Argentina’s unemployment insurance system provides monthly benefits to workers laid off from private formal jobs who made a minimum number of contributions.
Herrera, Gonzalez-Rozada, and Ruffo (2026) study a marginal increase in the level of these benefits. The paper sets out, within a single framework, the alternative ways the literature accounts for what an extra peso of UI actually costs the government, and then evaluates all of them on the same Argentine data before combining them with estimates of the consumption drop at job loss to estimate the MVPF.
The paper uses administrative records covering all UI beneficiaries from 2005 through 2007, with estimation focused on the sample of workers displaced with more than 12 contributions in the prior 36 months and with reference wages more than 5 percent above the bottom kink. Identification comes from a regression kink design. The paper’s estimates focus on the high kinks, which lie close to the median of the reference-wage distribution and therefore provide a good approximation of the effects for the median beneficiary.
The paper finds that raising the initial benefit by ARS 40 (10 percent) extends benefit duration by about one day (equivalent to an elasticity of 0.05, well below the 0.3 median for developed countries reported by Schmieder and von Wachter (2016)). The elasticity of non-employment duration, censored at 24 months, is 0.22, against roughly 0.57 in developed countries in the same survey. Each additional peso of initial benefit raises total UI paid over the spell by ARS 5.92 for an elasticity of 0.99, meaning almost the whole of an extra peso is a direct transfer with little added distortion. Re-employment wages rise by ARS 0.28 per additional peso of initial benefit (an elasticity of 0.08). The paper attributes the muted behavioral responses partly to how small Argentine benefits are, and notes that high informality is part of the picture: 48.6 percent of UI recipients simultaneously hold an informal or self-employed job, earning on average ARS 748 (which is nearly double the transfer itself).
MVPF = 1.4
The net cost is estimated as the mechanical cost of paying one extra peso of benefit to inframarginal beneficiaries plus three potential fiscal externalities:
The paper uses the following calibrated statistics: mean benefit duration 7.48 months, mean non-employment duration 19.26 months, assumed employment duration 45 months (roughly the observed pre-unemployment tenure), monthly wage ARS 1,200, initial benefit ARS 400, total UI paid per spell ARS 2,586, and a labor tax wedge of τ = 0.40, set to roughly balance the government’s budget and in line with the observed gap between labor costs and net wages in Argentina.
The paper notes that the same government budget can be written in four ways, each using a different set of the estimates above, and it evaluates all four side by side rather than choosing one:
(A) works with average durations and the linear RKD elasticities. (B) rewrites the same budget in terms of transition rates, taking the responses of the UI exit rate and the job-finding rate from proportional-hazard models. (C) imposes an actuarially fair budget in which total time is fixed, so a longer non-employment spell mechanically eats into time spent employed and the non-employment response matters even without other government spending. (D) replaces the separate duration responses with the single elasticity of total UI paid, 0.99, which bundles the mechanical peso and the behavioral response together.
The willingness to pay for $1 of additional UI benefits is estimated as
where γ is the coefficient of relative risk aversion, set to 2, and Δc/c is the proportional gap in consumption between the employed and UI beneficiaries.
The paper’s preferred input for Δc/c is the gap for household heads receiving UI: per equivalent adult, non-durable household consumption is 14.8 percent lower for a household head collecting UI or other public transfers than for a formally employed head, from the ENGHO 2004–05 cross-section with demographic controls. This yields a willingness to pay estimate of 1.296.
The paper measures the drop at job loss several ways.
Dividing a willingness to pay of 1.296 by each of the four net cost estimates gives the paper’s preferred estimates:
The average of these four is 1.38.
The paper also reports alternative MVPF estimates built from alternative cost estimates (including or excluding the fiscal and wage externalities) paired with alternative measures of willingness to pay (changes in household income after separation from a formal job, household income losses counting only households with a worker receiving UI, and consumption losses after separation). These alternative MVPF estimates are summarized in Table 9 (the last row of which documents the MVPFs noted above):

Herrera, Joaquín, Martin Gonzalez-Rozada, and Hernán Ruffo (2026). “The Fiscal Costs of Unemployment Insurance.” Economía LACEA Journal, 25(1): 15–39. https://doi.org/10.31389/eco.533
Schmieder, Johannes F., and Till von Wachter (2016). “The Effects of Unemployment Insurance Benefits: New Evidence and Interpretation.” Annual Review of Economics, 8: 547–581. https://doi.org/10.1146/annurev-economics-080614-115758