While I was in Arkansas this week, I estimated that two-thirds of the state's teachers won't work long enough to qualify for adequate retirement benefits. Despite basing my estimates on the state pension plan's own actuarial assumptions, some legislators and audience members pushed back, arguing that the pension plan itself helps keep turnover down. As we've written before, the evidence does not support this argument, at least not for workers of all ages and experience levels. But the Arkansas data provide a few more reasons to doubt how much pension plans affect retention rates.
First, Arkansas teacher retention rates have fluctuated over time, even under the same pension plan. Using the state's own actuarial assumptions (drawn from actuarial reports here)*, I calculated five-year survival rates for teachers. The five-year survival rate is a nice round number, but it's also when teachers "vest" into the Arkansas Teacher Retirement System and first qualify for a retirement benefit. Here are the expected five-year survival rates for teachers at various points in time:
Five-year survival rate for Arkansas teachers, as of 2007: 51 percent
Five-year survival rate for Arkansas teachers, as of 2017: 60 percent
This may be contrary to public perception, but the Arkansas pension system at least is seeing lower rates of turnover than it did 10 years ago. These forecasts are based on historical data and are used in the state's official financial projections. The state's assumed teacher retention rates climbed 9 percentage points (about 17 percent) during the last decade, at a time when the pension plan benefits stayed more or less the same.
Second, as we've noted before, retention rates vary widely across districts, schools, and roles, and Arkansas is no exception. Arkansas doesn't release enough data to dig in too far, but its "teacher" pension plan also includes school support workers, and it disaggregates retention rates specifically for those workers. In contrast to teachers, school support workers in Arkansas have much higher turnover rates. Here are their five-year survival rates:
Five-year survival rate for Arkansas school support workers, as of 2007: 28 percent
Five-year survival rate for Arkansas school support workers, as of 2017: 23 percent
To put a finer point on it, just 23 percent of Arkansas school support workers stay long enough to qualify for any pension at all. They're paying into the pension system, and their employer is contributing on their behalf, but they'll never get anything out of it. Again, the pension plan stayed the exact same throughout this time period, and it was the same pension plan for school support employees as it was for teachers.
Third, I also looked at the assumed retention rates used by the Arkansas Public Employees Retirement System (APERS). In benefit terms, the ATRS and APERS plans are extremely similar, except that APERS covers state and local government employees. For those workers, APERS assumes a five-year survival rate of 27 percent, and it hasn't changed this assumption in the last 10 years.
Now, none of this proves pension plans do or do not have an effect on retention. That would require a much more rigorous study than the superficial look I've presented here. It's possible that pension plans do exert some retention effects on certain groups of teachers, but that those effects are being swamped by other factors like salary or working conditions. But neither of the Arkansas plans assume that workers change their behavior around the vesting period in order to qualify for a pension. And Arkansas' diverging trends do suggest that pension plans are not enough to protect against fluctuations in retention rates over time, or against some groups of workers having much higher or lower retention rates even in the same pension plan.
*Note: Throughout this post, I use the state's turnover estimates for females. The rates for men are even higher.
The following piece originally appeared on The 74.
For nine days, West Virginia teachers went out on strike over concerns about salaries and health care benefits.
Although it’s true that West Virginia has low average salaries, this statistic misses out on lots of other things. For example, I’ve written before about how growing retirement costs are eating into teacher salaries, and it turns out West Virginia is a prime example of this.
In fact, while West Virginia ranks in the bottom five states in teacher salaries, it ranks in the top five in terms of retirement costs. (Most of that money is going toward paying down unfunded liabilities, but it’s a real expense incurred by the state and its school districts.)
Overall, West Virginia climbs to 32nd in terms of salary plus retirement costs. That’s a very different story than the one being told by the media.
Of course, this also doesn’t get into cost of living.
West Virginia lands in the middle of the pack on cost of living, and if we adjusted the raw salary figures based on how far $1 would go, West Virginia teacher salaries would rank even higher.
None of this is to make a judgment about how much West Virginia teachers should be paid, but that argument shouldn’t hinge on the “average teacher salary” metric.
Without looking at all forms of compensation or adjusting for cost of living, average teacher salary rankings don’t tell us all that much.
