IMMIGRATION POLICY AND THE AGRICULTURAL LABOR MARKET: SPECIALTY CROPS IN THE UNITED STATES

Transcription

1 IMMIGRATION POLICY AND THE AGRICULTURAL LABOR MARKET: SPECIALTY CROPS IN THE UNITED STATES Nobuyuk Iwa Internatonal Agrcultural Trade and Polcy Center Food and Resource Economcs Department PO Bo Unversty of Florda Ganesvlle, FL Robert D. Emerson Internatonal Agrcultural Trade and Polcy Center Food and Resource Economcs Department PO Bo Unversty of Florda Ganesvlle, FL Orachos Napasntuwong Internatonal Agrcultural Trade and Polcy Center Food and Resource Economcs Department PO Bo Unversty of Florda Ganesvlle, FL Selected Paper prepared for presentaton at the World Trade Organzaton Impacts on U.S. Farm Polcy Conference, New Orleans, Lousana, June 1-3, 2005 Copyrght 2005 by Nobuyuk Iwa, Robert D. Emerson, and Orachos Napasntuwong. All rghts reserved. Readers may make verbatm copes of ths document for non-commercal purposes by any means, provded that ths copyrght notce appears on all such copes.

2 IMMIGRATION POLICY AND THE AGRICULTURAL LABOR MARKET: SPECIALTY CROPS IN THE UNITED STATES Nobuyuk Iwa, Robert D. Emerson, and Orachos Napasntuwong Unversty of Florda ABSTRACT Mode 4 of the GATS addresses the temporary movement of natural persons. Snce WTO negotatons have yet to sgnfcantly address unsklled labor moblty, our analyss s based on U.S. mmgraton polcy and labor market rules wth the presumpton that t wll be nformatve for forthcomng WTO negotatons. Of course, t s also drectly relevant for U.S. mmgraton polcy and labor markets. The paper eamnes the epected ob length for farm workers under varyng degrees of legal status. The Natonal Agrcultural Workers Survey (NAWS) found that 78 percent of agrcultural workers n the U.S. were foregn-born for the years Of those, 68 percent were undocumented. Current mmgraton reform proposals vary n how they would alter the m of workers under varyng forms of work authorzaton. The concern for agrcultural employers, and producers of hghly labor-ntensve specalty crops n partcular, s the potental effect on the avalablty of labor for tme senstve actvtes. Our mcro-econometrc analyss of the NAWS data addresses the problem n two steps. Frst, an ordered probt model s estmated to characterze workers by legal status (unauthorzed, authorzed, permanent resdent, or ctzen), and then a duraton model s estmated condtonally upon legal status of the worker. Ths approach permts askng the relevant what f queston such as the epected ob duraton n agrculture for an unauthorzed worker, were the worker to become legal as a result of legslatve or polcy changes. Frst, our fndngs are that ob duraton s affected by legal status and that the estmaton adustment va the ordered probt model s relevant. Second, we fnd that epected ob duraton would generally be longer for unauthorzed workers f they were workng n a legal status rather than n ther unauthorzed status.

3 IMMIGRATION POLICY AND THE AGRICULTURAL LABOR MARKET: SPECIALTY CROPS IN THE UNITED STATES Introducton WTO negotatons can potentally mpact the terms of avalablty of labor for agrculture. Mode 4 of the General Agreement on Trade n Servces specfcally addresses the temporary movement of natural persons. Whle most estng commtments have addressed sklled labor, there s consderable potental for pressure from developng countres to lberalze less-sklled labor mgraton. Wnters, et al., estmate that: f quotas [on temporary nflows of workers] were ncreased by an amount equal to 3 per cent of developed countres labour forces, there would be an ncrease n world welfare of $156 bllon per year. Both developed and developng countres would share n these gans and they would owe more to greater moblty of less sklled workers than to that of more sklled workers. (p. v) Clearly, these gans arse from economy-wde gans, not ust agrculture. However, agrculture has a long hstory of employment of unsklled foregn labor, and the pattern has been towards the ncreased employment of foregn workers. Ths paper addresses the current stuaton for specalty crop agrculture, and focuses specfcally on legal status of the worker and ts relaton to the epected duraton of a person s ob n specalty crop agrculture. Our emphass s toward U.S. mmgraton polcy and labor market rules rather than WTO negotatons snce the former s what s currently n estence, but may be nformatve for forthcomng WTO negotatons. Specalty crop agrculture, whch faces ntense competton from nternatonal markets, s a very labor-ntensve ndustry. The nature of most specalty crops s that they are hghly labor-ntensve n contrast to most U.S. agrcultural producton. Labor ependtures account for The authors are grateful to Susan Gabbard, Trsh Hernandez, Alberto Sandoval and ther assocates at Agurre Internatonal for assstance wth the NAWS data, and to Danel Carroll at the U.S. Department of Labor for grantng access and authorzaton to use the NAWS data. Ths research has been supported through a partnershp agreement wth the Rsk Management Agency, U.S. Department of Agrculture; by the Center for Internatonal Busness Educaton and Research at the Unversty of Florda; and by the Florda Agrcultural Eperment Staton. The authors alone are responsble for any vews epressed n the paper. 1

