Labor Demand: Lecture 8. Empirical Evidence of Effects of Immigration, continued

Transcription

1 abor Demand: eture 8 Empirial Evidene of Effets of Immigration, ontinued Today we pik up where we left off on our disussion of the effets of immigration on labor supply and wages of natives. Reent immigration in the U.S. is onentrated in ertain oupations, partiularly low skilled oupations and agriultural work. lease see Table 4.: erentage Distribution of Immigrants and atives by Eduational Attainment, United States and California. In Smith, James., and Barry Edmonston, eds. The ew Amerians: Eonomi, Demographi, and Fisal Effets of Immigration. anell on the Demographi and Eonomi Impats of Immigration, ational Researh Counil, 997. Available online at: Consider the model, where losed eonomy produes one output: hilip Oreopoulos abor Eonomis otes for 4.66 Fall eture 8 Y = F K, ) (

2 = ( e ) is the elastiity of substitution between oupation groups. e is a ity-oupation augmentation fator. Sine wage is equal to marginal value produt in equilibrium, w = q F e rearranging: log = θ + ( )loge This is not a proper labor demand funtion, beause we have not solved for expressed employment as a funtion of ity effets, ity/oupation effets, and wages. F. evertheless, we ve Define total number of workers in ity as Solve for log : and add and subtrat = ( θ log ( ) ) + log e log( ) Define the ity-skill augmentation fator as: log e = e + e + e = u + u + d log f + u, where f is the fration of workers in ity in oupation (or skill group). hilip Oreopoulos abor Eonomis otes for 4.66 Fall eture 8 2

3 For employment, define: the number of people in ity, oupation (labor fore) the number of people in ity abor supply depends on wages: log( ) = ε, So ε is the elastiity of labor supply (in our stati model). Subbing in: log( / ε ) = ( θ log ) + ( )loge log( ), or ε + log( ε ) = v + v log f ' ε + + v = ( θ log ) + ( )loge log( ) ε + = u + u log f ' + u ε + ote, we have not mentioned immigration in the model yet. But onsider two skill groups: skilled and unskilled. Suppose immigration leads to an inrease in the population of unskilled labor. Impliitly, this assumes immigrants and unskilled natives are perfet substitutes. To the extent that this is not true, the implied inrease in the population of unskilled labor is less, and the implied effet is less. The model predits a derease in wages if the elastiity of labor supply or the elastiity of substitution are not infinite (similar to the Johnson model). Ignoring demand (as this model does), an inrease in hilip Oreopoulos abor Eonomis otes for 4.66 Fall eture 8 3

4 the unskilled labor allows firms to lower the wage and still hire the same number to produe the same output. How muh of a derease depends on how sensitive the urrent native workers are to that hange. If unskilled labor is a substitute for other inputs in prodution, firms an take advantage of lower wages for unskilled workers also by substituting the use of other inputs and using more unskilled labor. The more unskilled labor is a substitute for other inputs, the more unskilled workers demanded, putting pressure to drop the wage by less. ative unskilled labor employment only hanges from immigration to the extent that immigration affets wages. A number of papers define simply in terms of 2 groups: skilled and unskilled. The main empirial approah therefore is to regress wages and unemployment levels of unskilled natives at the ity level on the ity immigration level: log y = X β + f δ + e (ross setion) n n n (note: most studies swith to using f instead of logf to ease interpretation). We an t use ity fixed effets beause f is measured aross only one or two skill groups. Even if f is the fration of unskilled immigrants, we an t inlude the skilled worker observations bease, = f unskilled f, skilled log y = X β + f δ + e (first differene) n n n e.g. Altoni and Card (99), Buther and Card (99), aonde and Topel (99), Shoeni (997). The main advantage with the first differene analysis is that it eliminates bias from not inluding time invariant ity speifi fixed effets. Transitory effets (assoiated with transitory flutuation in the hilip Oreopoulos abor Eonomis otes for 4.66 Fall eture 8 4

5 demand for the output of speifi ities remain). Altoni and Card suggest instrumenting with the initial period level of immigration. ote, if short and long run differenes exist, the first differene approah measures more short run effets. <Altoni and Card, Table 7> The general onlusion with this approah has been that immigration has had little or no signifiant negative effet on native wages and employment. Main onerns with this approah: ) the possibility that mobility offsets immigrant inflows 2) the possibility of remaining downward bias from loal demand shoks Card, defines skill groups as oupations. This allows him to inlude ity fixed effets in the ross setional regressions. Also, grouping by oupation allows for more speifi labor supply inrease, and may allow more preision in estimation, and more useful variation. log w = u + u + d ( ) + u log( ) = v + v + d2 ( ) + v Data: 990 U.S. Census 6-68 year olds, no students 75 largest MSA s 6 oupation groups hilip Oreopoulos abor Eonomis otes for 4.66 Fall eture 8 5

