My recent paper in Economic Journal: In the Econ job market men and women are recommended differently: letters for women stress hard work, those for men more likely highlight brilliance. This affects job placement. Link to Paper: t.co/HUieaMuERd With Giovanni Facchini and Markus Eberhardt - ThreadSky

valerueda.bsky.social • 490 days ago

My recent paper in Economic Journal:
In the Econ job market men and women are recommended differently: letters for women stress hard work, those for men more likely highlight brilliance. This affects job placement.
Link to Paper: https://t.co/HUieaMuERd
With Giovanni Facchini and Markus Eberhardt

1 / 2

Comments

peterzorn.de•490 days ago

In what context do children show up so prominently in letters for female candidates?

eckelc.bsky.social•490 days ago

We have them, you know. Men are parents, too, but this isn't mentioned so much.

peterzorn.de•490 days ago

yes, that's my hunch. could also be women work more on children-related topics. word cloud measures frequency, not context.

valerueda.bsky.social•490 days ago

We were careful to focus on sections of letters more likely to talk about the person instead than the research. But some research words make it. Research terms related to research VERY dominated by men/women, eg work on children/ math methods, make it to this graph.

caglaronal.bsky.social•490 days ago

Interesting findings! If I understand correctly, are all these letters from a single institution? Also, is the code for your methodology available somewhere?

valerueda.bsky.social•490 days ago

Thanks!! Re/institutions: These are the letters received by 1 institution. B/c of the Econ Job Mrket Structure, these are broadly representative of the letters received by research intensive institutions. It's 600->800 applicants per year, >12,000 letters. Most from top-100 institutions (see below)

bgreiner.bsky.social•490 days ago

I am still wondering whether you got approval from these letter writers to use their confidential letters in your research, or whether the letter writers were informed about this possibility when they submitted their letter.

valerueda.bsky.social•490 days ago

Re/ Methods: Replication codes are available in EJ's data rep, although not the raw letters for obvious confidentiality reasons. Always happy to talk about methods on this, there is still much to do, & potential for constant improvement given fast moving NLP landscape.

caglaronal.bsky.social•490 days ago

Thank you! I really liked this. Congratulations.

agregory.bsky.social•490 days ago

I once sat on an economics job panel - this is scarily true ( and brilliance as ‘potential’ rather than evidenced by actual output). I think it is a bit less true in history but not absent.

Comments

Posting Rules

Reply