4 37% of U.S. frut, vegetable and hortcultural crop producton ependtures (2002 Census of Agrculture). By contrast, labor ependtures for all of U.S. agrculture represent only 13% of producton ependtures. Moreover, the labor requrements often vary consderably among the varous specalty crops, and even wthn a crop there can be consderably dfferent labor requrements dependng on whether the crop s produced for the fresh or for the processng market (processng tomatoes are mechancally harvested whle fresh market tomatoes are hand-harvested). Mechancal harvestng s the norm among maor U.S. farm commodtes such as corn, wheat, soybeans, and cotton, for eample; however, a substantal porton of U.S. frut and vegetable acreage remans totally dependent on hand harvestng (Sarg, et al.). Florda, for eample, s a maor producng state for fruts, vegetables, and hortcultural products. Labor ependtures represented 33% of Florda agrcultural producton ependtures across all commodtes n the 2002 Census of Agrculture. Vegetable and melon producers devoted 40% of producton ependtures to labor, and the fruts and tree nuts group allocated 38% of producton ependtures to labor n These two commodty groups are maor Florda commodtes, and domnate agrcultural employment n the state. The two commodty groups accounted for 46% of Florda agrcultural labor ependtures n 2002: frut and tree nut farms, 28%; and vegetables and melons, 18%. Both commodty groups are domnated by ntense seasonal harvestng labor requrements. Not only s the product sde of the specalty crop market operatng subect to strong nternatonal compettve forces, but the labor market s also an nternatonal labor market. The agrcultural labor market s heavly dependent on foregn-born workers. Accordng to the Natonal Agrcultural Workers Survey (NAWS) report (Carroll, et al., 2005), 78 percent of agrcultural workers were foregn-born for the years In addton, most of the foregn-born workers 2

5 were undocumented: 68 percent for the same perod (Carroll, et al., 2005). 1 Therefore, the effects of mmgraton polcy change on the agrcultural labor market have receved much attenton both economcally and poltcally. The most mportant mmgraton polcy change n recent years for the agrcultural labor market was the Immgraton Reform and Control Act (IRCA) of IRCA granted amnesty to a substantal number of undocumented agrcultural workers, enttlng them to work legally n the Unted States. Just before the passage of IRCA, many farmers and legslators epressed concern about ts possble effect on the agrcultural labor market. Ther predcton was that undocumented agrcultural workers who receved amnesty would leave agrculture for other employment opportuntes, whch would lead to serous labor shortages and wage ncreases n agrculture. 2 Lmted emprcal work has been done on the relatonshp between legal status and farm work duraton (Hashda and Perloff 1996, Tran and Perloff 2002, and Emerson and Napasntuwong 2002). In addton, no duraton study has focused on specalty crop agrculture whose producton depends more on physcal labor than other agrcultural sectors. The prevous studes whch address the overall agrcultural sector conclude that estmated duratons for documented, n contrast to undocumented, workers are sgnfcantly longer. Among these, the most comprehensve study s Tran and Perloff (2002). Usng the Natonal Agrcultural Workers Survey (NAWS) data for the years , Tran and Perloff estmate a statonary, frst-order Markov model of employment turnover, and calculate the steady-state probablty for each demographc group to work n agrculture. They conclude that Predctons made when the 1986 Immgraton Reform and Control Act was passed that grantng people amnesty would nduce most of them to leave agrculture were ncorrect, (p. 427) and.. the steady state probablty of 1 Carroll, et al. (2005) note that 53 percent of all workers were unauthorzed for work n the U.S. Snce only foregn-born workers can be unauthorzed, t follows that 68 percent of the foregn-born were unauthorzed. 2 See Tran and Perloff (427-28) for a detaled dscusson of ndustry and legslatve concerns. 3

6 workng n agrculture s hgher for someone wth amnesty than for an undocumented worker, so that IRCA ncreased the long-run probablty that people granted amnesty stayed n agrculture. (p. 437) However, ths concluson s a lttle problematc. As the authors mentoned n ther work, the porton of undocumented workers n the agrcultural labor force grew substantally n the 1990s. Accordng to the NAWS data, the porton of undocumented workers n specalty crop producton rose from 16% for the years , to 43% for the years and to 53% for the years Ths mples that there has been a large-scale nflow of undocumented workers nto the specalty crop agrcultural labor market and a large-scale outflow of documented workers from t. The latter mght mean that documented workers tend to leave specalty crop agrculture n the long-run: the opposte observaton to ther concluson. There are some concerns that mght lead to statstcal problems n ther work. Frst, a data sample ( ) s taken n a transtonal perod n the sense that workers granted amnesty mght not have had enough tme to move to other ndustres. It s also a transtonal perod n another sense that the legal status of many workers changed. The study s unable to control for ths status change usng the observed status at the tme of ntervew, the only legal status nformaton avalable n the NAWS data. As a result of the sample used, the study cannot capture the maor nflow of undocumented workers from foregn countres after IRCA and who have become a maor component of the labor force n agrculture. The most serous problem, however, s that the study tres to estmate a probablty matr and a steady state for the whole mgraton process usng data from only a small sector (the agrcultural labor market). Most mgraton for any status of worker would be from non-agrculture to non-agrculture, and most would not work n agrculture at all. It may be dffcult to estmate the whole mgraton pattern wthout data from all sectors. In ths presentaton, we present an alternatve method (duraton model wth sample bas correcton) to 4

