. The CPS focus is on all U.S. households, but itsdata for occupation of employment permits generation of farm work data.
The USDA produces regular farm labor estimates from an employer or farm operator based survey. NASS started conducting probability labor surveys in January 1974, but estimates based on these surveys were not published until 1975. Survey respondents are scientifically selected from two sources–a list sampling frame and an area sampling frame of farms. NASS maintains a list of farms that hire farm workers, and farms on this list are classified by size and type. Those expected to employ large numbers of workers are selected with greater frequency than those hiring few or no workers. A second sample consists of segments of land scientifically selected from aerial photographs. Each June, highly trained interviewers locate each selected land segment and identify every farm operating land within the sample segment’s boundaries. The names of farms found in these area segments are matched against the list of farms; those not found on the list are included in the labor survey sample to represent all farms not on the NASS list. This methodology is known as multiple frame sampling, with an area sample used to measure incompleteness of the list (USDA) .
NASS controls two types of errors, sampling and nonresponse. The probability sampling method provided an improvement in reliability over previously used methods by assuring that large and specialized farm labor employers are properly represented. By applying a sampling process, standard errors or reliability of estimates can be constructed and reported. Nonsampling errors can occur in complete censuses as well as in sample surveys. They are caused by the inability to obtain correct information from each farm operation sampled, differences in interpreting questions or definitions, and mistakes in coding or processing the data. Special efforts are taken at each step of the QALS to minimize nonsampling errors.
In the list sample, which contains about 11,000 farms for each survey, operations are stratified into groups of similar farms by gross value of sales or number of hired workers. Operations having estimated gross sales over $100,000 and those having hired workers are eligible for selection into the sample. These farms account for about two-thirds of the hired workers. A simple random, stratified sample is drawn in replicates for each state. Two replicates were surveyed each quarter, and they are rotated quaraterly. The U.S. list sample included about 11,000 farms each survey.
For the labor surveys, a simple stratified sample of farms that are not accounted for by the list-frame are selected from the area frame. Since 1984, the peak number of hired workers reported on the Farm Costs and Returns Survey has been used to post-stratify the farms in preparation for sampling. About 3,500 farms are selected each quarter; they mainly represented the smaller operations which had about one-third of the hired workers. Beginning in 1989, the June Agricultural Survey placed the FCRS as the source of farms in the labor sample.
Farm labor surveys are currently conducted in January, April, and July, and October (quarterly) by the State Statistical Offices (OSS) in all States, except Alaska. The reference week is the Sunday through Saturday period that includes the 12th day of the month. This corresponds to the week specified in the general employment and wage series of other Federal agencies.
In recent years, about 70 percent of the data have been collected by telephone. Face-to-face interviews are used for operations having large numbers of workers and those requiring special handling or nonresponse follow-up. Mail questionnaires now have limited use, and in fifteen SSO’s, data are collected using Computer Assisted Telephone Interviewing. Survey data are reviewed and edited by SSO statisticians for consistency before summarization.
In QALS, farm work includes any activity linked to producing crops or livestock including nursery and greenhouse products, and animal specialities such as catfish, furs, honey, or maintenance of farm equipment. Marketing, record keeping, or other phases of farm business management, even if done off the farm, are also considered farm work. Capital improvements, housework, and home making are not included. Some work on farms is considered agricultural services rather than farm labor. It is classified as agricultural services when the provider, e.g., crew leader, labor contractor, etc., is paid on a fee or contract basis for the use of materials, equipment, or labor. Some examples are picking fruit, custom combining grains, and veterinarian work done on the farm. When a farm labor contractor is used as the grower’s source of labor for picking fruit, he recruits, transports, supervises, and pays the workers, but these workers are not counted as hired farm labor.
Estimates regularly prepared from QALS include the following. Quarterly farm employment estimates have been prepared since January 1975. The quarterly estimates were discontinued over May 1981 to July 1984. After July 1984, the farm labor employment estimates have been a joint effort of the Departments of Agriculture and Labor. Estimates are published for the United States and for California, Florida, Hawaii, and 15 regions covering the remaining states (see table 1).
