20 ms·
Results of Stack Overflow survey of 20,000 developers
- norswap 12y agoWould love to see some analysis of the data. Too lazy to do it myself.
- Fuzzwah 12y agoThere is a subtle but statistically significant relationship between effort and reward.
- eykanal 12y agoThere is a significant relationship between willingness to invest effort and potential payoff. Spending ten minutes playing with what is essentially an online pivottable to possibly find learn an interesting but ultimately useless factoid counts for me as a large investment for a low payoff.
- danielweber 12y agoThe video helped tremendously. Without it I was completely lost.
- notastartup 12y agowhat video?
- danielweber 12y agoAnalyzing data in Statwing Select variables Analyze Video Tutorial [ VIDEO TUTORIAL HERE ]
- ambiate 12y agoThere is a significant relationship between internet points and effort.
- ambiate 12y agoI enjoyed 'hours surfing the internet' for those 'currently looking for a job.' If you have enough time to stay employed and surf the internet: you probably just feel like you're employed to receive a paycheck. Not challenged at all and seeking adventure in a new shiny grass field. Let's read about those greener fields (or burn time) instead of acting on it.
- Gracana 12y agoAs someone with debt and family and school obligations... It can be very hard to act on something like that. I feel trapped.
- hoopism 12y agoI wanted to make fun of it too but started playing with the data and found it to be really easy and actually a lot of fun. How dare they ruin a good cynical take.
- glaugh 12y agoHa. We apologize. Oversight on our part, we're normally more loyal to our good friend cynicism. For what it's worth, this demo is a bit less fun than typical because all the data is categorical (i.e., almost all of it stems from multiple choice questions). Our more typical demo dataset--every project run on the DonorsChoose, like KickStarter for teachers--includes numbers and times, such that you get a lot more variety in output: https://www.statwing.com/demo https://www.statwing.com/demo
- Gracana 12y agoAt first I was disappointed that the dataset wasn't analyzed for me; I wanted to see what an expert (or at least someone who had put in effort) had to say about it. The statwing tool is awesome though... I would absolutely love to use this for something useful. Hopefully I'll remember it if/when that time comes.
- jgmmo 12y agoSeemed like an awfully large amount of developers make less than 20 grand. I did some analysis that I think is a little more useful. Compensation & Job => filtered by 'US only', 'not looking for a job', and 'provided a job'. https://www.statwing.com/demos/dev-survey-2#workspaces/18770 https://www.statwing.com/demos/dev-survey-2#workspaces/18770
- matthewbauer 12y agoAre those people all part-time developers? A $20,000 salary would equate to $9.62 per hour given a 40 hour work week. That doesn't seem to make since for a full-time developer.
- owenversteeg 12y agoExactly - as a 13 year old I made $12 an hour. There's no way that 13-year-old me was making more than 55% of programmers.
- harmegido 12y ago....Do you live in India?
- owenversteeg 12y agoNo (I live, and lived at the time, in the US), but I'm assuming you asked because you were surprised I was working at 13. I was freelancing for some local businesses at the time, and I definitely didn't need the money.
- aiiane 12y agoThey might have been pointing out that 13-year-old you in the US may very well make more than an adult in India.
- 12y ago
- joshfraser 12y agoCompare OS and compensation. Apparently all the highest paid developers use Windows 7.
- dzhiurgis 12y agoEnterprise?
- indenturedsmile 12y agoEnterprise.
- akiselev 12y agoWhat is "enterprise" in this context? Do they work for an enterprise SaaS/vendor or for an actual enterprise, aka big businesses that use software? I don't think either of those are particularly better paid than the other groups. I think it's more likely that the best paid programmers are working in a field with specialized software that only runs on Windows. For example, EEs use Windows almost universally for Altium/Cadence/Mentor Graphics software which means that the majority of commercial firmware compilers and IDEs are Windows only too (even though many are forks of cross platform tools). The same is pretty much true of biotech too: a majority of the software that interfaces with hardware, especially lab automation, is on Windows so anyone dealing with those specialized fields would be start with Windows and run VMs for everything else. Another explanation would be hardware support. My workstation has a high end but not a server mobo (with funky features like Marvell's weird simulated RAID) and dual SLI K-series GPUs (professional/super computing). Getting Ubuntu 14.04 to even boot up, let alone properly deal with monitors and peripherals like a 3D mouse coming and going, takes at best an entire work day. Windows 7 with NVIDIA's drivers and VMWare leave much to be desired if you're a fan of Linux but they just work, especially if you have to do anything with CUDA at the same time as your main line of work. Also: Visual Studio.
- ajryan 12y agoIt was a joke.