7 estmate the effect of the legal status of a worker on duraton n farm work. Based on the estng studes whch have used the duraton model (Hashda and Perloff 1996, Emerson and Napasntuwong 2002), we develop the Heckman-type two-stage method, wth the ordered probt model n the frst stage and the duraton model n the second stage. The sample selecton bas ssue should be nvestgated frst. Duraton for a worker wth a legal status s observed only f the worker s n that legal status. Each foregn-born worker chooses hs/her legal status, consderng condtons such as hs/her ndvdual demographc characterstcs, cost of applcaton, and beneft of the status. Wthout correctng for ths selecton process, the duraton model wll yeld based estmators. Hashda and Perloff (1996) correct selecton bas usng Lee s etenson of Heckman s two-stage sample selecton method (Lee 1983). In the frst stage, the multnomal logt model s run to calculate a correcton term assumng the error term has a Gumbel dstrbuton. The second-stage duraton model wth ths correcton term does not generally yeld consstent estmates wth the normal dstrbuton assumpton of error term n the duraton model. 3 We wll use the ordered probt model n the frst stage for two reasons: (1) ths s consstent wth the assumpton about the error term n duraton model n the second stage and (2) the multnomal logt does not account for the ordnal nature of the legal status. Consderng the advantages n the labor market, they can be ordered as ctzen, permanent resdent, authorzed, and unauthorzed workers. 4 Net, treatment of completed and uncompleted employment spells of workers should be consdered. Hashda and Perloff (1996) and Tran and Perloff (2002) use only completed spells, whle Emerson and Napasntuwong (2002) use only uncompleted spells. There are further dstnctons n how spells have been defned n the lterature. Hashda and Perloff (1996) defne 3 Lee s method yelds consstent estmator under very restrctve condton (Bourgugnon et al. 2004). 4 The defnton of each legal status s gven n the Data secton. 5

8 the duraton varable as the average duraton of completed spells of farm employment by a worker. Tran and Perloff (2002) work wth employment transtons among three types of spells: agrcultural employment, nonagrcultural employment, and unemployed or abroad. They recorded a transton on a monthly bass over a two-year work hstory among the three above types of spells wthout regard to employer. Emerson and Napasntuwong (2002) defne the duraton varable as the number of years reported workng n U.S. agrculture. At ths pont our estmaton uses multple completed spells per worker of agrcultural employment at a sngle task. Our current defnton s closest to the one used by Hashda and Perloff (2002), and specfcally addresses varatons n ndvdual ob duraton by farm workers. Methodology The basc structure of the Heckman-type two-stage method s specfed wth the ordered probt model for the frst stage and the duraton model for the second stage. The ordered probt model s used to eplan the legal status of worker as a functon of the ndvduals demographc and polcy varables denoted as vector ). A foregn-born worker s legal status (J) takes on four ( values: 0=unauthorzed, 1=authorzed, 2= permanent resdent (green card holder), 3=ctzen. Wth the famlar argument of latent regresson (Greene 2003), we can assume that an unobserved varable J * s censored as follows: J J J J = 0 = 1 = 2 = 3 * f J µ 0, * f µ < J µ, 0 f µ 1 < J f µ < J 2 * * µ,. 1 2 where J * = + ; s a vector of eogenous characterstcs of ndvdual ; and s a dsturbance term. The characterstcs nclude gender, martal status, Englsh speakng ablty, race 6

9 (black, whte, and other), ethncty (Hspanc and other), age, age squared, educaton, educaton squared, US farm eperence, US farm eperence squared, and the year of ntervew (before 1993, after 2001, and n-between). 5 We assume that s normally dstrbuted wth a mean of zero and a standard devaton of. Then the lkelhood functon can be epressed as, 1 ),, ( Φ Φ Φ Φ Φ Φ = = = = = J J J J data L µ µ µ µ µ µ µ (1) where ndcates the cumulatve dstrbuton for the standard normal. Φ( ) Suppose the cumulatve dstrbuton functon of farm work duraton (t ) for person wth legal status s gven as ) Pr( ) ( t t t F < =. We denote ts densty functon as The probablty for the spell to be of length of at least t, usually called the survval functon, s gven as (t). f ) ( 1 ) ( t F t S =. Suppose that the log of the spell s normally dstrbuted wth mean τ ln and varance. Then, the survval functon s epressed as Φ = t t S lnτ ln 1 ) (. The hazard rate, the rate at whch the spell s completed after duraton t, s ) ( ln ln ) ( t S t t t h τ φ =, 5 See the Data secton below for addtonal detal. 7