An overall average wage rate and wage rate by method of pay have been estimated since 1974. In 1982, the pay categories were modified to make the series similar to wage rates reported for other industries. Rates are reported for workers paid by: “Hour,” on “Piece rate,” and by all “Other methods” of pay. “Piece rate” and “Other methods” are converted to hourly rates. Wage rates are also prepared for different types of work performed during the survey week. Since July 1982, the categories have been: “Field,” “Livestock,” “Field and Livestock,” “Supervisory,” and “Other.” In 1987, the type of worker classification was changed to what the employee was hired to do, rather than work actually performed during the survey week. Since then, machine operators have been counted as either “field workers” or “livestock workers” depending on the type of farm where they worked. Previously they were included in “Other workers.”
Since 1982, hours worked estimates have been constructed for “Hired workers,” “Self employed workers” and “Unpaid workers.”
Advantages and disadvantages of QALS are as follows. The probability sampling method of the survey permits standard errors of estimates to be constructed. These statistics are the primary indicator of reliability of estimates. The sampling error for the number of workers is about 2 to 3 percent at the national level and ranges between 10 and 20 percent for most regions. Wage rates have sampling errors of about 1 to 4 percent at both the U.S. and regional levels. Wage rates by type of farm and economic class of farm at the regional level have sampling errors ranging between 4.0 and 6.0 percent.
Farm operators who report no work during a particular survey week are not counted in the farmwork force. Currently, farm operations having gross value of sales below $5,000 account for about one-third of the total number of farms. Many operators of these small farms may not have any farm activities during the survey week and therefore are not included in the count of self employed workers. Some duplication of farmworkers may, also, exist because they could have worked on more than one farm during the survey week. No demographic data about the workers is collected in QALS, and employee benefits are not included in the compensation used to estimate average wage rates, and for some workers, this may be significant.
Some hired farm labor seems likely to be missed by QALS and other is misappropriated to agricultural services. Migratory workers who work for short time periods on any one farm and then move to another farm are sometimes missed by QALS. Furthermore, when growers used farm labor contractors for obtaining their hired labor, the hired (crop labor) is counted in expenditures on agricultural services but the workers and hours of farm work are not counted in the hired farm labor category. This treatment of workers hired through farm labor contractors seems unfortunate and can lead to misleading interpretation of long term trends in farm hired labor use.
The National Agricultural Workers Survey (NAWS) is a U.S. Department of Labor survey focused on crop workers and contracted out to Aguirre International (AI), San Mateo, CA, and started in 1989. The USDL pursued the survey with the perspective that seasonal agricultural service laborers are a difficult population to reach by traditional survey methods. It is a worksite or employer-based survey.
AI chose to develop an employer list in order to obtain a list of seasonal agricultural service (SAS) workers for its SAS labor supply surveys. AI first ranked U.S. counties by their total crop labor expenditures in the 1982 Census of Agriculture, grouping them into crop reporting districts, stratified the crop reporting districts by labor expenditures, and then selected sample counties or crop reporting districts from high, medium, and low farm-labor expenditure strata. The selected crop reporting districts included 160 to 200 counties across the United States, and AI settled on an initial sample of 60 counties in 34 site areas scattered across 25 states. The sample was later expanded to 73 counties to obtain better coverage of SAS labor. These counties have been organized into 12 SAS labor regions for preparing statistical estimates (see table 1), and we were given a CD where the data had been collapsed from 12 to 6 regions.
The Midwest Region is an aggregate of USDL regions 7 (MI, MN, and WI) and 8 (IL, IN, OH, IA, MO, KS, NE, ND, and SD). Although the Midwestern Region covers 12 states, NAWS surveys are actually completed only in six states (MI, WI, IL, IN, IA, and KY).
