Replication Package for "Selection in Surveys: Using Randomized Incentives to Detect and Account for Nonresponse Bias"

Replication README. Selection in Surveys: Using Randomized Incentives to Detect and Account for Nonresponse Bias Authors: Deniz Dutz, Ingrid Huitfelt, Santiago Lacouture, Magne Mogstad, Alexander Torgovitsky & Winnie van Dijk Please read this document carefully. The folder contains all the replication files for the paper results in form of tables and figures. Most of the individual data cannot be shared publicly due to individual privacy. The estimation codes won't run for external users unless they apply for the data, see details below. However, most exhibits are reproducible, as we share most of the estimation results. Also, to decrease reproduction time, users can toggle on and off the reproduction of estimates (see analysis/Master.R) Instructions to Replicators Two software are used to produce results: Stata and R. All reproduction is called from analysis/Master.do and analysis/Master.R scripts, which further call several smaller scripts in analysis/code folder. Stata is mostly used to produce descriptive statistics and some figures. Most of the content up to section 4 is produced with Stata. R is mostly used for estimation. Most estimations in Section 4 and all output for sections 5 and 6 are produced using it. Therefore, if the user wants to run everything again, she must run analysis/Master.R first and then call analysis/Master.do. Finally, see analysis/Selection_in_Surveys _figures_tables.pdf in analysis folder for enumeration of Figures and Tables, referenced in the replication scripts. --------------------------------------------------------------------- Replication Package Structure This replication package is structured as follows: - /analysis/ o Contains the scripts that execute the reproduction of our analysis of the NiK Survey. o Codes/ ▪ Contains the a set of scripts that reproduce our analysis of the NiK Survey. o Output/ ▪ Contains estimation output used to reproduce exhibits in the paper - /CoffmanReplication/ o Contains the data used for our re-analysis of Coffman et al (2019). - /nik-data/ o Contains scripts and data shareable used to construct the main data analyzed in this paper - /sos-data/ o Contains raw and analysis data used for our review of survey use in economics. ------------------------------------------------------ Computational Requirements Data preparations, estimations and preparation of figures were done in Stata/MP (version 16.1) and R (version 4.1) accessed using RStudio (version 1.4.1103). The programs took approximately 4 hours to run. The data preparation were carried out at the Statistics Norway’s restricted-access data server, which is equipped with a high performance computer a 132T disk space and a 1.6T memory. The analysis was carried out in personal laptops, using Stata/MP (version 18.1), R (version 4.1) accessed using RStudio (version 1.4.1103) and calling Gurobi 9.1.2. ------------------------------------------------------ Data Availability Statement: restricted-access data In the following, we provide data availability statements for the restricted-access data sources used in this study. 1. Restricted-Access Data from Statistics Norway We used restricted-access micro datasets extracted from Statistics Norway's Population Statistics (Statistisk sentralbyrå, 2000a), Employment Statistics (Statistisk sentralbyrå, 2020a), and (Falnes-Dalheim and Krawczynska, 2020). All our analyses were carried out at Statistics Norway's restricted-access data server. Researchers can gain access to these datasets from Statistics Norway by submitting a written application to [email protected]. Applications must be certified by the Norwegian Data Inspectorate in order to ensure that data are processed in a manner that protects the personal integrity of individuals surveyed. Conditional on this approval, Statistics Norway determines data access. For more information on how to gain access to data from Statistics Norway, see description here: https://www.ssb.no/en/data-til-forskning/utlan-av-data-til-forskere. In the following, we describe the restricted-access micro datasets from Statistics Norway needed to replicate our analysis: A. National Education Database and Central Population Register - input/alle_pers_g2020m04d30.dta This extract contains population register data including personal identifiers (fnr, snr), household information (pers_i_hushnr), civil status (sivilstand), birth dates (foedselsdato), death dates (DOEDS_DATO), immigration category (invkat), and spouse identifiers (ektf_fnr) as of April 30, 2020. B. Employer-employee registry - input/ameld_statdata_YEAR_mM_lag2.dta This extract contains monthly wage and employment data from the A-melding register, including employment start/end dates (arb_start, arb_slutt), working hours (arb_arbeidstid), wages (lonn_ialt, lonn_kontant, lonn_natural, lonn_godtgjorelse), employment status variables, and firm identifiers. Available for years 2017-2020. C. Statistics Norway Survey ``Norway in Corona Time'' input/covid19_4.sas7bdat and input/covid19_5.sas7bdat These extracts contain the survey data as collected by Statistics Norway. 2. Restricted-Access Data from the Norwegian Labour and Welfare Administration (NAV) We used restricted-access micro datasets from NAV covering employment status data(Statistisk sentralbyrå, 2020b). A. SOFA/ARENA Employment Status Data - input/sofaYYMM.dta These extracts contain detailed labor market status information including unemployment categories (as_gr, as_f), benefit types (as_yte), and program participation (tiltak). Available monthly for 2019-2020. --------------------------------------------------------------------- Directory Structure and Dataset List in restricted-use server Our projectfolder/ on Statistics Norway's data server was organized as follows: projectfolder/code/ projectfolder/input/ projectfolder/output/ projectfolder/temp/ projectfolder/survey/dta/ projectfolder/survey/wk48/ The scripts collecting and preparing the data for analysis are all found in the “nik-data” folder. ------------------------------------------------------ Data Availability Statement: primary and secondary data shared in this replication package Raw data collected for Section 2 is shared to the public in this replication package. This data can be found in the “sos-data” folder. See Online Appendix of the article for details on its collection and homogenization. Two additional secondary datasets are included in this replication package: 1. Replication data for Coffman et al (2019), found in the “CoffmanReplication” folder. 2. Local government data in Norway from Fiva et al. (2020), found in the “nik-data” folder. -------------------------------------- References Coffman, L. C., J. J. Conlon, C. R. Featherstone, and J. B. Kessler (2019). Liquidity Affects Job Choice: Evidence from Teach for America. The Quarterly Journal of Economics 134 (4), 2203–2236 Falnes-Dalheim, E., & Krawczynska, M. A. (2020). Norge i koronatid (SSB Notater 2020/XX). Statistisk sentralbyrå. ISBN 978-82-537-XXXX-X. Fiva, J. H., A. H. Halse, and G. J. Natvik (2020). Local Government Dataset. www.jon.fiva.no/data.htm. Accessed: 2021-05-23 Statistisk sentralbyrå (2000): "Dokumentasjon av BESYS - befolkningsstatistikksystemet", Notater 2000/24. Statistisk sentralbyrå (2020a): "A-ordningen. Data om personers tilknytning til arbeidsmarkedet." https://www.ssb.no/data-til-forskning/utlan-av-data-tilforskere/variabellister/a-ordningen (accessed: April 30, 2020). Statistisk sentralbyrå (2020b): " Kodeliste for ARENA as_stat (arbeidssøkerstatus, aktivitet og ytelse)». https://www.ssb.no/klass/klassifikasjoner/393/versjon/1076/koder (accessed: April 30, 2020) -------------------------------------- Citation Please cite as: Dutz, Deniz; Huitfelt, Ingrid; Lacouture, Santiago; Mogstad, Magne; Torgovitsky, Alexander & van Dijk, Winnie (2025). Replication Package for "Selection in Surveys: Using Randomized Incentives to Detect and Account for Nonresponse Bias". DOI: 10.5281/zenodo.16259430