Here is the methodology used in compensation analysis calculations. For links to the data sources mentioned here see “References/Data Sources”
First, a note on data accuracy. Posting accurate data is very important to Transparent California, which is why the data is collected directly from each agency using a legal California Public Records Act request. This data is theoretically the same data that was reported to the IRS for tax purposes for each employee and therefore is the most reliable source of actual compensation data.
“Full time employee” definition
All employees – those making more than the equivalent of a full year’s pay at the then-current minimum wage in California. If an employee’s regular pay did not exceed that, we do not consider them as full-time employees.
Certificated employees – those making equal to or more than the contracted minimum starting rate for such employees, based on California Department of Education “Certificated Salaries and Benefits” data. If the employee did not make the minimum salary for their district, we do not consider them full-time employees.
Employee group classification:
Employee position classifications (e.g. “Administrative”, “Certificated”, etc) are defined by job titles. For example, a job with “Teacher” in the title is assumed to be “Certificated”, while a job with “Principal” in the title is assumed to be “Administrative”.
This can be somewhat unclear at times. For certificated and classified employees this classification is often detailed in labor group agreements. A best effort has been made to connect job titles to the appropriate classification, however any feedback on mis-classified jobs is welcome.
Total pay is all paycheck compensation, for any reason. This is usually mostly the value of periodic paychecks but can also include one-time payments, stipends, bonuses, etc.
Total compensation includes not only paycheck compensation but also all money paid by the employer toward non-paycheck benefits. Typically this means contributions to retirement plans (usually pensions, but sometimes defined contribution “401k-style” plans or deferred compensation), plus money paid by the employer toward the employee’s healthcare insurance premiums.
In public employment a much larger portion of total compensation is given in the form of contributions to retirement plans than is typical in private industry. A full discussion of the nature of public employee benefits in education can be found here.
A comparison table contrasting contributions made by private employers vs. those made by public employers (from 2021) is below.
In it we can see that the normal contribution by private employers toward employee retirement is 10.20% of the employee’s pay.
In public employment, for employees covered by the California State Teacher’s Retirement System (CalSTRS), the total contribution is 26.48% of employee’s pay.
This means public employees receive a “benefit advantage” of 16.08% of their pay, in the form of contributions to their retirement.
For someone with a total pay of $90,000/year, that advantage is $14,472 per year. That amount generally grows annually as the pay total grows.
If the employer contribution stayed consistent at that level for a 30 year career and if that benefit were put in the average 401K plan earning average market returns, that is a benefit worth $2.6 million at retirement.
Given the magnitude of this benefit, to compare public employee compensation to private one must include this value in the calculations, which I call “comparable pay”.
For an employee making $90,000/year, total comparable pay would be $104,472, meaning a private employee would need to make $104,472/year (and make their own contribution of $14,472/year to their own retirement plan) to receive the same level of compensation.
|Data For FY||2021|
We often hear that public employees (particularly teachers) are poorly paid based on what they would make if they had instead chosen to go into private industry. It is commonly accepted if a teacher were to use their education to work in private industry instead of teaching the resulting compensation would be higher, creating a “teacher penalty” in compensation.
Fortunately there is a somewhat easy way to validate that. The CDE gives data on educational attainment by district annually, and the US Census Bureau publishes data on median income by educational attainment – both are linked in the Resources page.
If we use the Census Bureau numbers as a starting point and weight them based on the educational attainment of teachers in a district, we can come up with a comparable number.
For example, let’s assume the district has 40% teachers with bachelor’s degrees and 60% higher education.
Let’s say the Census Bureau numbers show the median income for someone with a bachelor’s degree is $50,000 and for those with advanced degrees $100,000.
If we weight this number 40-60, the median comparable wage for private employees with education comparable to teachers in this district would then be:
($50,000*.40) + ($100,000*.60) = $20,000 + $60,000 = $80,000
We know that teachers take an additional year of education to obtain their certificate. That means their education process takes 25% longer than a bachelor’s degree only.
