Andrea Ichino, Fabrizia Mealli, and Tommaso Nannicini, "From Temporary Help Jobs to Permanent Employment: What Can We Learn from Matching Estimators and their Sensitivity?", Journal of Applied Econometrics, Vol. 23, No. 3, 2008, pp. 305-327. The file tempworkdata.zip contains the data sets used in the analysis, while the file tempworkcodes.zip contains the Stata codes that replicate the results. All zipped ASCII files are in DOS format. ----- DATA DESCRIPTION Data collection was directly realized by the authors as part of a large-scale evaluation project on temporary employment financed by the Italian Ministry of Welfare and the Tuscany Region. The data were collected through Computer Aided Telephone Interviews (C.A.T.I.). See the article (Section 2) for a precise description of the data collection strategy. The units of observations are the universe of temporary agency workers in 9 Italian provinces (treated units) and a random sample from the labor force of the same provinces (comparison units). The final sample contains 2,030 units: 511 treated and 1,519 comparison units. See the article (Section 2) for a description of the data-cleaning steps that lead to the final sample. The file tempwork.txt contains the data in ASCII format. The size of the file is 613 KB. Columns are divided by ",". The file tempwork.dta contains the data in Stata format (with appropriate labels). The size of the file is 829 KB. The Stata files mattosc.dta and matsici.dta identify the comparison units used as matches of the treated units by the nearest-neighbor propensity score matching algorithm. They are created from the original data set by the Stata routine program "attnd" (see Becker and Ichino, 2002). ----- LIST OF VARIABLES Note: "time 1" refers to the pre-treatment period, "time 2" to the treatment period, and "time 3" to the post-treatment period. age2 age at time 2 alwstud always student from time 1 to time 3 atyp1 atypical contract at time 1 blu1 blue-collar at time 1 cap recidency ZIP code at time 1 child number of children at time 1 cond1 employment status at time 1 cond2 employment status at time 2 cond3 employment status at time 3 contr1 contract type at time 1 contr2 contract type at time 2 contr3 contract type at time 3 ct dummy Catania dhire directly hired by previous firm dist distance from nearest agency edu education level emp1 employed at time 1 fblu father blue-collar fcond1 father employment status at time 1 femp1 father employed at time 1 fjob prevailing job of the father fledu father edu lower than high school fself father self employed fysch father's years of schooling go contacted a temp agency gr dummy Grosseto hour1 weekly hours of work at 1 hour2 weekly hours of work at 2 hour3 weekly hours of work at 3 id identification code italy italian nationality job1 type of job at time 1 know knew about existence temp agencies li dummy Livorno loc0 dummy macro-region at 0 - birth loc1 dummy macro-region at time 1 loc2 dummy macro-region at time 2 loc3 dummy macro-region at time 3 lu dummy Lucca male male gender manp 1 if comes from manpower sample manuf1 manufactury sectors at time 1 me dummy Messina mis_dur duration of temp contract ms dummy Massa-Carrara nofl1 out of labor force at time 1 nyu1 fraction of school-to-work without employment out1 permanent contract at time 3 out2 permanent or fixed-term contract at time 3 out3 employed at time 3 pa dummy Palermo perm1 permanent contract at time 1 pi dummy Pisa protreat dummy for province with agency at time 1 prov1 province of residency at time 1 pvoto mark in last degree as fraction of max mark reg0 region at period 0 - birth reg1 region at period 1 reg2 region at period 2 reg3 region at period 3 sect1 firm's sector at time 1 sect2 firm's sector at time 2 sect3 firm's sector at time 3 self1 self employed at time 1 serv1 public or service sectors at time 1 sici dummy Sicily single non married tosc dummy Tuscany tp dummy Trapani train1 received professional training before treatment treat temp at time 2 - treatment statuts unemp1 unemployed at time 1 wage1 monthly wage at time 1 wage2 monthly wage at time 2 wage3 monthly wage at time 3 whall weights with nostud nofl workers only manp temps whall2 weights with nostud nofl workers and all temps ysch years of schooling yu1 years without employment at time 1 white1 white collar at 1 other1 other sectors at time 1 emp2 employed at 2 unemp2 unemployed at 2 manuf2 manufactury sectors at time 2 serv2 public or service sectors at time 2 other2 other sectors at time 2 emp3 employed at 3 unemp3 unemployed at 3 nofl3 out of labor force at 3 perm3 permanent contract at time 3 atyp3 atypical contract at time 3 manuf3 manufactury sectors at time 3 serv3 public or service sectors at time 3 other3 other sectors at time 3 wage1b hourly wage at time 1 wage2b hourly wage at time 2 wage3b hourly wage at time 3 dloc01 loc0==nord dloc02 loc0==centro dloc03 loc0==sud dloc04 loc0==estero lic dummy for high-school education uni dummy for college education fedu father edu greater or equal high school ----- HOW TO REPLICATE THE RESULTS The Stata do-file des_base.do replicates the descriptive statistics and the baseline estimates presented in the article (Table 1 and Table 2). It makes use of the data sets tempwork.dta, mattosc.dta, and matsici.dta. The Stata do-file sens_calibrated.do replicates the sensitivity analysis with calibrated confounders presented in the article (Table 3 and Table 5). It makes use of the data set tempwork.dta described above. It also uses the Stata routine program "sensatt" to perform the sensitivity analysis (see Nannicini, 2007). The Stata do-file sens_killer.do replicates the sensitivity analysis with killer confounders presented in the article (Table 4 and Table 6). It makes use of the data set tempwork.dta described above. It also uses the Stata routine program "sensatt" to perform the sensitivity analysis (see Nannicini, 2007). Please note that the sensitivity results presented in the article are based on simulations. Even though point estimates, which are averaged across iterations, are very stable for a sufficiently large number of replications (e.g., 1,000), they are never exactly equal. ----- REFERENCES Sascha Becker and Andrea Ichino, 2002, "Estimation of average treatment effects based on Propensity Scores", The Stata Journal, Vol.2, 4, 358-377. Tommaso Nannicini, 2007, "Simulation-Based Sensitivity Analysis for Matching Estimators", The Stata Journal, Vol.7, forthcoming.