Amidst the strikes and protests about state education funding and teacher pay, the problem of rising benefit costs, such as pensions and healthcare, went largely unnoticed. That’s a mistake because spending on benefits affects how much money gets to the classroom and influences teacher pay.
In a new report, I look into just how much spending on benefits has changed at the district-level between 2005 and 2014. The results are troubling.
As shown in the graph below, a large part of the problem is that benefit costs are rising at a far faster rate than overall K-12 spending. After adjusting for inflation, nationally we spent just 1.6 percent more on K-12 education generally, but benefit spending increased by 22 percent. The net effect is fewer dollars getting to the classroom. Nationally, benefits consume 19 percent of all K-12 spending. That’s a more than 3 percentage point increase since 2005
The trend of benefits eating up a growing share of K-12 budgets is consistent across the country but varies significantly from state-to-state. In West Virginia, for example, 25 percent of K-12 spending goes to benefits. Once changes to states’ overall investment in K-12 education are taken into account, 23 states actually sent fewer dollars to the classroom in 2014 than they did in 2005 after adjusting for inflation. For instance, North Carolina increased its education budget by 2 percent, but its benefit spending rocketed up by 48 percent. As a result, the state actually sends approximately $589 million less into classrooms.
Rising benefit costs is a problem for educators and policymakers alike. For teachers, much of the rising costs go to pay down debts and thus actually don’t correspond with more valuable pensions or better medical benefits. And worse, without similarly rising overall K-12 spending, fewer dollars are actually spent in the classroom. For state legislatures, the growing benefit costs squeeze budgets. To actually increase K-12 spending, states need to growth their investments more significantly than they have in the past.
This problem likely will get worse before it gets better and should worry educators and legislators alike. There are no easy fixes to these problems, but it will be critical for legislators to find solutions that balance past paying down past obligations with contributing to the education of current students.
Stagnant teacher salaries have been a source of political consternation in recent months. Striking teachers in five states have pointed to low salaries as a key reason for walking out. Meanwhile, Los Angeles teachers are threatening to strike as well, and Miami voters will be deciding on a big increase for teacher salaries this fall.
While teachers make less money than similarly educated professionals in other fields, another claim often made by teachers is that their salaries have seen no improvement over the last few decades. They point to average teacher salaries, which have declined by $630 in real dollars since 1991, to bolster their claims. However, average teacher salaries alone do not tell the full story of what’s happening to teacher compensation.
Chad Aldeman explains this concept in more detail here, but to summarize, declining average teacher salaries may reflect changes in teacher cohorts rather than an actual decline in compensation for teachers at any given point in their career. Young, newly-hired teachers are paid less than experienced teachers. For this reason, if one experienced teacher retires and is replaced by a first-time teacher being paid an entry-level salary, average teacher salary declines despite the fact that no teacher’s pay was cut. This means that lower average teacher salaries don’t necessarily indicate worsening conditions for teachers. Instead, they could indicate that the profession has become more mobile.
Another way of looking at teacher compensation can be found in state teacher pension plans’ actuarial valuations. These publications include a list of assumptions used to estimate the plan’s future payouts, including assumptions for teacher salary growth rates. These assumptions are compiled through regular observations of what’s happening in the state, meaning that the salary growth rates should closely mirror the actual experiences of teachers at the time.
These salary growth rate assumptions can tell us a lot about how teacher compensation has changed over time. If the growth rates are going down, that means that teachers’ salaries are not growing at the same pace as they used to, bolstering claims that teachers are losing out.
So, do these salary growth rates tell the same story as declining average teacher salaries? Unfortunately, the overall answer appears to be yes. Not all states report the data every year or keep their reports online, but we took a random sample of five states (Arkansas, Connecticut, Michigan, Missouri, and Texas) with actuarial data going back to at least the mid-2000s. Each state’s nominal salary increase assumptions demonstrated a marked decline, and four out of five states saw decreases in real growth rates as well.
One way of visualizing this change over time is to look at how real salaries that started at the same level would progress over a career as each state’s assumptions changed. For example, here’s a graph of what a Missouri teacher’s salary growth would look like over a 30-year career after each revision of the state’s actuarial assumptions over the last decade, keeping their starting salary constant at $40,000:
In 4 of the other 5 sampled states, the graphs look similar to Missouri’s. In every state but Connecticut, real teacher salary growth is lower now than it was in the mid-2000s*. By the end of teachers’ careers, these differences translate to thousands of dollars less in real yearly salary:
In Arkansas, a teacher who began her career with a salary of $40,000 in 2004 would have been expected to make $95,824 after 30 years. Today, she would be expected to make $79,384.
