What is salary?
Salary represents the estimated annual base salary, excluding bonuses, benefits, or additional compensation. We show salary either as a median (50th percentile) or as a range (25th percentile to 75th percentile).
What can I use this data for?
You can use salary data to understand how compensation varies for different roles or locations. The salary we display is always for all employers for your searched role and location. In other words, we aren’t able to show salary estimates for specific employers.
Acquire: Where do you get your salary data?
We use artificial intelligence (AI) to identify and extract salaries from job postings. Extracted salaries are then added to our historical database, which contains salaries from the past few years. Learn more about how we find and process job postings.
Of course, not every job posting includes a salary. To fill in these gaps, we rely on statistical modeling, which attempts to predict real-world values based on a set of assumptions and a limited number of sample values. In the case of salary, we use our model (which is based on actual job postings) to predict salaries for job postings that didn’t originally include them. These modeled salaries are attached to job postings and stored in our database. To account for varying labor market realities in different countries, we develop a separate model for each country we provide salary data for. We validate modeled salaries using government data (for example, from the Bureau of Labor Statistics in the United States, StatsCan in Canada, etc.).
Organize: How do you prepare your salary data for analysis?
When our AI extracts a salary from a job posting, it runs a number of checks and conversions to account for differences in currency, pay period, and more.
Currency: For example, distinguishing between the United States dollar and the Canada dollar
Pay period: For example, converting hourly wages to annual pay
Abbreviations: For example, converting 10K to 10,000, and 1.5 lakh to 150,000
Numerical formatting: For example, converting 1.500,00 to 1,500.00
We also regularly compare salaries in our database to data from third-party websites to ensure their accuracy.
Analyze: How do you calculate salary?
Each search you run returns a certain number of relevant job postings. Each of these postings has a salary associated with it, either one that was included in the original posting, or one that we assigned to the posting based on our salary model.
To translate these salaries into a range that accurately represents all salaries for a role and location, we fit the salaries to a log-normal distribution, which is the most commonly-accepted distribution for salary information. From this distribution, we identify the 25th, 50th, and 75th percentile salaries.
Deliver: How do we communicate salary?
We show salary either as a median (50th percentile) or as a range (25th percentile to 75th percentile).
More about salary:
When was your salary data last updated?
We are constantly updating our database with jobs that are posted with salaries. In this sense, our salary data is regularly refreshed.
The frequency with which we update our salary models (which estimate salaries for jobs posted without salary data) is dictated by client feedback, major market variations, and our team’s capacity. Typically, we try to revisit our models every 1 to 3 years.
What countries do you have salary data for?
We have salary data for the following countries:
Why don’t you have salary data for my country?
In many countries, the salary data isn’t sufficient to build a strong and reliable salary model. However, we are constantly adding to our database and regularly evaluate the possibility of building models for additional countries once the data meets our quality standards.
What location types do you have salary data for?
We are able to provide salary data at the country, state, and metropolitan area levels.
What search filters impact salary?
If you recall, there are two types of salaries that you might come across in our platform: extracted salaries (which are included in job postings) and modeled salaries (which are estimated based on our salary model and assigned to job postings).
Extracted salaries are responsive to all of the job attributes in our search experience: location, function, occupation, employer, skills, credentials, title, experience level, education level, employment type, and keywords. In other words, for job postings that were published with salaries, you can see how salary varies according to each of these attributes.
Modeled salaries are responsive to some (but not all) job attributes in our search experience: location, occupation, and experience level. This is because our salary model is only able to estimate salaries for these particular attributes.
How can I convert an annual salary into an hourly wage?
To convert annual salary to an hourly wage, divide the annual salary by the number of “work hours” in a year.
To calculate the number of work hours in a year, multiply hours worked per week by 52. For example, 40 hours per week × 52 weeks per year = 2,080 hours per year.
If an annual salary is $75,000 for a job that requires 40 hours of work per week (or 2,080 hours of work per year), then the hourly wage for that job is $36.06 ($75,000 / 2,080 hrs = $36.06 per hour).
Do you have a minimum threshold for displaying salary data?
If a search returns fewer than 5 job postings, we don’t display salary data.
Why don’t you use self-reported salary data?
While we believe that people are generally honest when taking surveys, we also recognize that there can be issues with self-reported data, especially when it comes to salary.
People may be economically or socially incentivized to report an amount greater than their actual salary. For example, someone might include a higher salary on a social profile in hopes of increasing the amount they’re offered for a new job. They might also report a higher salary because they (consciously or not) perceive their value to be greater than their actual compensation.
People may also simply misremember their salary or report the figure in an incorrect format. For example, someone may include commission and bonuses, or report their take-home pay (salary minus taxes and benefits) instead of their pre-tax base salary.
In light of these issues, we don’t believe that self-reported salary data is reliable enough to use in our model.
How accurate is your salary data?
When comparing our salary values to those from a major third-party payroll provider, we identified a slight upward bias for 50th percentile salaries in our data. In other words, the average salary for a role in our database is slightly higher than the salary for the same role in the payroll provider’s database.
We believe there are two reasons for this. First, we hypothesize that companies with better pay are more likely to include salaries in job postings, causing selection bias in our model. Second, we hypothesize that salaries in job postings may sometimes be intentionally inflated in order to entice candidates to apply.
Why does your salary data differ from my organization’s internal data?
When conducting a compensation study, organizations often consider only a specific subset of competitors. In contrast, our salary data reflects all employers in a market, including small businesses, government organizations, universities, non-profits, and large corporations. (As a reminder, our salary data doesn’t take employer into account, even if your search includes one.) Because these inputs are so different, the salary from a compensation study may be very different from the salary returned for your search.
How do you account for limited salary data?
Even with a database of more than 600 million salaries, sometimes a search doesn’t return enough data to estimate salary using our standard approach. This typically occurs for very niche or very senior positions for which there aren’t a lot of online postings or postings with salaries.
In instances of limited data, we adjust the salary distribution and leverage Bayesian techniques (instead of our regular log-normal distribution), which provides the best results for smaller sample sizes.
What is a log-normal distribution? Why do you use this distribution in your salary calculations?
Statistical distributions allow researchers to summarize data by showing a range of possible values and how often those values occur. In this case, we are summarizing the salaries associated with job postings returned for your search.
Instead of using a normal distribution, which is symmetrical and distributes values evenly above and below the mean, we use a log-normal distribution, which is asymmetrical and features a long right tail. In other words, the distribution is skewed towards lower salaries.
This skew is appropriate because most salaries cluster at the lower end of the range, either due to the minimum wage or to pay bands. However, there are often a few salaries at the higher end of the range, perhaps because candidates negotiated or because they’re high performers. In other words, most people make just below the average salary, but a few people make considerably more. A log-normal distribution better accounts for this pattern, resulting in a more accurate representation of salary.
Does your currency converter reflect the most recent exchange rate?
The currency converter’s exchange rates are updated daily.