Thesis: Social insurances are an important component of welfare states in developed countries.They have been set up to meet the twofold objective of improving efficiency, in acontext where information is not available to all parties, and enhancing equity bysharing the cost of insurance across risk types. If this general principle achievesconsensus, the question of the optimal level of insurance that should be providedis at the centre of the policy and scientific debate. This dissertation addresses theissue of the optimal design of unemployment insurance (UI), taking into account howUI parameters can affect the behaviour of firms and workers either before, duringor after unemployment. It contributes to the literature in three ways: (i) it drawsthe attention on the interaction between UI design and on-the-job behaviours wherethe focus has mainly been on the response of job-seekers to UI parameters duringunemployment; (ii) using rich administrative data, it empirically reveals the existenceand potential costs of such behavioural responses; (iii) it tests the empirical validityof theoretical arguments usually put forward as a motivation for public interventionin the UI market, namely adverse selection and individual optimisation failures.Chapter 1 assesses the effect of a UI program letting some job-seekers choosebetween low benefits for a long duration or higher benefits for a shorter duration. Iuse this uncommon choice feature in nationally-mandated UI schemes to understandthe determinants of the choice and its consequences. Using a rich set of covariates, Idocument the existence of adverse selection, and relate the choice of the high-benefitoption to observable characteristics generally associated to a lower risk-aversion or ahigher impatience. My results also reveal a high moral hazard cost, as job-seekersopting for higher benefits are predicted to stay unemployed longer. This negativeimpact that is even higher for job-seekers with a high initial unemployment risk.Chapter 2 shows that a discontinuous increase in the level of UI benefits at atenure threshold leads to the strategic scheduling of layoffs in order to maximise thesurplus from separation. I use the bunching methodology to quantify the extensionof the employment spell in response to the increase in UI benefits. Results suggestthat this extension is the result of an individual bargaining process between theworker and the employer. I argue that workers trade higher UI benefits against areduction in the cost or the risk associated to the layoff, through a lower probabilityto claim damages, lower severance payment or a lower reputational cost. Chapter 3 analyses the effect of the UI eligibility criteria imposing a minimumemployment record to claim benefits on pre and post-unemployment outcomes. Ithighlights a separation response through a jump in employment outflows at theeligibility threshold. Exploiting a reform that reduced the minimum employmentrecord, we show that, in some sectors, this separation response translates into ahigher number of contracts whose duration exactly coincides with the new workhistory condition. This suggests that, in some sectors, the change in the UI eligibilitycriteria contributed to shape a new norm regarding the duration of short contracts.It would imply that UI parameters do not only affect the outcomes of job-seekersbefore, during and after unemployment, but also the outcomes of workers who donot experience unemployment. The last part of the chapter evaluates the extensivemargin effect of UI benefits. Receiving UI benefits as opposed to not receiving anybenefit at all has a negative impact on future employment probability, that does notseem to be compensated by an improvement in terms of job quality.