10 where φ ( ) s the probablty densty for the standard normal dstrbuton. Net, we assume that the mean duraton of a spell ( ln τ ) depends on ndependent varables z (gender, martal status, age, age squared, educaton, US farm eperence, Englsh speakng ablty, race, ethncty, avalablty of free housng, task, regon (Calforna, Florda, and other), the year of the ntervew (after 2001 or not), dummy varable for seasonal workers) so that ln τ = z β. Then, the duraton can be epressed as ln t = z β + u where u N(0, ). However, duraton ~ t s observed only f person has legal status. Ths s a typcal case for selecton bas. Assumng e and u are bvarately normally dstrbuted wth correlaton coeffcent ρ, the mean of the log of the duraton condtoned on the legal status of person s corrected as E [ ln t ln t s observed] = z β + ρ λ where λ s the correcton term for the selecton bas whch s gven as 6 µ µ 1 φ φ λ = µ µ 1 Φ Φ Note that we can use the result of the ordered probt model n the frst stage for µ and µ. Also note that µ 1 =,µ = from the assumpton of normal dstrbuton. In the 1 3 second stage of ths Heckman type two-stage method, we estmate equaton (2) below by ordnary least squares wth only completed spells. 6 Correcton term s set to zero for natve-born ctzen. 8

11 e z e z t + + = + + = λ β β λ ρ β λ ln (2) We can also show that the condtonal varance of the log of the duraton would be [ ] [ ] t t δ ρ s observed ln ln var =, where Φ Φ + Φ Φ = µ φ µ φ µ φ µ φ µ δ. Then a consstent estmator of s gven by 2 [ ] = + = n n e / ˆ ˆ ˆ ˆ δ β λ. We can obtan the estmator of the asymptotc covarance matr for by substtutng these results n the formulaton n Greene (2003). ] ˆ, ˆ [ λ β β The dffculty n the farm worker duraton model s that t has two sources of nconsstency. The observatons are censored n the sense that the duraton of a person wth a partcular legal status s observed only f the person has that status. Some observatons are also censored n the sense that they are ncomplete. On the other hand, the legal status model (ordered probt) does not have any restrctons on the observatons, so that t should be a consstent estmator. The above method takes care of the selecton bas by usng correcton terms for the mean duraton. The current estmaton approach drops uncompleted spells from the data set, ntroducng an unknown etent of bas n the estmaton. However, gven the sze of the data set, the bas s beleved to be mnmal. 9

12 Data The data used n ths study are obtaned from the Natonal Agrcultural Workers Survey (NAWS) (Offce of Assstant Secretary for Polcy). We used the study perod from 1989, when the NAWS was frst avalable, to the most recent year, Ths secton wll descrbe the defntons of each varable we used n our model. In addton, we use only the data of laborers who worked n specalty crop agrculture. Specalty crops nclude all crops ecept corn, wheat, barley, oats, rce, rye, cotton/cottonseed, sugar beets, tobacco, soybeans, sugarcane, and multple feld crops. Legal status s a dscrete varable rangng from 0 to 3. Status 0 = unauthorzed workers means that the worker s undocumented (dd not apply to any legal status or applcaton was dened) and also ncludes those who had no work authorzaton even f they were documented. Status 1 = authorzed workers or documented workers; these workers must have a work authorzaton and may fall nto any of the followng statuses: havng border crossng card/commuter card, wth pendng status, or temporary resdents holdng a non-mmgrant vsa. Status 2 = permanent resdents or green card holders who have the rght to resde and work n the U.S., and status 3 = ctzens who are a ctzen by brth or a naturalzed foregn born ctzen. The varable Englsh measures the capablty of speakng Englsh, and does not nclude Englsh readng sklls. The varable s a dscrete varable rangng from 1 to 4, where 1= not able to speak Englsh at all, 2 = speaks a lttle Englsh, 3 = somewhat able to speak Englsh, and 4 = speaks Englsh well. Hspanc s a dummy varable ncludng Mecan-Amercan, Mecan, Chcano, Puerto Rcan, and other Hspanc ethnc groups. Black (or Afrcan Amercan) and Whte are also dummy varables derved from a queston regardng ther race whch may also be Amercan Indan/Alaskan Natve, Indgenous, Asan, Natve Hawaan or Pacfc Islander, or others. Age 10

