Research Project Description Micro1
“Estimating partner preferences of online daters in high-paid and low-paid occupations”
Marriage partners generally match on intelligence, education, social background, ethnic origin, religion, height, weight, attractiveness, and no doubt other traits too.2 Why is this? There are a number of explanations. One is a preference story: if individuals prefer marriage partners with traits similar to their own (or prefer traits in partners in similar ways), marriage partners will match on similarities. Another story is one of search frictions: if individuals are more likely to meet their partner in self-selected environments (such as clubs, schools, jobs, neighborhoods), they are also more likely to meet a partner with traits similar to their own, regardless of partner preferences. It has proven difficult, however, to empirically isolate the impact of partner preferences using information on realized partner matches alone.
One possible solution for identifying partner preferences is to run a field experiment in the context of an online dating website and send random invitations from fictitious profiles to online daters. With profiles traits manipulated along two dimensions (attractiveness and education), the responses of online daters will then measure the true preferences for partner attractiveness and education. This is what Egebark, Ekström, Plug, and Van Praag (2021) do in their paper Brains or Beauty? Causal Evidence on the Returns to Education and Attractiveness in the Online Dating Market. It is highly recommended that you read this paper.
This project builds on the work of Egebark et al (2021). In particular, students are asked to test whether online daters who work in high-paid occupations respond differently to profile invitations than online daters who work in low-paid occupations. This project will provide insights on a number of theoretical marriage models suggesting that partner preferences might vary with the earnings potential of husbands and wives. In traditional models of household specialization, for example, women are said to prefer men in high-paid occupations over men in low-paid occupations because they can offer more resources for raising children (Becker 1981).
The two data files for this project are (i) a restricted version of the data file used in Egebark et al (2021), with information on the online daters’ age, attractiveness, education, gender; occupation (four-digit occupation classification ISCO 2008), type of profile invitation, and reply; (ii) a data file drawn from the Wage Indicator Survey, with information on hourly earnings per occupation (four-digit ISCO 2008).3
With these data files, students have to (i) distinguish high-paid (above €20) from low-paid (below €20) occupations for men and women separately by merging the datafiles; (ii) replicate the main summary statistics (Table 2, means only) and findings (Table 5, cols 1&2) reported in Egebark et al. (2021);4 and (iii) test whether the partner preferences are different for male and female online daters working in high-paid and low-paid occupations. As a bonus, (iv) students are encouraged to provide a design assessment by showing a table with balancing regressions, and robustness specifications with and without covariates. Finally, they have to describe these findings in a coherent paper.
The papers and data file are available for download on Canvas.
1 Developed by Erik Plug.
2 Economists call these marriage patterns positive assortative matching. Sociologists call these patterns homogamy.
3 For details on the Wage Indicator Survey, we refer to Tijdens, and Osse. (2015). Wage Indicator continuous web-survey on work and wages. Amsterdam: University of Amsterdam/AIAS and Stichting Loonwijzer.
4 Your results will be slightly different because not all control variables are included in your dataset.
To help you with the analysis, here are some tips:
1. Create a crwork.do file that imports the main e-dater data and merges it with the wages data. The codes in your second file are not unique, you will therefore not use the merge 1:1 but merge m:1. The rest of the syntax is similar (see clab 4.2 Q 5).
2. Which data do you not need anymore and can be dropped? (Hint: type tab _merge).
3. Generate a dummy variable named wage_missing. This variable should tell you whether or not wage information for a given person.
4. Generate a dummy variable wage_high that determines earnings higher than €20, but be careful not to include missing wages as well. Use the variable you created in the previous question to guarantee that only individuals for whom we know their wage will get a value. In the code below it is assumed that wage_missing==0 means that you have wage information. Verify whether you have used the same coding; otherwise you need to use the other value you gave to wage_missing to get the same result.
gen wage_high = 1 if wagehour>=20 & wage_missing==0 replace wage_high=0 if wagehour<20 & wage_missing==0
5. Inspect the data and compare with what you see in Table 2 of the Egebark paper. For your analysis you only need age, gender, education, attractiveness, wage_high, treatments (t_ variables), and reply. Save the selected data as work.dta . Start an analysis.do file that opens the working data. Inspect the means for all variables. Specify them by gender similar to Table 2 of the paper. This is your Table 1, in section 3.
6. You might want to create tables in which all models are summarized together. To do that, type first ssc install estout. With the eststo command you can indicate which models you want to put in one table (Cf. Clab 4.1 or Clab4.2).
7. Before you start your analysis you need to make sure that the treatments are balanced. This means that you need to test whether the all photos and profiles have been shown to similar individuals. You need to check whether it does not happen to be so that one treatment was mainly shown to older individuals and all young individuals ended up in one of the other treatments. If that is the case, then the design is not balanced. If the design is balanced then all p-values should be above .05. Report in section 3 whether the design was balanced.
8. First do a regression in which you analyze response behaviour for men and women in general (regardless of income). You have to make separate regression for men and women. Do this by typing at the end of the regression: “ if ‘variable’ ==1(or 0)”. Figure out which variable you have to use instead of ‘variable’. (The code will not work if you type ‘ variable’ as there is not variable called ‘ variable’ in your dataset). Next the two regressions again but this add the background variables of the participants. This is your table 2.
9. Next do a regression for high and low paid men and high and low paid women (four regressions in total). If you want to do a regression for high paid or low paid workers type at the end of the regression ‘ if 'variable1'==1 (or 0) & variable2==1 (or 0)’ at the end of the regression. This way you tell STATA only to include the variable that fulfill this criterion. This is your table 3.
CS 340 Milestone One Guidelines and Rubric Overview: For this assignment, you will implement the fundamental operations of create, read, update,
Retail Transaction Programming Project Project Requirements: Develop a program to emulate a purchase transaction at a retail store. This
7COM1028 Secure Systems Programming Referral Coursework: Secure
Create a GUI program that:Accepts the following from a user:Item NameItem QuantityItem PriceAllows the user to create a file to store the sales receip
CS 340 Final Project Guidelines and Rubric Overview The final project will encompass developing a web service using a software stack and impleme