The goal of the NAWS is to have a nationally representative random sample of workers. The surveys started as two basic types: initial surveys collected a variety of information including work history information for the previous two years and follow-up surveys. The two-type interviews were needed to obtain an initial handle on the seasonality of SAS work. AI located about 350 employers who were initially willing to participate in terms of giving names of workers. AI interviewed about 3,000 new workers in FY 1989. The workers were asked about their SAS activities in the previous fiscal year. To get the seasonal nature of SAS work, the NAWS classifies workers as “new entrants,” “stayers, and “exiters.” This information is important for being able to make an assessment of the net change in SAS workers. In recent years, the NAWS has interviewed about 2,000 crop workers across the U.S. each year (U.S. Dept. Labor 2000).
The population sampled by the NAWS consists of all farmworkers in crop agriculture, including ones employed directly by growers and by farm labor contractors. Fieldwork in crops covers all fieldwork in SIC code 01. The field work criterion includes field packers, supervisors working the field, and all other field workers, but excludes secretaries and mechanics, H2A workers, and unemployed agricultural workers.
In recognition of the seasonal fluctuations in the agricultural workforce, interviews are conducted in cycles three times per year (beginning in February, June, and October), and the number of interviews conducted during a cycle is proportional to the amount of crop activity at that time of the year. The USDA provides the NAWS with quarterly estimates of hired and contract employment for each of the 12 NAWS regions. These numbers form the backbone of the cyclical land regional interview allocations. The annual interview total is divided into allocations for each of the three cycles. Multi-stage sampling is used to choose respondents in each cycle. The number of sites selected is also proportional to the amount of farmwork being done during the cycle. The likelihood of a given site being selected varies with the size of its seasonal agricultural payroll. Because some states such as California and Florida have relatively high agricultural payroll throughout the year, several CRDs in these states are selected for interview during each cycle. Within each CRD, a county is selected at random. Farm employers within each of the selected countries are chosen randomly from public agency records. They are primarily unemployment insurance files, Agricultural Commissioners’ pesticide registrations, and a list maintained by the Bureau of Labor Statistics and various state agencies.
The pride of the NAWS is its collection of socio-demographic data: demographics, legal status, education, family size and household composition, participation in the U.S. labor force. It also collects data on wages and working conditions (Aguirre International 2000). The NAWS publishes standard errors of their estimates at the national level. For example, the standard error for the hourly wage is 2.6 percent but about 23 percent for number of weeks doing U.S. farmwork and hours worked per week in farmwork (USDL 2000). At the regional level, the standard errors are much larger especially for all regions except for California (region 1) and Florida (region 2).
Advantages and disadvantages or limitation of the NAWS are as follows. The main advantage of the NAWS is the large amount of socio-demographic information provided on farm workers. It seems to provide relatively reliable information on SAS worker attributes in California and Florida. Some disadvantages exist. Although the NAWS claims to be a random sample of SAS workers, it is not entirely clear that they live up to this claim. The use of post-sampling weights may be justified under stable labor market conditions, but using them will bias estimates when new changes occur. There also has been and continues to be issues about how replacement farms/workers are chosen. Second, the standard errors are very large even at the 6 region level for most attributes, and state level information would be totally unreliable. Third, the NAWS does not cover livestock workers or other farm workers. There may be good reasons for this limited coverage but this also reduces their general usefulness.
The Current Population Survey, conducted by the Census Bureau, U.S. Department of Commerce, is a household-based survey that can be used to generate data on farm workers. In particular, the earnings microdata file can be used to examine demographics, earnings, and geographic characteristics of hired farmworkers (Runyan 1998). The summary information on farmworkers is based on 12 months of data, with each month’s data representing the number of individuals hired for farmwork during a 1-week period during a particular month. Annual averages are computed by summing the estimates across all months and dividing by 12. The annual average represents the average number of individuals employed at hired farmwork per week, not the total number of individuals employed.