In Connecticut, a teacher who began her career with a salary of $40,000 in 2004 would have been expected to make $61,538 after 30 years. Today, she would be expected to make $74,988.**
In Michigan, a teacher who began her career with a salary of $40,000 in 2002 would have been expected to make $143,676 after 30 years. Today, she would be expected to make $140,272.
In Missouri, a teacher who began her career with a salary of $40,000 in 2007 would have been expected to make $88,361 after 30 years. Today, she would be expected to make $74,648.
In Texas, a teacher who began her career with a salary of $40,000 in 2006 would have been expected to make $82,052 after 30 years. Today, she would be expected to make $63,361.***
There’s one big caveat here: we don’t know how average starting salaries have changed over this timeframe. Even if salary growth rates are decreasing, it’s possible (although unlikely) that a large simultaneous rise in average starting salary could mitigate the slower growth.
That being said, there’s reason to believe that these declines really do signal a worsening of teacher pay. For one, it corroborates what we're seeing in the trends in average teacher salaries. In addition, despite education budgets increasing in many states over the last decade, these declines in salary growth are coming at the same time that benefit costs are ballooning massively, eating more and more into the money that could be available for teacher salaries. As contributions to unsustainable teacher pension and health benefit plans continue to rise, many cash-strapped municipalities are facing impossible choices between paying teachers what they deserve or maintaining adequate funding of other core services.
Sadly, it’s hard to see these declines reversing themselves in the near future. Benefit spending will continue to ramp up over the next few years in most states. And if teacher salaries have declined throughout a period of economic expansion, it’s frightening to imagine what might happen if another recession hits and budgets are slashed. Nevertheless, we owe it to our teachers—and our students—to find the money to ensure that teachers are compensated in ways that fully reflect their worth and skill.
*Years vary based on data availability and when the states updated their assumptions.
**Although Connecticut’s nominal salary growth rates have declined, the state lowered their inflation assumption from 4% to 2.75% over the same time period. For this reason, assumed real salary growth is higher today than it was in the mid-2000s.
***Texas reports separate salary increase assumptions for men and women until 2008. The pay growth rate used in 2006 is for women only and is slightly lower than the growth rate for men. This means that the fall in average salary increase over this time period is actually slightly greater than what is reported above.Taxonomy:
We also know that fewer Americans have access to defined benefit pension plans, putting more of the burden of saving for retirement on individual employees.
Are these two trends compounding each other? The short the answer is no, this is not happening. As I argue below, this is mainly due to our investments in progressive programs like Social Security and Medicare. The transition from defined benefit to defined contribution plans has been going on in the background, and it may seem at first blush like this transition could increase inequality, but that would be a misread of how the plans work.
First, consider the logic. Both defined benefit and defined contribution plans award benefits as a percentage of salary. Consider two workers in the same firm, a low-paid manual labor and a higher-paid executive. If, over time, the salary of the executive rises faster than that of the manual laborer, their retirement wealth will also become skewed as well. But crucially, this would happen in direct proportion to the change in salaries, regardless of plan type. Under either a defined benefit or a defined contribution plan, the executive gets a higher retirement benefit in direct relation to his or her higher salary.
If anything, the above example might be under-selling things a bit, because it assumes both types of workers stay with their company for the same length of time. Traditional defined benefit pension plans reward both high salaries AND long tenures, and we know that higher-paid executives have much more employment stability than lower-paid manual laborers. In this scenario, the manual laborer would be better off in a front-loaded defined contribution plan than a back-loaded pension plan.
The politics of this are confusing, since liberals tend to be the ones defending defined benefit pension plans, but there’s nothing special about them in this regard. Traditional pension plans are not inherently progressive. In fact, so-called “teacher” pension plans tend to reward higher-paid school administrators over more transient, lower-paid teachers.