Accordingly we’ve weighted our calculation for this.
Using the example above, the difference in income between advanced and bachelors’ is ($100,000 – $50,000 =) $50,000. If we add 25% to the “bachelor’s only” pay that gets us to a total equivalent pay of ($50,000 + ($50,000 * 0.25) = $62,500)
($62,500 * .40) + ($100,000 * .60) = $24,960 + $60,000 = $84,960
For calculations involving change over time, longitudinal analysis is necessary to find the rate of increase within a district.
One might think that it would simply be a matter of looking at the state’s “J90” reporting (“Certificated Salaries & Benefits”) and tracking the average rate of pay for a district to see how, for example, teacher pay has changed. This is misleading, however, for two reasons.
The first is the fact that the J90 averages are determined by having districts report how many employees they have at each point in the salary schedule. Not from actual compensation records. This fails to capture any additional pay given to employees for other reasons – stipends, bonuses, and other forms of income. The J90 average of a district is typically about 2-3% lower than the actual average from pay records.
Average is also a misleading way to track change over time, because the mix of employees changes. Having more senior employees leave and more junior take their place can affect the overall average in ways that do not actually reflect the real rates of change.
For example, let’s say in one year we have 100 employees, all making exactly $100,000/year. The total payroll for that year would then be $10,000,000, and the average would be $100,000/year.
The next year, 50 of those retire and are replaced with employees making a starting rate of $50,000/year. The remaining 50 are all given 10% raises and are now making $110,000/year.
The total payroll is now (50 * $50,000 = $2,500,000) + (50 * $110,000 = $5,500,000) = $8,000,000. That means the average is now $8,000,000/100 = $80,000.If we are looking at averages, it appears the district has CUT people’s pay – they have gone from an average pay of $100,000/year to $80,000/year.
If we used averages, since the average pay has declined from $100,000 to $80,000, the headline of this analysis might be “Employees see huge pay cut!” Reality is quite different.
In reality, those who stayed received what most would consider a very generous raise – 10%, while the new hires have gone from making nothing (presumably they just graduated from college) to making something – an infinite raise.
Longitudinal analysis avoids these inaccuracies by looking strictly at cohorts of individuals who have been with the same employer from the beginning to the end of the period being examined, and calculating actual change rates for those people.
Longitudinal analysis must be done by matching employee names since no other more specific data (employee ID) is available. This means all such analysis is likely partial given that names are sometimes changed and can be reported differently in different data sets (“Bob” instead of “Robert”, or “Robert” instead of “Robert T.”) however the matched sample sizes are typically large enough to be significant as an indicator of total numbers.
It is, for instance, very unlikely if the organization is giving 50% or more of it’s employees raises at a certain rate that they would not apply that same rate to most – if not all – of the other employees of the company.
Employees who recorded base pay declines between the beginning and latest data periods were excluded from the longitudinal analysis, on the assumption that these declines were due to changes in job responsibility, retirement or termination, or voluntary action (like job sharing.) Cases of involuntary pay reductions or demotions are exceedingly rare, almost unknown.
Median vs. Average:
In my analysis, I use medians whenever possible. Median means “mid point” – half of the data is above the median, half is below, and is a far more accurate indication of values applicable to most of a population.
Imagine a group of 10 employees. 9 of them make $50,000/year, and one (the CEO) makes $500,000/year.
Total payroll would therefore be $50,000 * 9 = $450,000 + $500,000 * 1 = $500,000, total $950,000.
The average of this – $950,000/10 – is $95,000. Which not one employee actually makes.
The median of this group, however, would be $50,000. Which most of them make.
If one were looking for employment, saying this company’s average pay was $95,000 would be deceptive since no one in the company makes that much except one person – who makes far more.
Saying this company’s median is $50,000, however, would present a much more accurate picture of what to expect working there.
The US Census Bureau typically uses medians in it’s summary data, which allows for more direct comparisons.
If you have any questions about this methodology feel free to raise them by email to email@example.com