13 was calculated from the dfference between the date of ntervew and the date of brth, ecept for the earler years when age was asked drectly n the questonnare. Educaton s the hghest grade level for educaton, and t ranges from 0 to 20. Eperence s the number of years dong farm work n the U.S. (not ncludng farm work eperence abroad). Task s the task at the tme of ntervew. Although task s also asked for each perod of work n the past two years, we use only the task at the tme of ntervew for each duraton. Although the orgnal questons have over 100 task codes, tasks are grouped nto s categores as follows: 1 = pre-harvest, 2 = harvest, 3 = post-harvest, 4 = sem-sklled, 5 = supervsor, and 6 = others. They are argued to be ordered by ncreasng skll requrements. Seasonal Worker s a dummy for workers who were workng on a seasonal bass for the employer at the tme of ntervew. Free housng s a dummy varable for workers (or workers and ther famly) who receve free housng from ther current employer. It does not nclude those who own the house or lve for free wth frends or relatves. It also ecludes those who pay for housng provded by employers or by the government or charty. The dummes for Florda and Calforna are the state for each work duraton, and not necessarly the state at the tme of ntervew. The Before 1993 dummy varable s for all years pror to 1993 when the maorty of IRCA legalzaton was granted, and After 2001 s the years after the September 11, 2001 event. Duraton or farm work spells s a varable created from the work grd n the questonnare. It s the dfference between the endng dates and startng dates for each farm work spell, and only ncludes completed spells (all spells completed at the tme of ntervew). 11

14 Ordered Probt Model for Legal Status Here we estmate the ordered probt model for legal status for foregn-born farm workers usng NAWS data. Table 1 shows the estmates for parameters and asymptotc standard errors (gven n the parentheses) usng 29,194 observatons of foregn-born farm workers. Usng a 0.05 sgnfcance crteron, we fnd that all coeffcents ecept educaton squared are statstcally sgnfcant. The thrd column of Table 1 shows the margnal effect of each varable on the probablty of a worker beng legal. The probablty of worker beng legal s gven by * Pr ob( J > µ 0) = 1 Φ( µ 0 ). Then the margnal effect of varable k evaluated at the mean s φ ( µ 0 ) k for the contnuous varables 7 and Φ( µ 0 k k ) Φ( µ 0 k k k ) for the dummy varables, where k and k are varables and coeffcents ecludng k. Females, marred, workers wth hgher Englsh speakng ablty, non-black, whte, and non-hspanc are statstcally sgnfcantly more lkely to have a more advantageous legal status, all else beng the same. We also fnd that both age and US farm eperence have a sgnfcant nonlnear effect on legal status. US farm eperence has a postve effect on legal status throughout a person s workng lfe (up to 161 years), and age has postve effect on legal status up to 71 years, both of whch are postve over the entre relevant range. Educaton has a sgnfcantly postve lnear effect on legal status. We fnd that the greatest postve margnal effect s from the female dummy, followed by Englsh speakng ablty and the before 1993 dummy. The greatest negatve margnal effect s from the Hspanc dummy, followed by the after 2001 dummy. Note that, holdng all other characterstcs the same, the workers ntervewed before 1993 are twelve percent more lkely and those ntervewed after 7 The margnal effect for varables wth a squared term s gven by φ ( µ 0 )( k + 2 k _ sqk ) where k _ sq s coeffcent for the squared varable. Also, we treated Englsh speakng ablty as a contnuous varable. 12

15 2001 are fourteen percent less lkely to be legal, compared to those ntervewed between these two perods. Fnally, Table 2 shows the actual-predcted legal status table. A worker s predcted to be status 0 (unauthorzed) f ˆ < ˆ µ 0, and s predcted to be status 1 (authorzed) worker f ˆ µ ˆ 0 < < ˆ µ 1 and so on. Table 2 shows that 79 percent of unauthorzed workers are correctly predcted to be unauthorzed. In the same way, 21 percent of authorzed workers, 70 percent of permanent resdent and 21 percent of ctzens are correctly predcted n ther legal status. Our ordered probt model does a very good ob n dstngushng status 0 workers from legal workers, but many of status 1 workers and status 3 (ctzen) workers are mstakenly predcted to be status 2 (permanent resdent) workers. Duraton Model wth Selecton Bas Correcton Here we estmate the duraton model wth selecton bas correcton usng the results from the ordered probt legal status model n the frst stage. Table 3 shows estmates for parameters and asymptotc standard errors (gven n the parentheses) for farm workers wth each legal status. Status 0 (unauthorzed) workers have 39,907 observatons, status 1 (authorzed) workers have 14,988 observatons, status 2 (permanent resdent) workers have 33,346 observatons, and status 3 (ctzen) workers have 15,012 observatons. Based on asymptotc standard errors usng a 0.05 sgnfcance crteron, the coeffcents on the selectvty varable, λ, are all sgnfcant ecept for permanent resdent workers. That s, usng ordnary least squares wthout correctng for selectvty would lead to bas n all equatons ecept for permanent resdent workers. Many varables have a statstcally sgnfcant effect on duraton n a common drecton for all equatons. Regardless of the legal status, workers n tasks requrng hgher skll, workers wthout free housng from employers, workers n Calforna, workers n Florda, and workers 13

