Replication Materials for "Occupational Selection and the Reliability of Position Generator Measures of Social Capital"

Software Requirement

To undertake replication, you need R. Required packages are listed at the top of the R scripts.

Data

The folder "PG-Reliability/Data and Spreadsheets/" contains a data frame (SSND_PGitems_Replication_Data.dta) for the replication data. It contains 998 rows (one for each respondent, excluding those who did not provide a valid response to one or more of the 30 PG items) and 31 columns, as follows: v48_1 to v48_30 provide information on the relational criterion that links a respondent to an occupation, with values:

• 0: no connection
• 1: connection via a relative (and possibly via a friend or acquaintance as well)
• 2: connection via a friend (and possibly via an acquaintance as well), but not a relative
• 3: connection via an acquaintance, but not via a relative or friend

PS_EDU is an indicator variable scored 1 if the respondent has some post-secondary education.

To replicate our analyses, variables v48_1 to v48_30 must be recoded to specify the relational criterion. For the acquaintance criterion, recode values 1-3 to 1 and 0 to 0. For the friendship criterion, recode values 1-2 to 1 and values 0 and 3 to 0. For the relative criterion, recode value 1 to 1 and values 0, 2, and 3 to 0.

ISEI scores, Sixma-Ultee prestige scores, EGP class codes, and other occupational information (including stata assigments) for the 30 occupations are given in separate spreadsheets ("Strata_assignment.xlsx" and "EGP_assignment.xlsx") that the replication scripts read and use in order to construct the PG measures.

To replicate the analyses discussed in Section 8 for respondents who have some post-secondary education, specify an acquaintance criterion and drop respondents who have values of 0 on variable PS_EDU.

We perform the steps above in the script "PG_prepare_data.R" in the folder "PG-Reliability/Scripts/".

Replication

Below are steps for replicating the results presented in the article:

1. Run the script "PG_prepare_data.R" in the folder "PG-Reliability/Scripts/" to prepare the data for analyses.
2. Run the scripts "PG_pairs.R," "PG_pairs_RelcritCollege.R," "PG_triples_pairs.R," and "PG_triples.R" in the folder "PG-Reliability/Scripts/" to create replicate groups. Refer to Sections 5.1, 5.2, and Appendix A for details.
3. Using the functions "PG_measures_fun.R" and "PG_measures_equalscores_fun.R" in the folder "PG-Reliability/Functions/", run "PG-Reliability/Scripts/PG_funrun.R" to construct the position generator measures for the replicate groups defined in step 2.
4. Using the data files created in step 3, run "PG-Reliability/Scripts/PG_analyses.R" to construct ICC measures for each position generator measure in each replicate group.
5. Using the data files created in step 4, run "PG-Reliability/Scripts/PG_organizefindings.R" to organize the results.
6. Using the organized results from step 5, run "PG-Reliability/Scripts/PG_prophecy.R" to project the reliability of longer position generator instruments. Refer to Section 7.3.1 for details.
7. Run the script "PG_agreement.R" to create the Jaccard Coefficient column of Table 1 as well as Figure 1. You will also need the function defined in "PG-Reliability/Functions/PG_gplot_alt_fun.R" to replicate Figure 1.