4 ms·

An Analysis of Market Demand for Web Programming Languages (October 2012)

6 points by marcamillion 14y ago

paulhauggis 14y ago
My experience is much different (and I've looked in many different sections on craigslist). Python, PHP, Java, and C# were the top. Ruby was almost non-existent.
- marcamillion 14y ago
  Well...you agree with my findings on the PHP, Java & C#. Those are some of the more popular. Python & Ruby are less represented than the tech blog press leads you to believe. Either way, one of the things I am looking forward to discovering is seeing if there are fluctuations throughout the year - simple cohort analysis and such.
owenjones 14y ago
I see you have 'closure' listed under client side languages, are you referring to Clojure? In which case it is most definitely server-side, unless you meant Clojurescript?
- marcamillion 14y ago
  I meant 'Clojurescript' - the Clojure to JS compiler.
zissou 14y ago
What about the wages/salaries + the rest of the labor supply side? Comparing the raw quantity of job postings (labor demand) across cities doesn't have much value. The suggestion you received in your last post about collecting this data over time was the best one you've gotten. Building on that, you need some way to proxy for the number of relevant job seekers in each location (and for each programming language in perfect world) because 10 times more jobs in city A than city B doesn't matter if there are 10 times more potential workers in city A. The other piece that you could think about adding is a way to detect when a job match occurred. Although it's far from perfect, a start would be doing something like inferring from recently removed Craigslist posts that "Hey this post that used to be here is now gone. They must have found a job candidate." Obviously it's a strong assumption since posts can be removed for a number of reasons, including ones that would lead to a false-positive under the assumption I just described. On the other hand, one metric you'd get out of your data if you setup the system this way is the length of time jobs across programming languages and cities remain active (e.g. "Jobs for Python developers in Austin, TX appear to get filled N times faster than jobs for C# developers.") If nothing else, this time-based metric could be used to explain the supply side -- cities that fill jobs for a certain language faster on average may be doing so because there are more potential employees in that area. You could even use Google Trends data to see how trends for searches for "python jobs" or whatever in each city affects the story. It's something to think about and play around with. I've read published economics research papers that used the "post removed" assumption in the market for used products on Amazon.com -- when a unique URL for a used product would disappear, the authors concluded that to be a sale [at the price it was last observed at].
- marcamillion 14y ago
  Well...the suggestions sound good - but they are a bit outside of what I am trying to capture. All I am trying to do is to get a sense for the absolute demand (especially relative to other types of gigs) for web development gigs. I think the only way to get a sense of Wages/Salaries would be to start looking at 'Jobs' - which is outside of the scope of the analysis I am doing. One flaw with the 'post removed' though is that sometimes the posts are removed because of spam (i.e. it was created by a spammer or someone with fraudulent intent). So I am not sure how much of a valid proxy that may be here...especially given that another real assumption for people removing posts is that they get inundated with unqualified leads - so they remove it to stop the inbound flood.

6 points by marcamillion 14y ago

My experience is much different (and I've looked in many different sections on craigslist). Python, PHP, Java, and C# were the top. Ruby was almost non-existent.

marcamillion 14y ago
Well...you agree with my findings on the PHP, Java & C#. Those are some of the more popular. Python & Ruby are less represented than the tech blog press leads you to believe. Either way, one of the things I am looking forward to discovering is seeing if there are fluctuations throughout the year - simple cohort analysis and such.

owenjones 14y ago

I see you have 'closure' listed under client side languages, are you referring to Clojure? In which case it is most definitely server-side, unless you meant Clojurescript?

marcamillion 14y ago
I meant 'Clojurescript' - the Clojure to JS compiler.

zissou 14y ago

What about the wages/salaries + the rest of the labor supply side? Comparing the raw quantity of job postings (labor demand) across cities doesn't have much value. The suggestion you received in your last post about collecting this data over time was the best one you've gotten. Building on that, you need some way to proxy for the number of relevant job seekers in each location (and for each programming language in perfect world) because 10 times more jobs in city A than city B doesn't matter if there are 10 times more potential workers in city A. The other piece that you could think about adding is a way to detect when a job match occurred. Although it's far from perfect, a start would be doing something like inferring from recently removed Craigslist posts that "Hey this post that used to be here is now gone. They must have found a job candidate." Obviously it's a strong assumption since posts can be removed for a number of reasons, including ones that would lead to a false-positive under the assumption I just described. On the other hand, one metric you'd get out of your data if you setup the system this way is the length of time jobs across programming languages and cities remain active (e.g. "Jobs for Python developers in Austin, TX appear to get filled N times faster than jobs for C# developers.") If nothing else, this time-based metric could be used to explain the supply side -- cities that fill jobs for a certain language faster on average may be doing so because there are more potential employees in that area. You could even use Google Trends data to see how trends for searches for "python jobs" or whatever in each city affects the story. It's something to think about and play around with. I've read published economics research papers that used the "post removed" assumption in the market for used products on Amazon.com -- when a unique URL for a used product would disappear, the authors concluded that to be a sale [at the price it was last observed at].

marcamillion 14y ago
Well...the suggestions sound good - but they are a bit outside of what I am trying to capture. All I am trying to do is to get a sense for the absolute demand (especially relative to other types of gigs) for web development gigs. I think the only way to get a sense of Wages/Salaries would be to start looking at 'Jobs' - which is outside of the scope of the analysis I am doing. One flaw with the 'post removed' though is that sometimes the posts are removed because of spam (i.e. it was created by a spammer or someone with fraudulent intent). So I am not sure how much of a valid proxy that may be here...especially given that another real assumption for people removing posts is that they get inundated with unqualified leads - so they remove it to stop the inbound flood.