Unemployment Insurance Duration Extensions in Poland

The net cost is expressed per unit of mechanical transfer: 1 for the extra benefits paid to claimants who would otherwise have exhausted their benefits, plus the behavioral cost per unit of that transfer. For a marginal extension, the mechanical increase in benefit receipt,  dD_b/dP|_M, is the share of recipients who exhaust their benefits: 0.62 months per month of extra potential duration.

The behavioral cost has two channels: additional benefit payments, as more workers separate and claimants stay on benefits slightly longer; and lost tax revenue and extra social assistance, as higher separations and longer spells shrink employment. The paper writes the ratio of behavioral to mechanical cost as

\frac{BC}{MC} = \frac{1}{\frac{dD_b}{dP}\bigr|_M}\left(\eta_{\delta,P}\frac{D_b}{P}e + \frac{dD_b}{dP}\Bigr|_B – \frac{dD}{dP}u_b + e\frac{\tau}{b}\left(\eta_{\delta,P}\frac{D}{P} + \frac{dD}{dP}\right)\right)

where P = 6 months; η_δ,P = 0.119 is the elasticity of separations with respect to potential duration, from the inflow estimate; dD/dP = 0.41 is the additional months of unemployment per month of potential duration; dD_b/dP|_B = 0.02 is the behavioral increase in benefit receipt, in months per month; e = 0.9 and u_b = 0.05 are the shares of the workforce that are employed and that receive benefits; D = 11.9 months is the average unemployment duration; and D_b/P = 0.92 relates the average benefit duration to the potential duration. The third term is negative: longer spells leave fewer employed workers to separate and start new claims.

The ratio τ/b compares what the government loses when a worker moves into unemployment with the UI premium over social assistance. Here τ is the worker’s tax payments plus the social assistance the government saves while that person works, and b is the UI benefit over and above social assistance. From tax and replacement-rate statistics for an average-wage worker (an average tax rate of 35 percent, and replacement rates of 41 percent with UI and social assistance and 21 percent with social assistance alone), the paper obtains τ/b = 3.72, against 1.31 if social assistance were ignored.

The paper calculates a behavioral cost of 3.64 per unit of mechanical transfer, almost all of it lost tax revenue (about 3.48) rather than additional benefit payments (about 0.15). The net cost is therefore 1 + 3.64 = 4.64.

Dividing willingness to pay by net cost gives an MVPF of 1.2346 / 4.64 = 0.27. The paper reports MVPFs for every combination of its three risk-aversion coefficients and two consumption losses without designating a headline; this figure uses its model with endogenous separations, a coefficient of 2, and the benchmark 10 percent loss. With a 10 percent loss, the MVPF is 0.24, 0.27, and 0.37 for coefficients of 1, 2, and 5; with a 30 percent loss, it is 0.31, 0.44, and 1.28.

Holding inflows into unemployment fixed, as in Schmieder and von Wachter (2016), the behavioral cost is 2.48 and the MVPF is 0.32, 0.36, and 0.49 with a 10 percent loss, and 0.41, 0.59, and 1.71 with a 30 percent loss. Counting the separations of employed workers raises the behavioral cost by almost half and lowers every MVPF by about a quarter.

The extension would be welfare-neutral only if exhaustees consumed 53 percent less than employed workers with a risk-aversion coefficient of 2, or 26 percent less with a coefficient of 5. The paper concludes that, unless workers are highly risk averse, shorter benefit durations would raise welfare in Poland, but that extending duration is a less costly way to make UI more generous than raising the benefit level, because it draws far fewer employed workers into unemployment. Under the same assumptions, the paper’s MVPFs for a benefit-level increase are 0.08 for claimants with six months of potential duration and 0.11 for claimants with 12 months.