The CPS collects information on demographic, social, and economic characteristics of the employed, unemployed, and persons not in the labor force. It is a probability sample of households, designed to represent the U.S. civilian, noninstitutional population. Each month, about 50,000 households are sampled in all 50 states and the District of Columbia. Only about 1,300 of these households supply hired farm labor. Once a household is selected, it is interviewed for 4 consecutive months, dropped from the survey for 8 months, then interviewed for a final 4 months. Approximately one quarter of the sample is changed monthly. This survey design provides for about three-quarters of the selected households to be interviewed the following month, and one-half to be interviewed the next year. During each month’s visit, trained enumerators complete a questionnaire for each household member age 15 and older. Questions focus on each household member’s labor force activity during the survey week, the calendar week containing the 12th day of the month. Information from this sample of households is expanded to provide national-level estimates.
The CPS earnings microdata file is constructed as follows. Each month, workers in about one-quarter of the CPS households are asked additional questions about weekly hours worked and earnings. For example, the 1996 CPS earnings microdata file consists of all records from the monthly quarter-samples of CPS households that were asked the additional questions during 1996. The data file contains 430,000 individuals, including over 1,290 who were employed as hired farmworkers.
The CPS was redesigned in 1994, and the redesign affects virtually every aspect of the survey, including the questionnaire, data collection methods, and the processing system (U.S. Dept. Labor 1993). As a result, data for 1994 and later years are not directly comparable to those for 1993 and earlier.
In the CPS, employed persons are defined as: Persons 15 years of age and older who, during the survey week (1) did any work as paid employees, (2) worked 15 hours or more as unpaid workers in a family enterprise, or (3) were not working but had jobs or businesses from which they were temporarily absent because of illness, bad weather, vacation, labor-management disputes, or personal reasons, whether they were paid for the time off or were seeking other jobs. Hired farm workers are employed persons who, during the survey week, did farmwork for cash wages or salary, or did not work but had farm jobs from which they were temporarily absent. Hired farmworkers include persons who manage farms for employers on a paid basis, supervisors of farmworkers, and farm and nursery workers. Hired farmworkers are classified according to the industry of the establishment where they worked: “Crop production” refers to employment in establishments primarily engaged in producing crops, plants, vines, and trees, excluding forestry operations. “Livestock production” refers to employment in establishments primarily engaged in the keeping, grazing, or feeding of livestock. “Other agricultural establishments” refers to employment in establishments primarily engaged in agricultural services, forestry, fishing, hunting, trapping, landscape and horticultural services, and other agriculture-related establishments.
The CPS collects sizeable amounts of socio-economic information on individuals, including those who are farmworkers. For example, age, gender, race, schooling, place of birth, marital status, number of children.
Some disadvantages or potential limitations of the CPS for farmworker data are as follows. First, the CPS classifies employed persons according to the job at which they worked the largest number of hours during the survey week. If an individual worked in two (or more) jobs and one of them being farmwork during the survey week, then if his/her hours of work during the survey week is smaller for farmwork than nonfarm work, he/she will not be classified as a farmworker. Hence, using the CPS definition of farmwork leads to some potential undercounting of farmworkers. Second, the CPS is a household survey, and if an individual lives in nontraditional quarters, is migratory or undocumented, he/she is unlikely to be counted. This seems to apply especially to Hispanic farmworkers (Runyan 1998). Furthermore, undocumented workers of all types try to avoid survey enumerators, and this is another source of undercounting of farmworkers from the CPS. Third, with only 1,200 households in the survey supplying hired farmworkers, reliable national and regional estimates are possible, but state level estimates would not be reliable. Although standard errors are available for national-level estimates derived from the CPS, regional level standard errors have not generally been provided but would be useful to data users.
II. Trends in Hired Farm Labor During the 1990s
QALS, NAWS, and CPS provide the basic data needed for examining trends in hired farm labor usage over the 1990s. Also, QALS and CPS can also be used to provide information on farm labor supplied by self-employed and unpaid workers, i.e., farm operators and family members, but this option is not available in the NAWS.