- return0 12y agoDid they do some filtering of their users or just raw data? In that case they asked 20000 stackoverflow users, and their results will not be representative of developers as a whole.
- owenversteeg 12y agoSome things I noticed: - 55% of programmers surveyed are making less than $20k/yr, which is ridiculous. - 4.9% of programmers surveyed are women - 2.7% of programmers surveyed are 51+ - 17.5% of programmers earning less than $20k/yr live in California. That last fact caught my eye, so I did some math: This article (http://www.wired.com/2014/04/no-exit/ http://www.wired.com/2014/04/no-exit/) says that the author pays $1250/mo "for a mattress on the floor, behind a panel of imbricated torn shower curtains, in an unheated rabbit warren of 20 bunk beds under a low converted-warehouse ceiling." Assuming they all live in the same space that the author lived in, they spend $15000/yr on rent alone. Assuming that they have health insurance, a cell phone plan, and pay $3/day for transportation, they have (and I swear I'm not making this up) about $3.50 per day to spend on food.
- planckscnst 12y agoOr a bunch of them live off their parents. It would make sense that there is a large population of students living in California using SO.
- lucideer 12y agothe 55% includes unemployed programmers worldwide - 16.7% of programmers with jobs earn <$20k - 2.2% of US programmers with jobs earn <$20k
- ramayac 12y agoI'm a full time developer Java/SQL, finishing a MBA... and I still make < $20k/year. It's ridiculous in USA, but in LATAM it's sad reality.
- thiagocmoraes 12y agoIn which city do you work? If you go to a big IT center in Brazil such as São Paulo or Campinas, $20k/year is just a little over the starting salary for a new grad from a top school. There's LOTS of worse positions, but this is not really that hard to find, considering that HN crowd is probably more qualified than the average developer...
- withme 12y agoOnly ~13k of the 20k reported a StackOverflow reputation, and none of them had more than 10k rep? That seems odd.
- jacquesm 12y agoNo, that makes perfect sense. The ones with the high SO rep are the ones that have the most time on their hands, the others are clocking up paid hours.
- withme 12y agoJoking aside, I'm genuinely confused as to how you survey that many of the site's users and get no one with rep above 10k. Especially considering that high rep users are much more engaged and invested in the site, and so I would expect a much higher response rate from them.
- jacquesm 12y agoWhat's my rep?
- recursive 12y agoAccording to this: http://blog.stackoverflow.com/2014/02/2013-stack-overflow-user-survey-results/ http://blog.stackoverflow.com/2014/02/2013-stack-overflow-us... 10.5% had over 10k reputation.
- glaugh 12y agoYeah, sorry, just looking at this now. These are all represented in the raw data as "10,000" and "5,000", etc., as though they were actual numbers; they should actually be ranges, though, as indicated by that post. Thanks everyone, sorry I just now saw this.
- Aardwolf 12y agoThe data itself is really interesting! But the UI, gee, lots of frustration with tiny scrollbars, things not fitting on my large monitor, mal-aligned tables if scrolled right, gigantic help screen appearing everytime you close something, ">" and "$" of the values used in alphabetical sorting, ... I'd prefer oldschool HTML tables over this...
- glaugh 12y agoHey, can you shoot me an email at contact @ statwing? I'd love to hop on a screenshare for 10 minutes. Agreed that the gigantic help screen thing upon close is dumb. But a couple of the things you mention sound more like browser-related issues/bugs, we'd love to see them live for a second.
- Aardwolf 12y agoHello, Thanks for seeing the feedback! I'm sorry, screenshare will not work out, but the platform I used is Firefox on Linux. The misaligned tables happen when the data is wide, so there is a thin horizontal scrollbar at the bottom that only appears on hover (which my personal taste doesn't like, sorry about that). When using that scrollbar to go all the way to the right, the top labels of the table are misaligned. The thing is, the data would be easier to view if it would simply be wider and the browser could arrange the scrolling of the whole screen(e.g. left/right arrow keys would do it).
- glaugh 12y agoPoint taken about the scrollbar only showing on hover, definitely an issue. For context, there's some tricky issues that come up in building a way to view tables that can scale to 10,000s by 10,000s (e.g., you can't just plop down the html, even though that would suit this particular dataset well). Definitely still got some kinks to work out. And thanks for reporting the misaligned tables on Firefox/Linux thing. Cheers, and thanks for the feedback.
- lifeisstillgood 12y agoAaaargh. This really frustrates me. I would rather have a bunch of text files and a link to the R manual. Really, this is Stack Overflows annual "OK Cupid" moment and instead of a really linkable blog post with two gee-whiz insights I get "here you go, find an insight by spending two hours learning a UI you will never need again" Did I miss the blog where they had already looked at the data themselves?
