We also noted that operations receiving notices of violation were nearly four times more likely to have farms in more than one county than. were noncited operations. Similarly, operations with multiple violations were approximately 50% more likely to have farms in more than one county than were operations with a single violation.First is the sampling procedure. Ideally, one would have a valid list of all California agricultural operations using hired farm labor, from which a random sample of farms could be selected for inspection. From this sample one could obtain valid estimates for the frequency of workplace violations and the characteristics of operations at which they are found. Currently, no universal listing of California farm operations exists. We used the California Farm Operators Database developed and maintained by CIRS. This database is not a comprehensive listing of all farm operators in California, and it is likely that smaller operations are underrepresented in the CIRS database. National daia suggest that smaller farms have higher injury rates.” However, larger farms are more likely to employ farm workers. The likely effect of the selection against small farms on the observed results is unclear. Although the lack of a universal listing of agricultural operations does not present insurmountable difficulties, the nonrandom nature of the sampling process does present significant problems for epidemiological use. Whether or not an operation is targeted for inspection may be influenced by its past practices, location, size of workforce, and other factors. Consequently, the study sample of farms with citations may not be representative of all farms with violations,bato bucket and the study sample of farms without citations may not be representative of all farms without violations. The end result is that there may be real differences that we are unable to detect.

In addition, the California Farm Operators Database data were from the most recent year available, i.e., not necessarily from 1993 and 1994, the two years of TIPP data. There were also several limitations to the acreage reported by agricultural operators. In some cases, the operator did not report acreage, so the database indicated only that the crop was grown. For these cases, no comparisons on acreage could be made. Operators could also report several crops that would be grown over the course of the year. In some instances the operator may have reported total farm acreage for each crop reported, so the reliability of total farm acreage for each operator is limited. Limitations also exist for the TIPP data relating to the information collected and the manner in which it was collected and stored. For example, report forms did not include adequate identifying information or descriptive characteristics of the operation . TIPP developed a data collection form in 1993that facilitated information gathering. We recommend that this concept be further expanded to include more identification information and descriptive characteristics. A form could be developed reflecting the needs of participating agencies. To facilitate computer entry and analysis, these forms could be put on media suitable for reading by optical mark readers. We recognize that choice of information to be collected must reflect primarily regulatory rather than epidemiological considerations. Nevertheless, we feel that it would aid the regulatory mission by having specific information and having it collected and stored in a manner that facilitates analysis. From our results and experience with these data, we make six recommendations, The TLPP model of inter agency cooperation has demonstrated its effectiveness in the large percentage of TIPP reports that include multiple citations from different agencies. However, in some areas collaboration is incomplete. For example, the database of TIPP reports for 1994 and subsequent years does not include health-and-safety violations, which are reported to the OSHA IMIS database.

The TIPP program should make a concerted effort to engage all agencies with regulatory responsibility in this effort. For example, the California Highway Patrol should be involved to address vehicular safety concerns, such as those related to transportation of field workers to and from work. Increased inter agency collaboration is helpful in maximizing the utility of existing resources. However, the program and the agricultural community would benefit by increases in resources devoted to preventive education and field enforcement activities. Credibility and effectiveness of the program would be advanced by assuring that participating staff have linguistic fluency, cultural knowledge, and an understanding of the local agricultural employment systems. Agriculture is the dominant source of employment in poor countries. Agriculture accounts for nearly half the labor force in India and more than 70 percent across several countries in SubSaharan Africa.1 At the same time, average labor productivity in agriculture is lower than in other sectors . The gap in wages between rural and urban activities suggests a possible misallocation of labor across space . Yet, the explanations offered by the literature for the rural-urban wage gap often do not imply any misallocation. For instance, the gap could merely reflect the self-selection of heterogeneous workers into sectors . Or, people may forgo higher wages because they prefer amenities in rural over urban areas . Our paper studies a different possible reason for the agricultural wage gap. Specifically, why do the poor remain in agriculture relative to taking local non-agricultural jobs that do not require migration? The rural non-agricultural sector gives employment opportunities in many contexts . Unlike for urban non-agricultural jobs, not taking these jobs cannot be explained by a desire to stay in rural areas. With this in mind, we first ask whether there is a gap between agricultural and rural non-agricultural wages, even after adjusting for selection of heterogeneous workers. If so, what prevents labor reallocation and therefore allows this gap to persist? Our comparison within rural areas eliminates many of the common explanations for low levels of rural-urban migration. This allows us to explore a different explanation: perhaps attributes of rural non-agricultural jobs are less desirable.

