briatte / apsaconf

R code to download conference panels ('sessions') and papers from APSA meetings from 2015 to 2021 (T = 7 years).

The data cover over 28,500 papers presented by approximately 19,000 participants in over 8,500 sessions .

Year	Sessions	Participants	Papers
2015	972	5,518	3,861
2016	1,131	6,085	4,126
2017	1,221	6,076	4,038
2018	1,294	6,571	4,201
2019	1,499	6,865	4,420
2020	1,301	5,716	3,479
2021	1,368	7,030	4,456

This is work in progress, so the counts are approximate until further data cleaning and other corrections are performed.

Running the download scripts takes roughly two days (the scripts leave 1.5 second between each download to avoid server choking).

Data

programs.tsv contains all datasets below in a single (large) file.

• papers.tsv -- information on papers, identified by their session id ("session" means "conference panel") and paper id. Large file because it contains abstracts.
• participants.tsv -- information on participants (authors/presenters, chairs, discussants), identified by pid, their "person id". There are a few likely homonyms.
• roles.tsv -- the role of each participant (pid) in each panel (session): presenter (p, in which case the id of the presented paper is listed in paper), chair (c), discussant (d), or "else" (e) for very few special cases.
• sessions.tsv -- information on sessions (conference panels), identified by their session id.
• years.tsv -- a short summary of each conference year.

All files are TSV-formatted. Missing values are denoted NA.

Notes:

• Identifiers session, paper and pid are variable-length numbers, but are better handled by treating them as strings to avoid issues with e.g. leading zeros.
• Some sessions (e.g. all-member meetings) have no participants, so the raw data contain more session pages than there are unique sessions in the parsed data.
• A few participants have two pid identifiers in a same conference year, most likely because they created two conference user accounts.
• Session and paper identifiers (session and paper) might repeat over years, which is why the data contain less unique values for those than listed above.
• Similarly, pid is unique only per conference year: it is not fixed through time, and so cannot be used to identify people longitudinally.

Code

• Scripts 01-03 download the raw data, parse it, and create the datasets
• Scripts 04-05 sample conference panels, papers and participants

Script 05 is in draft form and does not yet do much.

Notes

On scraping the website interface:

• 2020 and 2021 require a time zone setting
• 2018 and 2019 use the 'new' All Academic Inc. interface
• 2016 and 2017 use the 'old' interface
• 2015 uses an even slightly 'older' interface that works exactly like the 'old' one

Main recurring session types for recent years, with counts:

type	2018	2019	2020	2021 (in-person)	2021 (virtual)
Author meet critics	59	64	59	24	27
Business Meeting	156	140	80	12	51
Created Panel	611	641	577	319	400
Featured Paper Panel	6	9	.	1	2
Full Paper Panel	376	411	334	159	141
Poster Session	58	63	52	1	65
Reception	77	73	25	30	11
Roundtable	137	167	127	51	87
Short Course Full Day	14	8	2	1	.
Short Course Half-Day	12	12	23	9	.
TLC Full Paper Panel	.	1	1	5	5
TLC Workshop	.	2	8	3	3

"TLC" means "Teaching and Learning Conference". There are many more session types, including some that were replaced by the "Created Panel" type in 2018.

TODO

Other candidate conferences:

• MPSA (All Academic Inc.)
• SPSA (PDFs)
• WPSA (own website, has participant emails)
• SWPSA (All Academic Inc. for some years, PDFs otherwise)
• NPSA (All Academic Inc.)
• NEPSA (PDFs)
• PNWPSA (PDF, no past archive)
• SPPC (different websites, some missing)
• PolMeth (many different meetings, SPM one on different websites)

MPSA and NPSA might be doable, as might some years of SWPSA.