Corrigendum

We are grateful to Lahav Karady of the Hebrew University of Jerusulem for taking the time to replicate our event history analysis of the likelihood of taking a first documented trip to the United States, which appeared in the third and fourth columns of Table 2 in Massey, Durand, and Pren (2015). In his reanalysis Karady detected some anomalous data issues and brought them to our attention. Upon investigating we discovered that a programming error had led to the selective dropping of cases during a file merge. That mistake is here corrected in Table 1, which compares the original published model estimates to new estimates based on the full dataset.

Table 1.

Discrete Time Event History Analysis to Predict the Odds of Taking a First and an Additional Documented U.S. Trip in year=t+l from independent variables defined in year t: 1972 to 2011

		Corrected Model			Published model
		Departed on first trip with documents (year=t+l)			Departed on first trip with documents (year=t+l)			Difference Between Models
	Independent Variables	β		SE	β		SE	t-test
Demographic Background
	Age	0.0565	**	0.0170	0.0721	***	0.0175	-0.9114545
	Age-squared	-0.0010	***	0.0002	-0.0010	***	0.0002	0
	Female	0.0723		0.1330	0.7394	***	0.1362	-4.9880663	***
	Married	-0.3160	**	0.1132	-0.2380	**	0.1148	-0.6868035
	No. of minors in household	-0.0632	**	0.0301	-0.0805	**	0.0299	0.57563436
Human Capital
	Labor force experience	-0.0056		0.0084	-0.0148	*	0.0079	1.10957941
	Education	0.1065	***	0.0111	0.1283	***	0.0113	-1.9558235	*
	Agricultural job	---			---
	Unskilled job	-0.1243		0.0981	-0.2619	**	0.0978	1.40364192
	Skilled job	-0.2493	*	0.1351	0.0231		0.1358	-2.0138738	*
Social Capital
	Parent a U.S. Migrant	0.4326	***	0.1184	0.0256		0.1149	3.46113452	**
	No of U.S. migrant siblings	0.0662	**	0.0318	-0.0476		0.0312	3.594282	**
	Spouse a U.S. migrant	-0.624	**	0.2473	-0.7714	**	0.2418	0.59911515
	No. of U.S. migrant children	0.2885	***	0.0421	0.1569	***	0.0434	3.10375162	**
	No. of U.S. born children	-1.1886	**	0.3703	-1.8225	***	0.4666	1.61484547
	Prop U.S. Migrants in Community	0.0362	***	0.0033	0.0198	***	0.0035	4.91125306	***
Physical Capital
	Land	-0.1930		0.1648	-0.1750		0.1627	-0.1095458
	Home	-0.3153	**	0.1025	-0.2514	**	0.1043	-0.6208959
	Business	-0.2474	+	0.1517	-0.1985		0.1529	-0.3217588
Documentation
	Temporary Visa	5.9135	***	0.1232	4.6272	***	0.1229	10.446623	***
	Legalized under IRCA	-0.2856		0.2342	-0.6433	***	0.1699	1.63075168
U.S. Social Context
	Border Patrol Budget ($2013) [Divided by 1000]	-0.2605	+	0.1596	-0.2765	*	0.1628	0.09978845
	Rate of Employment Growth	0.0579	*	0.0334	0.0623	*	0.0338	-0.1313731
	Residence / Work Visas per Capita	0.0001		0.0002	0.0002		0.0003	-0.352601
Mexican Context
	Crude Birth Rate	-0.0261		0.0173	-0.0326	*	0.0176	0.37422349
	Rate of GDP Growth	-0.0274	**	0.0118	-0.0295	**	0.0119	0.17761839
	Homicide Rate	0.0029		0.0191	0.0053		0.0195	-0.1250495
Community size
	> 100,000	---			---
	10,000-99,999	-0.7756	***	0.1161	-1.0666	***	0.1183	2.49553633	*
	2501-9,999	-1.0444	***	0.1182	-1.3150	***	0.1215	2.27467021	*
	<=2500	-1.0072	***	0.1470	-1.3807	***	0.1472	2.54001804	*
Intercept		-6.8186	***	0.9094	-5.1997	***	0.9265	-1.7724857
Likelihood Ratio		1937.1301	***		1700.6557	***
Wald		3071.1172	***		2115.5368	***
Total number of person-years		641,587			193.012

⁺p=.10;

^*p<.10;

^**p<.05;

^***p<.001

Fortunately the correction does not alter our basic conclusion that the “circularity [of migration] has declined markedly for undocumented migrants but increased dramatically for documented migrants.” The re-estimation did nonetheless yield significant changes in the values of certain coefficients, as indicated by the t-tests reported in the last column of Table 1. Although little change occurred in estimates of the effects of physical capital or U.S. and Mexican contextual variables, the effects of demographic background, human capital, social capital, documentation and community size did register significant changes in how they predicted the likelihood of taking a first documented trip.

Even though changes coefficients associated with the dichotomous indicators of community size were statistically significant, however, they were not substantively important since the direction and significance of all effects was the same, just somewhat greater in magnitude in the corrected model. Among indicators of human capital, the negative coefficient for labor force experience became significant in the re-estimation; the significant negative effect of coming from a skilled occupational background turned positive and lost significance; the negative effect of coming from an unskilled background increased in size and became significant. Although the coefficient for education increased in size, it remained positive and significant in both models.

With respect to demographic characteristics, the positive effect of being female increased in size and became highly significant. Women are more likely than men to undertake a first trip with documents, in keeping with the fact that migration from Mexico is generally a male-led phenomenon in which men migrant first, manage to acquire documents, and then sponsor the entry of spouses and children (Massey and Cerrutti 2001). The fact that most legal migrants enter the United States via family reunification provisions within U.S. immigration law is indicated by the strong and significant positive coefficients for having a U.S, migrant parent, siblings, and children, effects that were not significant in the original model. Although the coefficient for the relative share of migrants in the community was significant and positive in the original model, its size increased significantly upon re-estimation.

Finally, although the effect of having access to a temporary work visa increased significantly in size in moving from the original to the corrected model, the effect was very strong and significant in both equations. As noted early, none of these changes affects our principal conclusion that the likelihood of return migration increased for documented migrants but decreased for undocumented migrants, it does change the profile of those likely to depart Mexico on a first documented trip in a more sensible direction. Those most likely to initiate legal migration tend to be relatively educated urban females living in households and communities with a high degree of migratory experience. The corrected estimates underscore more strongly the importance of family ties in the initiation of documented migration.

The erroneous file merge primarily affected the dataset used to estimate the model of first trip probabilities. Our re-analyses detected no significant shifts in estimated coefficients for the model predicting the likelihood of returning from a first U.S. tip and only one significant shift in each of the models predicting the likelihood of taking and returning from an additional U.S. trip. In the former, the coefficient for the homicide rate shifted from significantly negative to significantly positive, inducing that lethal violence promoted greater repeat migration to the United States (going from -0.108 to 0.0954, with the corrected estimate significant at p<0.001). In the latter model, the coefficient for months of U.S. experience fell from statistical significance to insignificance (going from 0.002 to 0.000). Since no other significant shifts were detected, the full tables are not reproduced here but will be sent upon request.