USDA Estimates Using QALS
The QALS provide a strong foundation for all farm labor estimates because it performs quarterly surveys and collects information on employment and hours worked of self-employed, unpaid, and hired farm workers, and on wage rates for hired farm workers. The USDA has regularly reported employment and hours data at each survey dates in each quarter, but it was not until 1996 that it started reporting an estimate of the annual number of workers employed and hours worked. Before that data, the primary way that annual estimates can be derived is to average the information over the four quarterly estimates. The recent provision of annual estimates at the national and regional level seem to be a useful addition. Hourly average wage rate data, however, have a long history of being published by the USDA for the U.S., regions, and states.
In table 2, we see that the total number of U.S. hired farmworkers seems to have peaked in the early 1990s at about 880,000, were a little lower during 1993-96, and then increased to an intermediate level at the end of the 1990s of about 875,000 workers. The Midwest region accounts for 22 to 23 percent of the total U.S. hired farm workers over 1990-1996 than the share declined to 19 percent in 1998. Also, the number of farmworkers in the Midwest Region most likely trended downward over the 1990s, being about 15 percent lower in 1998 than in 1990. During the 1990s, the U.S. real average hourly wage rate for all hired farmworkers seems to have been roughly unchanged until 1997 and 1998 when it increased by perhaps 5 percent. The average wage rate for field crop workers is always lower than the average wage rate for all farmworkers. The average hourly wage rate for hired farm labor in the Midwest Region seems to have been a little lower than for the U.S. during the first have of the 1990s, but this was reversed in 1996-1998 when the average hourly wage rate for all farmworkers in the Midwest Region was slightly higher than for the whole U.S.
The data reported in the issues of Farm Labor on hours worked are not as useful as they might be, and economists at the USDA, Resources and Technology Division, have worked to create annual estimates of hours worked for family labor (self-employed and unpaid family) and hired labor by state and for the U.S. (see Ball et al.). For individuals interested in aggregate labor supply/demand data, these numbers are potentially much more useful than those that are reported regularly by the USDA. Also, Ball et al. have constructed estimates of the opportunity wage/cost of self-employed and unpaid farmworkers which are also useful to broad issues associated with farm labor that arise in agricultural productivity and commodity cost and returns analysis.
USDL Estimates from the NAWS
Some of the estimates from the NAWS are not quite as independently constructed as one might anticipate because the overall control totals on the national and regional size of the SAS workforce seems to be based on the quarterly estimates of QALS from the previous year. The special feature of the NAWS, however, is its detailed information on socio-demographic characteristics of hired farm workers in crops. I focus on information provided on the CD for the years 1993 and 1998, especially changes that occurred over this five-year period for the U.S. and Midwest Region.
The age distribution for U.S. SAS workers is surprisingly symmetric, with about 30 percent being in the 15-34 age group (see figure 1). The distribution shifted to the right over the five year period, indicating that SAS workers became somewhat older. For the Midwest Region, the age distribution is skewed more toward younger age groups, but the highest frequency is reported for the 25-34 year-old group (see figure 2). However, between 1993 and 1998 the distribution shifted rather dramatically to the right, indicating workers became generally older.
In the NAWS, the ethnicity of U.S. farmworkers is predominately Mexican, 63 percent in 1993 and 78 percent 1998 (see figure 3). The share that is Western European, African, or Asian (as indicated by the category “none of the above”) decreased from 26 percent in 1993 to 11 percent in 1998. The share that is Mexican-American is 6 percent in both years. For the Midwest Region, the distribution of SAS workers by ethnicity is dramatically different than for the rest of the U.S. (see figure 4) In 1993, 51 percent were “none of the above” and this share declined to 27 percent in 1998. The share that is Mexican was 40 percent in 1993 and increased to 53 percent in 1998, and the share that is Mexican-American increased from 9 to 16 percent over the same time period. Hence, the NAWS shows that the share of Mexicans in the SAS workforce has significantly increased over 1993 to 1998 in the U.S. and in the Midwest Region.