6 ( = 75 x 6) Question: how to measure? City-oupation speifi labor fore. Approah: At the ountry level, for the sample working, estimate r( o = ) = X β + e i i i (multi-nomial logit). X inludes rae, eduation, age, marital status, ethniity, and ity. Use predition equation to estimate probability of being in oupation for entire sample (even those not working). is the sum of the probabilities: = πˆ i i Cities with more immigrants, individuals with less eduation, will have higher for low skill oupations. Finally, Card instruments the ity oupation labor supply,, using the predited hange due to predited inrease in immigrant labor supply: S = M g g λ g τ g M g is the ountry-wide total immigrant inflow of ethni group g, between 85 and 90. λ g is the fration of pre 85 immigrants of ethni group in ity τ g is the ountry wide fration of immigrants in ethni group g in oupation hilip Oreopoulos abor Eonomis otes for 4.66 Fall eture 8 6

7 City speifi variation in S arises only from a variable that is measured before 85. Thus, S is independent of any ity speifi demand shok between 85 and 90 that would have led to differenes in immigration omposition. Table 6 and 7 show results with and without IV ote, first stage not shown, but T-stat from regressing log on S is about 5. Even though the analysis is at the oupation/ity level, the ATE is the effet of wages and employment from an inrease in the ity-oupation labor fore due to immigration between 85 and 90. Approah generates more preise results, but onlusion that immigration effets on labor market are small remains. Conern about mobility remains. Card tries to address this by atually measuring population inflows and outflows of ities, in response to ity speifi immigration. The ensus provides information on loation 5 years ago. Can use to alulate ity-oupation population inflows and outflows. 90, 85, = Z β + γr + d + θ + e Figure and Table 4. y is omponent of population growth for oupation group in ity Boras: The demand urve is downward sloping hilip Oreopoulos abor Eonomis otes for 4.66 Fall eture 8 7

8 Boras takes the mobility onern seriously. He argues strong fores exist to equalize eonomi onditions aross ities. (ote, see Shoeni who takes Altoni and Card and adusts for ost of living by ity). Boras approah is analogous to Card s, but uses variation by experiene and eduation group rather than ity, to measure impats of immigration at the national level. Output is time speifi, rather than ity speifi: Y = F K, ) t ( t t t = ( est s E E st ) E E, like before. st is now the number of workers with eduation s at time t, and E is the eslatiity of substitution aross these eduation groups. Boras normalizes the skill augmentation fators to sum to one: ow he goes one step further to allow for workers with same eduation but different experiene not to be perfet substitutes: X X X X st = ( α xs ), x are the number of workers in eduation group s and experiene group x at time t, and X is the elastiity of substitution aross experiene lasses, within an eduation group. ike before, set the M = w to get the wage equation hilip Oreopoulos abor Eonomis otes for 4.66 Fall eture 8 8

9 w = F t t e st E st E α xs X (*) = ut + ust + uxt log + u X Boras main analysis estimates a slightly different model: M it = et + est + ext + β log + e M + ' it Whih is similar to what Card estimates. To get bak to the elastiity, define M it = vt + vst + vxt + β + v, where ' m tsx M it = is the perentage inrease in the group s labor supply due to immigration. ' Be areful to note whether elastiites are being measured (not here) or whether the studies are omparable. Sine d M it d log( M + ' it d = mit ) d log( + m d = ) dm ( + m ) 2, So we an divide Boras oeffiient by 2 ( + m ) to get d dm. hilip Oreopoulos abor Eonomis otes for 4.66 Fall eture 8 9

10 Boras uses m =.68, the fration of immigrants in the labor fore in 2000, for the omparison. ote, omparison between Card and Boras is made diffiult by Boras using eduation-experiene shares, and Card using oupation shares. Table 3 Table 4 Table 5: Main point, estimates muh different whether defining eonomies at the state level versus the national level. Implied effet from a 0 perent inrease in the flow if immigration on wages is.3 perent fall when a state s geographi boundary is used, and a 4-6 perent fall when onsidering the national market. For a really good overview of the eonomis of immigration in the United States, you are reommended to take a look at The ew Amerias, ational Researh Counil. The book an be read on onsulted quikly online at: hilip Oreopoulos abor Eonomis otes for 4.66 Fall eture 8 0