16 ntervewed after 2001 are statstcally sgnfcantly more lkely to have a longer duraton farm ob. Most of the sgns of these coeffcents are reasonable, ecept for the avalablty of free housng offered by the employer, whch we epected to have a postve effect on duraton. Ths may be because workers offered free housng are often mgratory, seasonal workers wth low skll and whose length of contract s generally short. An nterestng result s for Englsh speakng ablty. For unauthorzed workers, hgher Englsh speakng ablty s more lkely to lengthen the duraton n farm work. However, Englsh speakng ablty tends to shorten the duraton n farm work for other status workers. That s, legal workers leave agrcultural work earler as ther Englsh speakng ablty mproves, all else beng the same. Demographc varables tend to have varous drectons of nfluence on farm work duraton for each legal status. Beng female has a sgnfcantly negatve effect on duraton for authorzed workers, whle t has no sgnfcant effect for workers n any other legal status. Marrage has a sgnfcantly postve effect on duraton for authorzed and permanent resdent workers, whle t has a sgnfcantly negatve effect for unauthorzed and ctzen workers. Permanent resdent Hspanc workers tend to have longer farm work duraton than non-hspanc workers, whle unauthorzed and authorzed Hspanc workers tend to have shorter farm work duraton than non-hspanc workers. Educaton has a sgnfcantly postve effect on the duraton for all legal status, and eperence has a sgnfcantly negatve effect on the duraton for ctzen workers, but a sgnfcantly postve effect for workers n any other status. Age has a sgnfcant nonlnear effect on duraton for all equatons. The effect s postve up to an age of 84 years for unauthorzed, up to 69 years for permanent resdents, and up to 105 years for ctzen workers. On the other hand, the effect s negatve through 42 years for authorzed workers. 14

17 Net, usng estmates of each equaton, we calculate the predcted duratons of farm work by legal status by averagng the predctons over all observatons for each equaton (Table 4). The results ndcate that the average predcted duraton for unauthorzed workers s not necessarly shorter than those for legal workers (authorzed, permanent resdent, or ctzen). Actually, ts average predcted duraton s the second longest, and longer than for permanent resdent and ctzen workers. Fnally, we mplement a set of smulatons to eamne how farm work duraton of a typcal unauthorzed worker would be epected to change wth a change n legal status. Ths approach solates the effect of legal status of the worker from dfferng characterstcs of workers by holdng the characterstcs constant across varyng legal status. In addton, we vary the tme perod ( , , and ), the locaton (Calforna, Florda, and the rest of the U.S.), and the task (harvest or pre-harvest). We f each contnuous varable at the mean of unauthorzed worker observatons, and f each remanng dscrete varable at the category wth the mamum number of observatons of unauthorzed workers. The profle of the typcal unauthorzed worker s llustrated n Table 5. The epected duratons for ths typcal unauthorzed worker are shown n Table 6 (and Append Fgure 1) usng the equaton estmates for each legal status, condtonally upon beng an unauthorzed worker. For 14 out of 18 of the smulatons, unauthorzed workers workng as authorzed workers would have longer epected ob duratons. The largest effects were for unauthorzed workers workng under a permanent resdent status all were non-negatve, varyng from zero to 19 percent. For 13 out of 18 of the smulatons, unauthorzed workers workng under a ctzen status would have shorter epected ob duratons. Combnng the authorzed, permanent resdent and ctzen categores nto a sngle legal category, unauthorzed workers workng under a legal status would have longer epected ob duratons for 14 of the 18 smulatons. Noteworthy, s 15

18 that all of the smulatons for the perod ndcated qute large postve effects on epected ob duraton for an unauthorzed worker workng under a legal status. A strkng result of the smulatons s that epected ob duraton s markedly longer for work n Florda than for ether Calforna or for the rest of the Unted States, regardless of legal status (Fgure 1). Moreover, workng under a legal status n Florda has a consderably larger effect on epected ob duraton than elsewhere (last column of Table 6, and Fgure 1 for ). Fgures 2 and 3 llustrate the effects of legal status by tme perod for Calforna and Florda. The relatve dfferences between the and were mnmal for both Calforna and Florda. Calforna unauthorzed workers would have been epected to have a margnally shorter ob duraton under a legal status than as unauthorzed workers. In both states, epected ob duraton would have been longer followng 2001, and the effect of workng under a legal status would have been stronger n Florda. Fgures 4 and 5 contrast pre-harvest and harvest workers after 2001 for Calforna and Florda. There s very lttle dfference n the epected ob duratons for pre-harvest compared to harvest workers n ether locaton. Our estmated effect of a change n legal status from unauthorzed to a legal status s roughly consstent wth Tran and Perloff s result. In our case, epected duraton s somewhat longer when workng under a legal status; they report that IRCA ncreased the long-run probablty that people granted amnesty stayed n agrculture. (p. 437) Hashda and Perloff s result s n the same drecton, but larger. Emerson and Napasntuwong s result smlarly suggested a longer duraton for authorzed rather than unauthorzed workers. Ther result referred to the number of years workng n U.S. agrculture, rather than ndvdual obs as the above three 16