Given the ethnic composition of SAS workers, the native language distribution contains few surprises. For the U.S., the native language was Spanish for 71 percent of the SAS workers in 1993 and a larger 83 percent in 1998 (see figure 5). English was the native language for 25 percent in 1993 and declined to 12 percent in 1998. For the Midwest Region, English was the majority language in 1993; 50 percent of the SAS workers reported it as their native language (see figure 6). However, Spanish was reported as the native language for 49 percent. In 1998, Spanish was the majority language by a sizeable margin with 65 percent of SAS workers reporting Spanish and 33 percent reporting English as their native language.
For years of schooling completed, four to seven years of schooling is the most frequently reported category for all U.S. SAS workers and for those educated abroad (see figure 7). However, a larger share of those educated abroad report the lower completion levels than for all workers. Furthermore, the education distribution of SAS workers does not change much between 1993 and 1998. For the Midwest Region, the education distribution for SAS workers differs dramatically for those educated abroad compared to all workers (see figure 8). For all workers, the relative frequency is lowest for the 0-3 years of schooling completed category (7% in 1993) and it increases monotonically as one moves across the figure to higher schooling completion levels (38 % for 12+ years). For those that were educated abroad, almost a majority (49%) of them fall in the four to seven years of schooling competed category. The frequency distributions in 1998 are similar to those in 1993. Hence, there is no evidence that the education level of SAS workers is changing.
The wage rate data are graphed using hourly wage rate categories in the 1998 price level (see figure 9). In 1993, it shows that for the U.S., 33 percent of the SAS workers earned less than $5.15 per hour which became the U.S. minimum wage later. In 1998, 31 percent of all SAS workers earned less than this amount. Wage rates in the Midwest Region are skewed to the right relative to the U.S. in both years, but the difference is larger in 1998, when the Midwest Region seems to by paying significantly higher wage rates for SAS labor.
The basis for pay for the U.S. and Midwest Region is predominately hourly, and the second most frequently reported type is piece-rate pay. Piece-rate is somewhat more frequently reported in non-Midwest Regions, and over 1993 to 1998, the hourly pay base gained relative to the other pay bases for the U.S. and Midwest Region (see figures 10 and 11).
The NAWS also collects and reports information on the “legalized” versus “Unauthorized” status of farm crop workers. In spite of the glamorous goals of IRCA (Martin et al 1995), over 1989 to 1998, there has been a dramatic increase in the frequency with which SAS workers report being “unauthorized,” starting at 8 percent in 1989 and strongly trending upward to about 52 percent 1998 (see figure 12).
Estimates Based on the CPS
The hired farm labor data prepared from the CPS provide information on the number of workers, their distribution by type of work and by region, weekly earnings, and some socio-economic characteristics. Using the CPS, the total number of hired farm workers is about 890 thousand in 1990 and this number trends down and is about 10 percent lower in 1994 (see table 3). It then increases and is at the end of the period to the numbers at the beginning of the period. The distribution of workers by type of work or establishment has remained quite stable, about 50 percent in crop production, 41 percent in livestock production, and 9 percent in other. The regional distribution of hired farm workers has shifted over time with a greater share being in the West, an increase of roughly 10 percentage points over decade. This shift has been at the expense of the South and Midwest. For the U.S. the average real weekly earnings was relatively unchanged over 1990-93 and then increased by about 5 percent for 1995-1998.
For the U.S., the age distribution for hired farm workers from the CPS data looks similar to that of the NAWS, even though they cover a different population of hired farm workers (see figure 13). However, in the Midwest Region the CPS age distribution is quite different from the NAWS. In the CPS data, 34 percent are less than 20 years of age, compared to 14 percent in the NAWS, and only 20 percent are 25-34 years of age, compared to roughly 30 percent in the NAWS. The CPS data also show a larger share of hired workers in the over 55 years of age category.