Finally, under what circumstances do workers indeed move out of agriculture and into local non-agricultural activities? Our approach uses detailed data on labor allocation for a panel of agricultural workers during the peak planting and harvesting times of the 2014-2016 cropping seasons.2 Using six mobile phone surveys, one for each peak time across the three seasons, we collected information on the daily occupations and earnings for a period of one to two weeks for each worker. These data allow us to observe the sector of activity and earnings for the same individual within a very short period of time. We find that workers can indeed obtain higher earnings from working in the non-agricultural sector, and this wage gap is not due to selection on unobservable worker quality, as suggested by much of the literature on the rural-urban wage gap . Instead, a survey with workers reveals that the available non-agricultural jobs — most often in construction — are considered harder than agricultural work. Workers treat the local non-agricultural sector as a source of employment for times when agricultural work is less available. Therefore,dutch bucket hydroponic the sectoral wage gap we estimate seems more likely to reflect an equilibrium where workers prefer their familiar agricultural jobs, but are willing to do more difficult non-agricultural tasks if adequately compensated. We build this argument using three results. First, all six of our surveys collected information on agricultural wages and three of the surveys included information on non-agricultural wages. About 18 percent of laborers switch between agricultural and non-agricultural work across these three surveys. Some of this switching even occurs within a short period of just one to two weeks. Using these workers for identification, by including individual and survey fixed effects into the regression, we estimate the within-individual agricultural wage gap. Our data allow us to eliminate time-invariant correlates of unobserved ability, as well as those that vary over time, but remain constant within a short one to two week period. We estimate an agricultural wage gap of 21 percent. Put differently, the same worker can obtain 21 percent higher wages by moving out of agriculture and into non-agricultural work. The wage gap does not reflect migration costs. Rather, workers can obtain higher wages by taking nearby non-agricultural jobs, i.e. those in the same village or nearby villages that don’t require fixed migration investments. This result differs from recent evidence on the wage gains from rural-urban migration. Evidence from multiple countries tends to show that most of the rural-urban wage gaps disappear when focusing on “switchers” by introducing individual fixed effects . Therefore, sorting on unobservable ability stands out as a likely explanation for the gap between rural and urban wages.

Our data point in a different direction for the agricultural wage gap within rural areas: a meaningful gap persists even after eliminating the sorting explanation. Second, we examine what prevents labor shares from adjusting and eliminating this gap. We posed a simple question to workers: what is the top reason you work in agricultural jobs if wages in those jobs are a bit lower than non-agricultural jobs? The result reveals a disutility for certain aspects of non-agricultural work. The top explanation is that non-agricultural jobs are “too hard.” Laborers appear willing to accept lower wages in exchange for doing agricultural work — even if non-agricultural jobs are available in close proximity. We interpret this finding as suggestive that the agricultural wage gap reflects compensation for the difficulty of rural non-agricultural work — which in this setting is indeed physically demanding. The available jobs tend to involve construction, brick laying, and working in brick factories or coal mines. Alternative explanations exist. Perhaps search costs make non-agricultural work difficult to find, or some people are unable to perform the tasks required for non-agricultural work. Our third result is that workers move into the rural non-agricultural sector when they have to, i.e. when agriculture faces a bad year, their own farms are less productive, and agricultural employment is more scarce. Using variation in monsoon rainfall as a measure of agricultural productivity, we find that low agricultural productivity leads to a sharp increase in rural non-agricultural work. Going from the 90th to the 10th percentile of the rainfall distribution, conditional on village and year fixed effects, causes rice yield to decline by 63 percent, the probability of working in agriculture at harvesting to decline by 8.5 percentage points , and the probability of working in the non-agricultural sector to increase by 6.2 percentage points . Colmer uses data across all of India and similarly shows that people turn to non-agricultural work when temperature is unfavorable for agriculture. His district-level estimates include movement into the non-agricultural sector while continuing to reside in the village, and short-term migration to district towns. Our panel of workers sheds light on how workers use very local non-agricultural jobs during times of low agricultural productivity. In short, people take non-agricultural jobs when agricultural opportunities are less available. This finding suggests that the inability to find non-agricultural jobs, or the lack of ability to perform these tasks are unlikely explanations for the agricultural wage gap. Our paper adds to the literature on labor reallocation and development. This literature — focusing almost entirely on rural-urban migration — seeks to explain the large gap in ruralurban wages in developing countries. Selection on unobservable worker quality represents one of the leading explanations. In addition to this type of sorting, Bryan and Morten use data from Indonesia to show that costs of moving across space are a quantitatively important barrier. Finally, migrants may require compensation for a loss of certain rural amenities such as access to risk-sharing networks or high-quality housing .We contribute by considering an important alternative source of reallocation that has received little attention: the movement of labor from agriculture to non-agricultural work within rural areas — an important employment source across many countries .Little is known about heterogeneous selection and the barriers preventing movement from agricultural to rural non-agricultural jobs. In addition, different explanations are required to explain agricultural wage gaps within rural areas, compared to rural-urban gaps. Migration costs and differential amenities between rural and urban areas can not explain wage differences within a village. Our findings further indicate that selection does not entirely explain the agricultural wage gap within rural areas. More likely, workers require a compensating wage differential to take on hard non-agricultural tasks, or they choose to take those tasks when agricultural work is difficult to find.