Bruno Crépon, Marc Ferracci, Gregory Jolivet, and Gerard J. van den Berg, "Information Shocks and the Empirical Evaluation of Training Programs During Unemployment Spells", Journal of Applied Econometrics, Vol. 33, No. 4, 2018, pp. 594-616. The data used in this article are available for replications, but they cannot be posted here. We are delighted to guide any researcher who is interested in the data. In this file, we first describe the data used in our paper (DATA INFORMATION) and then we present the programs uploaded to the JAE site (STATA PROGRAMS). DATA INFORMATION The data that are needed to replicate our results consist of a single Stata data table: spell.dta. This table contains an administrative data set directly available from the French national public employment office "Pole Emploi". We now explain how to obtain the data from Pole Emploi. We then describe the content of the Stata data table. Access ------ French administrative registers are subject to French laws and regulations on data protection. Usage of the data used in our paper requires permission from the relevant authorities, in this case the statistical services of Pole Emploi. Pole Emploi makes the data available to researchers for replication purposes. In their request, researchers should make explicit reference to the replication of our article to ensure proper access to the data. Requests should be made to Pole Emploi, by contacting Stephane Ducatez: Stephane Ducatez Directeur des statistiques, des Etudes et de l'Evaluation Pole Emploi 1, avenue du Docteur-Gley 75020 Paris France E-mail: stephane.ducatez@pole-emploi.fr Researchers should ask for table spell.dta. Description -------------- The data in the table spell.dta come from the Fichier Historique Statistique, an exhaustive register of all unemployed spells recorded at the National Employment Agency, whether the individual receives unemployment benefits or not. We use data on all unemployment spells starting in 2003 or 2004. We follow these spells up to their end or to the 1st of January 2008, which is the date when the data was extracted (very few spells last until then). Variables --------- The variables in spell.dta are: ident individual identifier gbeg date when unemployment spell starts gend date when unemployment spell ends gprop date of notification gbegt date when training spell starts gendt date when training spell ends prop1-prop25 date of notification 1-25 action1-action25 date of action 1-25 sexe gender datnais date of birth nenf number of children nation nationality nivetude diploma gmot unemployment motive gcont XXX pas dans le prog prep gtime XXX pas dans le prog prep gqual qualification gmat married gtenu tenure gcom city identifier galoc unemployment benefit UPLOADED STATA PROGRAMS These programs are zipped in the file cfjv-programs.zip. They are all ASCII files in DOS format. Unix/Linux users should use "unzip -a". -- prepdata.do This program creates the main work table (table75.dta) from the source data (spell.dta) -- statdesc.do This program produces the descriptive statistics in Tables 1 and 2. -- estim_PminZY.do This is the main estimation program. It maximises the likelihood of the benchmark specification model (K=2) and produces the results for Table 3, Table 4 (last panel), Figure 1 and Table 5 -- estim_PminZY_treatflex.do This program estimates the model with time-varying effects of notification and produces the results in the first panel of Table 4 -- estim_PminZYcs.do This program estimates the model where durations are censored 30 days after notification and produces the results in the second panel of Table 4 -- results_la.do This program uses some results from estim_PminZY to make Figure 1 -- estim_APBminZY.do This program estimates the model with additional information shocks from section 4.3 and produces the results in Table 6 -- estim_ZY_treatflex.do This program estimates the model without notification shocks and produces the results in the left column of Table 7 -- estim_PZY_treatflex.do This program estimates the full model, including the effect of training programs, and produces the results in the right column of Table 7 -- expected_durations.do This program computes the counterfactual expected unemployment durations shown in Table 8, using the results from estim_PminZY.do