We built a Game Industry Salary Explorer. It enables you to instantly search over 5,000 actual game developer salaries. In this post, I detail how the data was gathered, provide a brief analysis, and offer several points of caution when interpreting.
As an employee, it can be difficult to know how much you could be getting paid. Maybe you've asked a few friends how much they make, explored self reported salaries on Glassdoor, or looked at annual surveys. Another source, that many people aren't aware of, is H-1B data.
As part of the H-1B Visa program in the United States, companies are legally required to report salary information for foreign workers. The raw data is published by the Department of Labor.
Christer Ericson was one of the first to analyze game industry H-1B salaries back in 2008. We built upon his work by filtering through H1-B datasets from 2009 through 2015, resulting in 5,230 salaries at 301 game companies.
Of the sample we looked at, the median salary for H1-B workers in the games industry from 2009 through 2015 was $99,266.
The top 25% earned between $120,733 and $250,000. The bottom 25% earned between $29,615 and $77,636.
Using the keyword classifier from our game job search engine, we looked at specific job categories. All salaries shown below are expressed in thousands of US dollars per year.
A box plot makes it easier to examine earning potential across categories.
Here are the salary distributions for the three largest categories. The left vertical axis is the salary bucket. So 90 is the bucket for all salaries between $90,000 and $99,999. The labels on the right side of each graph indicate how many salaries fall into each bucket. Note that the horizontal scale of the three graphs varies. The bucket in which the median occurs is highlighted in darker blue.
If you use our Salary Explorer, you can view other distributions. They update in real time as you search and filter.
It was quite a bit of work to prepare the data. Here is the process I followed detailing challenges and assumptions made at each step:
- Downloaded and combined datasets. Handled schema changes between years.
- Filtered game companies. There were 239,380 companies overall. Using the list of companies from our job search engine was a useful starting point. My general criteria was that a company must make games or primarily serve the games industry. Middleware, hardware, and service companies, such as Twitch, were included. Gambling and casino companies were excluded. I included companies that have since gone out of business. For massive companies like Microsoft, Google, and Apple, it's near impossible to determine which roles are games related. Most of the job titles are fairly generic (e.g. "Software Engineer") so I only included those with an obvious keyword in the title (e.g. "games" or "xbox"). I did my best to exclude companies with similar names. For example, there is a financial company in Chicago named Twitch but the one we're interested in is referred to as Twitch Interactive in the H-1B data.
- Filtered rejected applications. Some of the H-1B's were not approved.
- Fixed incorrect data and typos. There were a fair number of typos across title, company, and location that I corrected (e.g. Diesigner => Designer). I also standardized companies that had multiple name variations or aliases. There were one or two cases where the salary period seemed incorrect (e.g. $50,000/hour => $50,000/year).
- Categorized based on title keywords. Over time, we've developed a list of title keywords per category that are used for assignment. Jobs can be tagged with multiple categories. Classification is challenging. Some job titles are vague (e.g. "Senior Associate" and "Localization QA Analyst (Engineering Writer)"). The actual role performed by a given job title can vary from company to company. At some companies a Test Engineer is performing manual testing and at others they are primarily writing code. Some roles are truely cross discipline, while others are single discipline even though the title suggests otherwise. These are just some of the examples. Improving the classifier is a continued effort.
- Converted salaries to a standardized format. Hourly, weekly, and monthly wages were converted to yearly. I assumed 40 hours per week and 52 weeks of work per year. Most of the H-1B records listed an exact salary but some specified a range. For the ranges, I used the midpoint of the minimum and maximum. Often the range was small but some companies, such as Valve, have large ranges (e.g. $70K-200K) which is a potential source of error. Finally, I used US Consumer Price Index data to convert all salaries to 2015 US dollars. At most, this increased salaries by 10.5% (those from 2009). The CPI adjustment was for the purpose of this analysis only. Search results show the original salaries.
- Filtered extreme outliers. There were 14 positions, mostly executives, making greater than $252,960. You can still find them in the search results but I excluded them from the box plot.
Gamasutra's annual survey is another popular source of salary data. The 2014 report (of 2013 salaries), found an average US salary of $83,060 (sample size of 1,246 according to footnote).
This table compares survey results with H-1B findings.
|2013 Survey Average||2009-2015 H-1B Median|
|Business and Management||101||105||Business|
|Programmers and Engineers||93||106||Engineering|
|Artists and Animators||74||79||Art|
|QA Testers||54||82||Quality Assurance|
Some category comparisons are within the same ballpark while others are substantianlly different. Here are a few considerations:
- Classification is potentially the biggest source of error. It's unclear if the survey classified based on title or allowed respondants to self classify. Did they allow multiple classifications per job? I expect self classification might be more accurate but you still have job definition differences from person to person. It's also unclear what types of companies the survey considered "true" game companies and which ones they filtered out. We included producers in the product management category, where as I suspect the survey might have included product managers in their business and management category.
- Small sample sizes - Many of the H-1B job categories have a small number of samples. Be wary of drawing too many conclusions for N < 100. I suspect the survey suffered from this issue since it only had 1,246 responses as compared to 5,230 H-1B's.
- H-1B skew - It's expensive, in time and money, to hire H-1B workers. I expect that larger companies are more likely to hire H-1B's. And I expect the roles to be more senior, in order to justify the increased hiring cost.
- Average vs Median - I would expect averages to skew higher since most of the outliers are large salaries. Since both studies removed extreme outliers, this impact should be minimized.
- 2013 vs 2015 USD - Adjusting the 2013 survey results to 2015 dollars (which we used for H-1B data) would be a 1.7% increase. This is relatively small adjustment. I suspect a bigger issue is doing a multi year vs single year comparison.
Keep in mind that the H-1B data is dominated by a few massive corporations. Electronic Arts, Activision, Blizzard, Sony, and Zynga account for over 1800 salaries.
The H-1B data is also heavily concentrated geographically. 75% of the salaries are from California. However, I think this is fairly representative of the overall games industry. Our job search engine shows that 61% of active US job listings are in California. If you're not willing to relocate to one of the major hubs (California, Washington, Texas), you're going to severely limit your employment opportunities.
If you enjoyed this post, please join our community of game developers.
Comments, questions, thoughts? I'd love to hear from you. My email is firstname.lastname@example.org.