Aico van Vuuren, "Using a Structural-form Model to Analyze the Impact of Homeownership on Unemployment Duration", Journal of Applied Econometrics, Vol. 32, No. 4, 2017, pp. 858-876. All files are zipped in the file av-files.zip. Because the programs are intended to be used on a Unix system, all ASCII files are in Unix format. Windows users have to use "unzip -a" or the equivalent. IMPORTANT: all .txt input files have as a first line the number of rows and columns in the data (matrix) of the file. When referring to row numbers, I refer to the rows starting from this row which is the line number of the document minus one. 1. DATA AVAILABILITY The data used in this paper are the property of the Dutch statistical office (CBS) who do not permit open access. Hence the data cannot be lodged here. Permission to use this data source is open for academic researchers and institutions. Applications should be directed to CBS. More information can be found at https://www.cbs.nl/en-gb/our-services/customised-services-microdata/conducting-your-own-research Questions concerning the data (access) can be asked to microdata@cbs.nl. A description of the data (in Dutch) can be found at https://www.cbs.nl/nl-nl/onze-diensten/methoden/onderzoeksomschrijvingen/korte-onderzoeksbeschrijvingen/inkomenspanelonderzoek--ipo-- As stated in the paper, I use the years 1989 to 2000. Details about the dataset can be found in the Web Appendix attached to the paper, which is also available on https://docs.google.com/viewer?a=v&pid=sites&srcid=ZGVmYXVsdGRvbWFpbnxhaWNvdmFudnV1cmVuc2hvbWVwYWdlfGd4OjNhYTdjNzQyMGE0NTU2MmM 2. SOURCE CODE FOR INDIRECT INFERENCE I provide c++-files for the indirect inference method described in the paper. Also the makefile is provided as well the file program.txt that explains execution of the file. The most important cpp-files are the following ones: cpb.cpp - Main file, handling the arguments and starting the indirect estimation process (using either simulated annealing or BFGS). IndirectEstimation.cpp - Defines the criterion function for indirect inference estimation. This function does the sampling, estimation of the sampled data and calculates the criterion function using the penalty function described in the paper. hazards_function.cpp - Estimation of the reduced-form hazards-rate model; see the paper for details. sample.cpp - Sampling class to do the sampling for indirect estimation. All other files are help files: probit.cpp - c++ implementation of probit. logit.cpp - c++ implemenation of logit. maximize(1).cpp - BFGS for c++ (1 is slight adaption) maxsan.cpp - simulated annealing retval.cpp - help file for maximize.cpp and maxsan.cpp NumJacobain - calculation of a jacobian RuntimeException.cpp - error handling BisectionSolver - bisection solver for c++ load/savemat - IO libraries for files. NormalDistribution - library for normal distribution function param_start.txt - starting values of the parameters: rows 1-11 Beta_lambda, rows 12-22 Beta_delta, row 23 constant of Beta_Lambda, rows 24 constant of Beta_Delta, row 25 log sigma_lambda, row 26 log sigma_delta, row 27 constant beta_eta, rows 28-38 Beta_eta, row 39 B, row 40 inverse logistic of P, row 41 log K, row 42 inverse logistic of (Rho + 1)/2, row 43 constant of Beta, row 44-54 Beta, row 55 log Sigma, row 56 log Tau. Please consult Table 1 for explanation of notation. After unzipping av-files.zip, and assuming that all files are in the same directory, the files can be compiled and linked by simply typing make (assuming that the standard g++ compiler is installed and working properly -- also OpenMpi or similar must be installed under /usr/local/include -- Windows users must change this). 3. INPUT FILES FOR SOURCE CODE - INDIRECT INFERENCE The paper is based on indirect inference based on the results of a mixed-proportional-hazards-rate reduced form model. See the paper for further details and also the paper by Van Vuuren (2014), which can be found at https://docs.google.com/viewer?a=v&pid=sites&srcid=ZGVmYXVsdGRvbWFpbnxhaWNvdmFudnV1cmVuc2hvbWVwYWdlfGd4OjZhZTM2MDhjMTZhNGViNWQ The results of these estimates can be downloaded from the file vp_theta.txt. It contains the following information (see also the results of the paper) Rows 1-3: impact of homeownership on hazard rate (local/non-local/employment) Rows 4-15: results of the probit regression of homeownership. These are coefficients of - Constant - Gender - Between 30 and 40 years of age - Between 40 and 55 years of age - Older than 55 - Presence of children in the household - Private sector - Civil Servant - Single - Head of household - Higher education grant Rows 16-27 : results for the hazard rate of leaving unemployment through local labor market. Coefficients are as above. Rows 28-39 : results for the hazard rate of leaving unemployment through non-local labor market. Coefficients are as above. Rows 40-51 : results for the hazard rate of leaving employment. Coefficients are as above. Rows 52-56 : results for the baseline hazard - homeowners - local labor market. Rows 57-61 : results for the baseline hazard - homeowners - non-local labor market. Rows 62-66 : results for the baseline hazard - employment. Rows 67-71 : results for the baseline hazard - renters - local labor market. Rows 72-76 : results for the baseline hazard - renters - non-local labor market. Rows 77-80 : mass points for unobserved heterogeneity distribution - local/non-local/employment. Rows 81-95 : probabilities for unobserved heterogeneity distribution. The results for the distribution regressions can be found in the files bx_dist.txt for the x-quintile of the distribution of the wages. Every file has 12 rows with coefficients of regressors as described above. The estimation of the indirect inference method also needs the inverse covariances of the first-stage reduced form models. These can be found in the files mhess.txt and mhess1_x.txt for the x-quintile of the distribution of the wages. These files contain the same number of rows (and columns) as the number of rows of respectively vp_theta.txt and bx_dist.txt The data on the x-variables can be found in the file g.txt. See the original document, i.e. footnote 6, for the way in which it is constructed. The number of observations of this file equals the actual number of observations (140082) and every row contains dummy variables of: 1) gender 2) between 30 and 40 years of age 3) between 40 and 55 years of age 4) older than 55 5) presence of children in the household 6) private sector 7) civil servant 8) self employed 9) single 10) head of household 11) higher education grant Aico van Vuuren University of Gothenburg aico.van.vuuren [AT] economics.gu.se