- glaugh 12y agoPre-analyzed results are available here, also downloadable: http://blog.stackoverflow.com/2014/02/2013-stack-overflow-user-survey-results/ http://blog.stackoverflow.com/2014/02/2013-stack-overflow-us... http://blog.stackoverflow.com/2013/01/2012-stack-overflow-user-survey-results/ http://blog.stackoverflow.com/2013/01/2012-stack-overflow-us... This was just us at Statwing grabbing their data (with permission) and sticking it into our software with the goals of making it easy to explore and exposing folks to Statwing.
- lifeisstillgood 12y agoThank you.
- skeoh 12y agoI was also overwhelmed when I first saw the UI but found it to be pretty simple after having a go at it. Step one: choose the variables you're interested in. Step two: Describe each variable or Relate all variables. I am very impressed by this app.
- zymhan 12y agoNot everyone wants to download a dataset and use a statistical software suite just to see the results of a survey. I found the site extremely helpful.
- jakejake 12y agoThis is really cool - I found it interesting relating the "job satisfaction" statistic with various other things like "hours per week: fixing bugs" or "hours per week: meetings." It goes pretty much as you would expect except that the two extremes seem to have the highest satisfaction. The only one that was exactly like I would have expected was "hours per week: technical support"
- btribbensee 12y agostatwing rocks...
- contingencies 12y agoAfter 6-10 years of experience, some bunch of people wind up self-reporting as 'full stack'. After 10 years, this declines and they become 'Manager / Team Leader' or 'Other', categories which previously did not self-report very much. Not super fascinating, but it confirms my own path (15 years commercially and counting).
- enodios 12y agoI find it interesting that developers at small companies (1-25) are apparently the happiest, but also spend the most time surfing the web.
- danford 12y agoI was surprised more developers used Linux than OS X.
- mqsiuser 12y agoE.g. Ryan Dahl uses Linux and has objections towards OSX. I trust him (and others) and I use Ubuntu on a MBPR (Linus likes 13inch screens with high res & low weight... he actually has a Mac Book Air). Now Ubuntu is maybe okay but for a professional SW-developer/hacker there are a couple of more roads to go down with (BSD, myriad of Linux distros / devices). There is a Mac standing in our living room and if you are an iOS dev ofc. Better than Windows in my opinion.
- sheepmullet 12y agoAs a developer I don't see any compelling reason to get a Mac unless you want to develop for iOS. In terms of dev work what does OSX provide that you don't get with Linux or Windows?
- pbnjay 12y agoIt's a fun data set to explore, but without the ability to compare interaction effects the UI doesn't allow you to get very deep into the data. Also, it'd be nice to be able to filter out the highly correlated tables. Otherwise a cool presentation, I particularly like the little arrows for significance p-vals and CI tooltips.
- glaugh 12y agoThanks. For now, the best way to deal with confounds and interactions is to run an analysis a couple times, filtering for a third variable in various ways to see how that affects the analysis. Obviously it'd also be nice to also have multivariate regression, or at least views that integrate 3 different variables.
- dangero 12y agoFull time remote workers have the highest average salary. That's an interesting data point. Correlation is not causation obviously, but it's interesting that the more time you spend out of the office, the more money you make. (Compensation plus bonus)
- Xixi 12y agoThe reason could simply be that the pool of jobs available when you work remotely is larger/of better quality than the pool of jobs available within a reasonable distance of where you live. Buying a house is also negatively correlated with salary [1] for a similar reason: you are less likely to relocate for a better job (whatever your definition of better is) when you own a house... [1] can't find the source for that claim, sorry.
- dangero 12y agoCould be. Could also be that people who work from home are more likely to ask for what they want. Working from home would be one indicator of that.
- devonkim 12y agoA lot of consultants that are either working out of their home office or at a client site only for so long count themselves as full time remote workers because they have no defined daily place. I know I and several other traveling consultants considered ourselves in this category. I believe a lot of people that are expected to travel frequently (conference speakers, so-called "community evangelists" might also count). I've gotten the impression that most programmers that are traveling consultants are probably able to negotiate with companies that really badly want their specific skills and as correlation to that need results in some wage boosts. I would have liked to see a better breakdown of the compensation side and if bonuses were a higher percentage for remote workers with lower equity to other groups, that'd closely fit the profile of most consulting groups (pre-sales might mix the results up too). The data points they have aren't exactly comprehensive though unfortunately.
- dbtc 12y agoThis might be because only the more senior devs are given the opportunity to work remotely full time.
- shekhar101 12y agoA bit off-topic but the page looks really heavy and responds slowly on a rather powerful machine that I have. Any idea if we can get a csv or raw data of this?
- doweig 12y agoCan someone give interesting stats out of this mess ?