For racial/ethnic group, the CPS data give a much larger share that is white than is implied by the NAWS at both the U.S. and Midwest Region level. The contrast is especially large for the Midwest region where the CPS data suggest that 96 percent of the hired farm workers are white (see figure 14). Part of the differences are undoubtedly due to differences in coverage of the two survey.
For schooling, the distribution of schooling from the CPS data shows a larger share of the workers at higher levels of education for the U.S. than for the NAWS (see figure 15). For the Midwest Region, however, the (1996) CPS and 1998 NAWS present a similar distribution for education.
Figure 16 presents information on the weekly earnings distribution. It shows that a much larger share of the earnings distribution for the Midwest distribution is in the lowest earnings category than for the U.S. The U.S. distribution has a spike in the 3rd of 7 earnings groups, while for the Midwest Region, the first three weekly earnings groups have about equal shares. These data, however, do not seem to be comparable to the NAWS’ hourly wage rate data.
IV. Conclusions and Implications
The methodology, goals, and summary data for three hired farm labor surveys have been reviewed. QALS and NAWS are quite expensive to collect, but the CPS data are largely taking advantage of an existing data set that has very little marginal cost. The major advantage of the NAWS has been the information on socio-economic characteristics of hired farmworkers, but it provides no information on these characteristics for the other major components of the farm labor workforce–self-employed and unpaid family workers. The NAWS seems to identify a large share of low-wage, Spanish-origin workers, but it may over estimate their relative importance in the hired form work force. The NAWS seems to have good reliability at the national level, but the regional estimates appear to be quite noisy. However, even for the national estimates, one must ask whether too much emphasis is being placed on the post-sample weights. These weights are clearly incorrect when weather or economic conditions in an area change dramatically from one year to the next, causing the timing of farmwork and the flow of farmworkers to deviate from the immediate past.
The CPS provides a useful alternative to the NAWS for social-demographic information on hired farm crop workers and new information for the self-employed and unpaid farmworkers. It, however, has good reliability at the national level, standard errors seem likely to be much larger at the regional level (e.g., for 6 or 12 regions), and it is totally unreliable at the state level. I, however, believe that it should be used regularly to provide information on farmworkers.
QALS is a relatively good data base to work from for quarterly estimates at the national, regional, and state level, but more effort needs to be placed on turning these into highly reliable estimates of annual employment and hours of work. This is something that the USDA can do, and the data should be made available regularly. Hence, one could easily make the argument for using the QALS and CPS for the needed farm labor data in the 21st Century, and possibly using the resources currently going into the NAWS to support research on farm labor issues.
Aguirre International, “National Agricultural Workers Survey: Public Access Documentation,” San Mateo, CA, March 2000.
Ball, V.E., F. Gollop, A. Kelly, and G. Swinard. “Patterns of Productivity Growth in the U.S. Farm Sector: Linking State and Aggregate Models,” American Journal of Agricultural Economics 81(1999): 164-179.
Bowie, C.E., L.S. Cahoon, and E.A. Martin. “Evaluating Change in the Estimates.” Bureau of Labor Statistics. Vol. 116, No. 9, Sept. 1993.
Bregger, J.E. and C.S. Dippo. “Why is it Necessary to Change?” Monthly Labor Review. Bureau of Labor Statistics. Vol. 116, No. 9, Sept. 1993.
Huffman, Wallace E. “An Assessment of the Process Underlying RAW Calculations,”in Report of the Commission on Agricultural Workers, 1989-1992. Appendix I: Case Studies and Research Reports. Washington, DC: U.S. Government Printing Office, Nov. 1992.
Martin, P.L., W.E. Huffman, R. Emerson, J.E. Taylor, and R.I. Rochin. Immigration Reform and U.S. Agriculture. Oakland, CA: University of California, Division of Agriculture, Publ. 3358, 1995.