19 analyses do. Ther model dd not drectly address the sample selecton ssue, as the other three analyses do. Concluson We have proposed and estmated a Heckman-type two stage model wth legal status as an ordered probt model n the frst stage and a duraton model n the second stage. Ths methodology ams at overcomng two sources of nconsstency of farm work duraton study focused on specalty crop agrculture: selecton bas and the censorng problem. Our frst methodology deals wth the former problem adequately, but t takes only a rudmentary measure on the second problem: we have used only completed spells. Our current estmaton result s based on ths method. The current estmaton has sgnfcant coeffcents on the selecton bas correcton term for all legal status equatons ecept for that of permanent resdent workers. That s, usng ordnary least squares would lead to nconsstent estmates n all equatons ecept for permanent resdent workers. Although most of the effects of swtchng from an unauthorzed status to a legal status result n an ncrease n epected ob duraton, there are some nstances where epected ob duraton s shorter under a legal status than an unauthorzed status. The most common occurrence of ths was an unauthorzed worker workng under a ctzen status. We demonstrate large, postve dfferences n epected ob duraton between Florda and other states. Epected ob duratons appear not only to have notably ncreased followng 2001, but the effect of havng a legal status s also more postve followng

20 Table 1. Orderd Probt Model for Legal Status for Foregn-Born Specalty Crop Farm Workers Parameter Margnal Effect Estmate Female (0.019) Marred (0.017) Englsh Speakng (0.010) Black (0.078) Whte (0.016) Hspanc (0.049) Age (0.004) Age ( ) Educaton (0.007) Educaton (0.0005) Eperence (0.003) Eperence ( ) Before (0.018) After (0.020) μ (0.093) μ (0.093) μ (0.096) 18

21 Table 2. Actual-Predcted Legal Status Table Predcted Legal Status Total Actual Legal Status % 10% 11% 0% 100% 1 43% 21% 36% 0% 100% 2 14% 15% 70% 1% 100% 3 8% 8% 63% 21% 100% Table 3. Duraton Model for Specalty Crop Farm Workers wth Each Legal Status Unauthorzed Authorzed Permanent Ctzen Resdent Constant (0.026) (0.037) (0.034) (0.038) λ (0.006) (0.008) (0.007) (0.005) Female (0.007) (0.007) (0.007) (0.007) Marred (0.006) (0.006) (0.006) (0.006) Englsh Speakng (0.003) (0.004) (0.004) (0.004) Hspanc (0.011) (0.012) (0.010) (0.012) Age (0.001) (0.001) (0.001) (0.001) Age ( ) ( ) ( ) ( ) Educaton (0.0009) (0.001) (0.0008) (0.001) Eperence (0.0004) (0.0007) (0.0005) (0.0005) Task (0.002) (0.002) (0.002) (0.002) Free Housng (0.007) (0.007) (0.006) (0.007) Calforna (0.005) (0.006) (0.005) (0.006) Florda (0.008) (0.008) (0.008) (0.009) After (0.006) (0.007) (0.006) (0.007) 19

22 Table 4. Average Predcted Duraton for Each Legal Status (Days) Unauthorzed 58.0 Authorzed 58.4 Permanent Resdent 57.2 Ctzen 56.1 Table 5. Profle of the Typcal Unauthorzed Worker Constant 1 Female 0 Marred 1 Englsh Speakng Hspanc 1 Black 0 Whte 0 Age Age Educaton Eperence Free Housng 0 20

23 Table 6. Smulated Changes n Job Duraton by Legal Status a Tme Perod Locaton Task Unauthorzed Calforna Harvest Florda Harvest Rest of US Harvest 42.1 Legal Status Legal b Authorzed Permanent Resdent (-8.1) (2.5) (4.6) (15.6) (7.8) (5.0) Ctzen 45.2 (-14.5) 77.8 (4.8) 40.7 (-3.3) Legal c 51.9 (-1.7) 82.6 (11.3) 44.6 (6.0) Calforna Harvest Florda Harvest Rest of US Harvest (-8.9) 78.5 (3.7) 45.9 (6.8) 54.2 (0.5) 85.9 (13.3) 44.2 (2.9) 45.0 (-16.5) 77.5 (2.3) 40.6 (-5.6) 51.9 (-3.8) 82.6 (9.0) 45.0 (4.6) Calforna Harvest Florda Harvest Rest of US Harvest (12.0) (27.4) 65.6 (31.3) 66.3 (5.8) (19.3) 54.1 (8.3) 56.6 (-9.6) 97.5 (10.8) 51.1 (2.2) 68.2 (8.8) (23.3) 49.6 (19.3) Calforna Pre-Harvest Florda Pre-Harvest Rest of US Pre-Harvest (-7.8) 73.5 (5.0) 43.0 (8.2) 50.8 (2.0) 80.5 (15.0) 41.5 (4.4) 40.6 (-18.5) 69.9 (-0.2) 36.6 (-7.9) 48.9 (-1.9) 77.8 (11.0) 42.0 (5.7) Calforna Pre-Harvest Florda Pre-Harvest Rest of US Pre-Harvest (-8.6) 74.4 (4.1) 43.5 (7.3) 50.9 (0.0) 80.6 (12.7) 41.5 (2.3) 40.4 (-20.5) 69.6 (-2.6) 36.5 (-10.1) 48.9 (-3.9) 77.8 (8.8) 42.4 (4.4) Calforna Pre-Harvest Florda Pre-Harvest Rest of US Pre-Harvest (12.4) (27.9) 62.1 (31.9) 62.2 (5.3) 98.6 (18.7) 50.8 (7.8) 50.9 (-13.9) 64.3 (8.8) (5.5) (23.2) (-2.7) (19.2) contnued 21

