Arkadiusz Szydlowski, "Endogenous Censoring in the Mixed Proportional Hazard Model with an Application to Optimal Unemployment Insurance", Journal of Applied Econometrics, Vo. 34, No. 7, 2019, pp. 1086-1101. All files are ASCII files in DOS format. They are zipped in the file as-files.zip, which contains two folders. Unix/Linux users should use "unzip -a". The article uses six datasets: 1) SIPP1.csv - dataset used in column (1) of Table 3 in the text - variables (in order of appearance): unemployment duration (Y) censoring indicator (D, =0 if censored, =1 otherwise) log UI benefit log annual pre-unemployment wage average unemployment rate age dummy marital status dummy constant (=1 for all observations) - discretization: 4 x 2 x 2 x 2 x 2 2) SIPP2.csv - dataset used in column (2) of Table 3 in the text - variables (in order of appearance): unemployment duration (Y) censoring indicator (D, =0 if censored, =1 otherwise) log UI benefit log annual pre-unemployment wage average unemployment rate constant (=1 for all observations) - discretization: 6 x 2 x 3 3) SIPP1_min.csv - same as SIPP1.csv besides log UI benefit being discretized to minimum within the quantile - this dataset is used in Appendix C.4 4) SIPP2_min.csv - same as SIPP2.csv besides log UI benefit and average unemployment rate being discretized to minimum within the quantile - this dataset is used in Appendix C.4 5) SIPP1_max.csv - same as SIPP1.csv besides log UI benefit being discretized to maximum within the quantile - this dataset is used in Appendix C.4 6) SIPP2_max.csv - same as SIPP2.csv besides log UI benefit and average unemployment rate being discretized to maximum within the quantile - this dataset is used in Appendix C.4 These datasets come from Survey of Income and Program Participation. They are modified versions of the dataset used in Chetty (JPE, 2008). The main difference is that we erase second and further spells for those who entered unemployment multiple times. Each dataset contains 3986 observations. In order to obtain the main results in Table 3, one needs to use the KMS package available from https://molinari.economics.cornell.edu/programs.html We used the CVX package with SDPT3 solver and University of Leicester's ALICE2 computing cluster in order to obtain the results. The attached MATLAB program files provide components to be used within the KMS package: -> KMS_EndCensAppl.m - main file, calculates the confidence set -> KMSoptions.m - options file -> moments_w.m - part of the moment condition(s) that depends on the data -> moments_theta.m - part of the moment condition(s) that depends on the parameters (theta) -> moments_stdev.m - standard deviation of the moment condition(s) -> moments_gradient.m - gradient of the moment condition(s) wrt theta For different models in the article: 1. Weibull hazard (columns (1)-(2) in Table 3) 2. Weibull hazard with c-dependence (columns (3)-(4) in Table 3) 3. Piecewise constant hazard (column (5) in Table 3) 4. Piecewise constant hazard with c-dependence (columns (6)-(7) in Table 3) The data files need to be saved in the MATLAB format (.mat) before running the KMS package.