Page 68 - IJPS-1-1

Undocumented migration in response to climate change

the date of the first move enabled us to construct an event-history file, indicating the
household migration status for each observational year during the study period of
1986–1999. This period was chosen as a time of relatively stable migration policies fol-
lowing the enactment of the Immigration Reform and Control Act (IRCA) in 1986
(LoBreglio, 2004) and because Mexico experienced conditions of increased temperature
and drought during the 1990s (Stahle, Cook and Villanueva Diaz et al., 2009) that resem-
ble conditions expected under climate change (Collins, Knutti, Arblaster et al., 2013;
Wehner, Easterling, Lawrimore et al., 2011). A reduction in the weather stations available
through GHCN after 1999 prevented the construction of the climate measures for later
years.

2.2 Outcome Variable
In the cultural context of Mexico, migration needs to be considered as a household-level
strategy (Cohen, 2004). A household sends a migrant to an international destination as a
self-insurance mechanism against local market failure, expecting the migrant to remit
money to support the household in Mexico (Massey, Arango, Hugo et al., 1993; Taylor,
1999). We therefore focused on the household as the unit of analysis, in line with prior
work (de Janvry, Sadoulet, Davis et al., 1997; Hunter, Murray and Riosmena, 2013;
Kanaiaupuni, 2000). We constructed an event history file (risk set) in which house-
hold-years are assigned a value of 0 when the household was at risk for international mi-
gration but no move occurred, a value of 1 if an undocumented international move oc-
curred, or a value of 2 if a documented international move occurred. Households were at
risk of migration if they did not send a member to the U.S. before 1986. Households were
included in the data set for the years after 1986 as long as the household heads were at
least 15 years of age, and after the date of their first union formation (household heads can
get divorced, widowed, and remarry in later years). These criteria ensured that households
were truly formed during the years when they were exposed to the risk of migration.
Households were removed from the data set following the year of the first move, when the
household head turns 65, when the household is censored at the survey year, or at the end
of the study period in 1999. Households may move in and out of the study community and
are only exposed to the risk of migration if at least one core household member (head or
spouse) was present during a given year.
Although other pathways are possible (Burke, Miguel, Satyanath et al., 2009; Naw-
rotzki, Diaconu and Pittman, 2009), we assumed that climatic effects lead to migration
through negative impacts on the agricultural sector (Mueller, Gray and Kosec, 2014). Ru-
ral households in Mexico heavily depend on agricultural production for income and sus-
tenance (Conde, Ferrer and Orozco, 2006; Wiggins, Keilbach, Preibisch et al., 2002;
Winters, Davis and Corral, 2002). As such, we focus our analysis on 68 municipalities that
contain rural MMP communities (population < 10,000) dispersed across the country. Fig-
ure 1 illustrated the location of the rural municipalities as well as the 214 weather stations
from which daily temperature and precipitation data were available.

2.3 Primary Predictors
Previous research has shown that temperature and precipitation above and below certain
thresholds have the strongest impact on agricultural production (Lobell, Hammer, McLean
et al., 2013; Schlenker and Roberts, 2009). As such, we employed two climate change in-
dices that reflect percentile-based threshold effects, namely the warm spell duration index
(WSDI) and precipitation during extremely wet days (R99PTOT). The warm spell dura-
tion index was computed as the annual count of days when at least six consecutive days of

International Journal of Population Studies | 2015, Volume 1, Issue 1 62