Ok, enough with the theory, what do the data show? Retirement wealth is a function of participation and contribution rates. Once you combine both types of plans, workers are no more or less likely today to participate in a retirement plan than they were 30 years ago (see Figure 20 here), and total retirement savings has also kept pace. That is, the type of retirement plan has not dramatically changed our collective retirement preparation.
In terms of results for individuals, earlier this year I wrote about a paper from Adam Bee and Joshua Mitchell from the U.S. Census Bureau. Among Americans age 65 and older at the 25th percentile in terms of income, they found income gains of 31 percent from 1990 to 2012. For those at the 75th percentile, the gains were similar, at 32 percent. They also found that poverty rates among those 65 and older declined from 9.7 to 6.7 percent. To be sure, these stats do not prove that there is no inequality in retirement wealth, only that we’ve provided retirees with some buffer against it.
Namely, Americans over age 65 have a giant social safety net through Medicare and Social Security. To show just how important Social Security is, consider the chart below, from the Federal Reserve. The first two columns show that the top 5 percent of households have about 13 times the income as the bottom 50 percent of households, and about 110 times (!) the retirement wealth in the form of DB and DC retirement plans. That is, wealthier people have much more in the way of retirement assets than lower-income people. This has always been true, and income inequality is making these trends worse. But look at the last column, which compares total retirement wealth (including Social Security) with income while working. As a ratio of retirement wealth to wealth while working, the bottom 95 percent is doing ok, mainly because of Social Security.
Unlike our other forms of retirement savings, Social Security is explicitly designed to be progressive, and it is effectively providing a floor of retirement benefits and keeping retirees out of poverty. It’s possible that defined benefit or defined contribution plans could be designed more progressively. Defined benefit plans could provide a floor of benefits to all workers, for example, or award proportionately higher benefits for lower-paid workers. Defined contribution plans could be framed in terms of level-dollar amounts for each worker, as opposed to a uniform percentage rate which benefit higher-paid workers more.
But these elements are rare, and in fact the opposite is more common, where stable workers get extra rewards for their stability and transient workers lose out. In the teacher space, many states give extra perks to teachers or administrators who reach thresholds like 20 or 25 years of service. Those are another way states have made their teacher pension plans regressive, not progressive, and they're not even effective as retention tools. Worse, more than 1 million teachers don’t have access to Social Security, leaving them particularly vulnerable to poorly designed retirement plans.
All told, fewer older Americans have defined benefit pension wealth, and the poverty rate among elderly Americans has fallen dramatically. Both things can be true at once.
As we have written about before, female educators in Illinois earn on average $7,775 less than their male colleagues. This disparity in salary translates into average annual pensions that are $3,800 less valuable. In a recent blog post, I found that at least some of the salary and pension gap is derived from the fact that most female educators work in elementary schools, which have lower average salaries compared with high schools where most men work.
In this post, we will look into whether different rates of educational attainment among men and women contribute to the salary gap. This is an important question since in most school districts teachers can earn a significant salary increase, the so-called “master’s bump,” once they earn their master’s degree.
To do this, we looked at the rate of female and male educators who hold at least a master’s degree at each year of experience, and then we compared that trend to the salary data. In both cases, higher numbers correspond with a gap favoring male educators, while lower numbers represent a gap favoring female educators.
As shown in the graph below, there’s a gender salary gap (the orange line) even among educators with one year of experience, and the gap steadily increases over time, and growing to $12,567 among those with exactly 30 years of experience. The pattern on educational attainment (blue line), on the other hand, shows a different trend: women tend to hold master’s degrees at slightly higher rates at most experience levels.
Educational Attainment Rates do not appear to Influence the Gender-Based Salary Gap
Source: Author’s analysis of data from Illinois Teacher Service Record (TRS) 2012. Data adjusted for cost of living using the Comparable Wage Index.
These data show that men with six or fewer years of experience have slightly higher educational attainment, but after that the advantage goes to women. One might expect that with higher educational attainment, women’s salaries would be higher. But no, men have higher salaries at every experience level despite having worse educational credentials. The higher rate of educational attainment for women is insufficient to overcome other barriers to higher salaries, such as working disproportionately at the elementary level.
In the end, educational attainment does not explain why there is a large and persistent pay gap between male and female educators in Illinois. We will further explore other features and potential explanations of the gender-pay gap in future posts.