Panel Paper:
The Link Between Welfare Programs and Informal Labor: Evidence From Argentina
*Names in bold indicate Presenter
This paper examines the effects of government lead formalization processes aimed at reducing labor informality on formalization rates, hours worked and wages of those affected by the policy change. Policy might have lower effects than expected given the existence of welfare programs that act as a fixed cost of operating formally. That means that targeted beneficiaries self-select into formal or informal sector which biases the true impact. Using the advents of Argentinean national Law 26.8444 of 2014 that regulates domestic workers (maids, nannies) and using having children under the age of 18 as an exogenous source of variation, this paper studies the impact of such processes through a diff-in-diff strategy. The strategy compares the mentioned outcomes between those eligible to receive the Universal Subsidy per Child program or AUH (biggest cash transfer in the Argentinean system which is dependent on the number of children under 18) and those non-eligible, before and after the policy change. The AUH is aimed at receiving a subsidy per child under 18 if the parent is unemployed, works in the informal sector or earns less than the monthly minimum wage.
The theoretical model developed in the paper predicts: a. the probability of having an “off the books” job is higher for those workers eligible to receive the AUH compared to the non-eligible ones; b. an ambiguous effect on equilibrium wages; and, c. conditional on being a formal worker, the hours of participation for those AUH eligible workers bunches at the number that gives them a monthly wage below the minimum wage cutoff.
The data for the estimations comes from the Argentinean Household Survey for the period January 2013-July 2015. Preliminary results show that the passage of the law reduces the probability that eligible workers will work on the formal sector, as predicted, and their equilibrium wages. The effect on number of hours cannot be determined since the parallel trend assumption, needed for the diff-in-diff strategy, doesn’t hold for that variable. The latter is probably due to the fact that number of hours usually contains measurement error when it is part of a household survey dataset. For example, for a particular job, income paid to a domestic worker might be split between “in the books” and “off the books” parts, and there could be bunching in the number of hours that are considered “in the books” (instead of total hours) - something that cannot be seen in the data.