AAS Job Trends

03 Sep 2017

Astronomy Job Trends

A look at the trend in astronomy jobs over the past decade, and a visual exploration of my data analysis thought process.

Monthly job trends.

I scraped every job listing from the AAS Job Register (note: after scraping, the AAS removed all old archives from their site) since 2003. Because the jobs are stored on pages by their month, I naturally looked at the number of jobs posted per month over time. Two things jumped out at me. The first is that the monthly data seem to show an increase over time, but more importantly that there was huge variation within each year. I thought I would check out the yearly cycle first to see if I could explain the scatter.

The yearly job cycle.

Since these are all academic jobs, there is a definite trend that follows the American School year. The majority of jobs are posted in September through December, and the rest of the year is mostly flat.

Yearly average

Now lets look at the average for each year. Here we can see that there is a definite upward trend over the years. This trend is even easier to see if we clear away the monthly points and look only at the yearly totals.

Yearly total

This trend is easier to see if we look at the total jobs each year instead of the average. 527 jobs where posted in 2003, which rose stedily to a height of 891 in 2008. You can see the effect of the recession on the number of jobs posted, which fell to 1289 by 2011. In 2016 1027 jobs were posted, finally recovering from the job losses during the recession.

By country

Hold up though. The AAS is an US based organization yet many jobs are posted from outside the US. Does this have any affect on the job trends? Turns out it does. Jobs outside the US have been steadily rising, while jobs inside the US showed a more pronounced dip after the recession. I can imagine two reasons for this. First, the recession might have hit US science and educaton spending harder than other countries. Second, the AAS job postings may be only recently gaining in popularity for institutions located outside the US. The data turn out to not be great for this analysis, see below for some difficulties.

I wanted to map the locations of all the jobs posted to the AAS job register. One difficulty I faced is that the job locations are listed manually and have large variability (for instance two jobs may be listed for "University of Wisconsin-Madison" and "UW-Madison" which are the same place). Instead of reaching for more complicated natural language solutions, or manually managed the data, I used google maps API to deal with duplicated names. For each entry, I queried the maps and recorded the locations. Locations within 2 km where considered to be the same institute.

Location of academic jobs

Hover over a point for more information.