Polivka, A.E. and J. Rothgeb. “Redesigning the Questionnaire.”Bureau of Labor Statistics. Vol. 116, No. 9, Sept. 1993.
Runyan, J.L., Profile of Hired Farmworkers, 1996 Annual Averages, ERS, USDA, Agricultural Economics Report No. 762.
U.S. Department of Agriculture. Farm Employment and Wage Rates, 1910-1990. National Agricultural Statistics Service, USDA. Statistical Bulletin No. 822, March 1991.
U.S. Department of Agriculture. Farm Labor. National Agricultural Statistical Service, USDA. Various issues 1995-2000.
U.S. Department of Labor, Monthly Labor Review. Bureau of Labor Statistics. Vol. 116, No. 9, Sept. 1993.
U.S. Department of Labor, “Findings from the National Agricultural Workers Survey (NAWS) 1997-1998: A Demographic and Employment Profile of United States Farm Workers,” USDL, March 2000.
Table 1. Comparison of Regional Groupings of
States for NAWS and QALS
USDL NAWS Farm Labor Regions
Regions States (QALS)
1 CA CA
2 FL FL
3 CT, ME, MA, NH,
NY, RI, VT Northeast I
4 DE, MD, NJ, PA Northeast II
5 NC, VA, KT, TN, WV Appalachian I & II
6 AR, LA, MS, AL, GA,
SC Delta & Southeast
7 MI, MN, WI Lake
8 IL, IN, OH, IA, MO, Corn Belt (I & II) &
KS, NE, ND, SD Northern Plains
9 OK, TX Southern Plains
10 ID, MT, WY, CO, NV,
UT Mountain I & II
11 AZ, NM Mountain III
12 OR, WA Pacific
Table 2. Number of Hired Farm Workers and Average Wage Rates: U.S. and Midwest, 1990-1998
Variable 1990 1991 1992 1993 1994 1995 1996 1997 1998
Total number workers
U.S. 892.2a 910.2a 866a 803 779 832 832 876.5 879.5
Midwest 205.5b 183a 185 180 188.4 169.8
Wage rate ($/hr)
U.S. - all 5.52 5.79 6.06 6.67 6.39 6.54 6.78 7.35 7.47
- all real 1998 dol. 6.88 6.93 7.04 7.52 7.03 7.00 7.04 7.47 7.47
- field crop 5.23 6.02 6.13 6.34 6.66 6.97
Midwest - all 5.31 6.15 6.55 6.89 7.44 7.80
Source: USDA, QALS
a Averages of quarterly survey employment members.
b Used 1986 regional employment shares as weights.
Table 3. Number of Hired Farm Workers, Distribution by Type of Work and Region, and Median Earnings, 1990-1998
Variable 1990 1991 1992 1993 1994 1995 1996 1997 1998
Total number workers 886 884 848 803 792 849 906 875
Type of establishment/work
Crop production 419 449 409 436 411 433 451 458
(47.3) (50.7) (48.2) (54.3) (51.8) (51.0) (49.8) (52.3)
Livestock production 390 363 364 313 315 345 369 368
(44.0) (41.0) (42.9) (39.0) (39.8) (40.6) (40.7) (42.1)
Other 77 72 75 54 66 71 86 49
(8.7) (8.2) (8.9) (6.7) (8.4) (8.4) (9.5) (5.6)
Northeast 6.9 6.1 6.1 6.1 6.0 7.1 7.2 7.3
South 35.6 37.1 37.8 37.5 39.4 32.3 30.9 31.4
Midwest 24.1 23.3 23.7 21.4 18.4 20.0 23.9 18.5
West 33.4 33.5 32.4 35.0 36.2 40.6 38.0 42.2
Median weekly 200 210 200 220 238 240 250 260
Real, 1998 dol. 249 251 232 248 262 257 260 260
Source: Runyan 1998