24 Table 6. Smulated Changes n Job Duraton by Legal Status, contnued Notes: a All other worker characterstcs are as n Table 5. b Numbers n parentheses are percentage changes from epected duraton n an unauthorzed status. c The legal category combnes the three prevous categores authorzed, permanent resdent, and ctzen. The calculaton s E[ln( t.1) authorzed]pr[ authorzed] + E[ln( t.2 ) perm res]pr[ perm res] + E[ln( t., 3 ) ctzen]pr[ ctzen] 1 Pr[ unauthorzed] 22

25 Days Rest of US Calforna Florda Unauthorzed Legal Fgure 1. Epected harvest worker ob duraton after Days Pre After 2001 Unauthorzed Legal Fgure 2. Epected harvest worker ob duraton: Calforna 23

26 Days Pre After 2001 Unauthorzed Legal Fgure 3. Epected harvest worker ob duraton: Florda Days Pre-Harvest Harvest Unauthorzed Legal Fgure 4. Epected farm worker ob duraton: Calforna after

27 Days Pre-Harvest Harvest Unauthorzed Legal Fgure 5. Epected farm worker ob duraton: Florda after

28 References Bourgugnon, F., M. Fourner, and M. Gurgand. Selecton Bas Correctons Based on the Multnomal Logt Model: Monte-Carlo Comparsons, DELTA Workng Papers , DELTA (Ecole normale supéreure) (September 2004). Carroll, Danel, R. M. Samardck, S. Bernard, S. Gabbard, and T. Hernandez. Fndngs from the Natonal Agrcultural Workers Survey (NAWS) : A Demographc and Employment Profle of Unted States Farm Workers. Research Rept. 9. U.S. Department of Labor, Offce of Asst. Secretary for Polcy, Washngton, D.C. March Emerson, R. D., and O. Napasntuwong. Foregn Workers n Southern Agrculture. Selected paper, Southern Agrcultural Economcs Assocaton Annual Meetng, Orlando, Florda (February 2002). Greene, W. H. Econometrc Analyss, 5 th Edton.Upper Saddle Rver, N.J.: Prentce Hall, Hashda, E., and J. M. Perloff. Duraton of Agrcultural Employment. Workng Paper No. 779, Department of Agrcultural and Resource Economcs (CUDARE), Unversty of Calforna, Berkeley (February 1996). Lee, L. F. Generalzed Econometrc Models wth Selectvty, Econometrca, 51(2) (March 1983): Offce of the Assstant Secretary for Polcy, U.S. Department of Labor. The Natonal Agrcultural Workers Survey: What s the Natonal Agrcultural Workers Survey (NAWS)? (May 2005). Avalable at Sarg, Yoav, James F. Thompson, and Galen K. Brown. (2000) Alternatves to Immgrant Labor? The Status of Frut and Vegetable Harvest mechanzaton n the Unted States. 26

29 Backgrounder. Center for Immgraton Studes, Washngton, D.C. Tran, L. H., and J. M. Perloff. Turnover n U.S. Agrcultural Labor Markets. Amercan Journal of Agrcultural Economcs. 84 (2) (May 2002): U.S. Department of Agrculture. (2004) 2002 Census of Agrculture. Vol. 1, Parts 9 and 51. NASS. Washngton, D.C. Wnters, Alan, Terre Walmsley, Zhen Kun Wang, and Roman Grynberg. Lberalsng Labour Moblty Under the GATS. Economc Paper 53. Commonwealth Secretarat, London

30 CA, <93, harvest FL, <93, harvest Rest of US, <93, harvest CA, 93-01, harvest FL, 93-01, harvest Rest of US, 93-01, harvest CA, >01, harvest FL, >01, harvest Rest of US, >01, harvest CA, <93, pre-harvest FL, <93, pre-harvest Rest of US, <93, pre-harvest CA, 93-01, pre-harvest FL, 93-01, pre-harvest Rest of US, 93-01, pre-harvest CA, >01, pre-harvest FL, >01, pre-harvest Rest of US, >01, pre-harvest Days Legal Unauthorzed Append Fgure 1. Farm worker ob duraton smulatons