Reviewing Recent Research on How Pension Reform Affects Public Sector Recruitment and Retention

Defined benefit pensions are often thought of as recruitment and retention tools for the public sector. Since retirement benefits are part of an overall compensation package, it is reasonable to assume that a change in any part of the benefit would alter the attractiveness of a job offer. However, there is a lack of any robust empirical evidence to support such a claim, as evidenced by a review of recent articles that discuss the consequences of retirement benefit changes for public service recruitment and retention.

Unfunded liabilities of many public pension plans have grown over the last decade, necessitating significant reform in most cases. But any discussion on proposed changes ought to include a thorough cost/benefit analysis. This begs for a data-driven investigation regarding pension effects on recruitment and retention to ensure policy decisions are guided by more than intuition and conventional wisdom.

Perhaps the most recent and relevant study that addresses the question directly is a recent paper published by the Boston College Center for Retirement Research (CRR) that uses an ordinary least squares (OLS) regression analysis to examine the effects of pension cuts on public sector competitiveness (the public sector’s ability to retain and recruit employees).

The authors use private sector wage changes as a measure of job attractiveness. Assuming that other factors—such as motivation for work— are not altered, they look at changes in private sector salaries during the timeframe in which the only thing that has been modified are pension benefits. Their theory holds that if workers value pension benefits, then cuts to public sector pensions will reduce the attractiveness of public sector jobs relative to those in the private sector. As a result, there will be a greater demand for private sector jobs, and workers with fewer private sector opportunities will tend to work within the public sector. Therefore, workers’ starting wages in the private sector will be lower.

They find that the private sector wage of new hires declined by 2.9 percent after cuts in pension benefits in the public sector. They conclude that this finding indicates that the public sector is having trouble recruiting the same type of employees as the private sector after pension reductions. They also look at the retention of public sector workers but find no statistically significant change in the salaries of the public sector employees who left for the private sector.

Although this research provides important insights, the underlying analysis and findings leave something to be desired in terms of explanatory power, suggesting that this research should be interpreted with caution. There are several statistical issues that are important to note:

• First, the model does not really fit the data well as evidenced by its low R-squared (0.25) – a measure of the strength of the relationships between the model and the dependent variable. This means that very little variance in the dependent variable is explained by the independent variables (in this case, wages, size of the public sector relative to the private sector, health coverage, demographic variables).
• Second, many independent variables used in the CRR analysis do not have a statistically significant effect on the dependent variable. This means that independent variables—such as health insurance coverage, the size of the public sector in proportion to the private sector, retiree coverage, and others—are not good predictors of wage change.
• Third, the model could have benefited from adding cohort or generational controls. It has been frequently cited that generations differ in their attitudes towards savings, retirement, and investment. Not controlling for generational effects could have resulted in omitted variable bias, which leads to attributing the effect of missing variables to included ones.
• A fourth, and last, critique is that the OLS regression analysis method requires strict assumptions, which are frequently violated in this analysis. Examples of such violations include missing or unavailable independent variables, the existence of influential outliers that skew the data, nonnormality of the dependent variable and others. Perhaps the CRR study could have benefited from using a non-linear type of regression in order to model this relationship and avoid such methodological issues.

Survey research offers another important type of study that can get us closer to the answer of how important of a behavioral motivator pensions really are. If we are trying to find how changes in retirement affect public jobs, what better method can one find than to ask the individuals who hold those jobs? Unfortunately, this sort of research is expensive and has to be custom-built, depending on the purposes of the study. As opposed to many empirical studies that use existing databases, surveys have to be specifically tailored to a research question at hand and are therefore difficult to come by.

An example of a study that is making such an attempt is a relatively recent paper by the National Institute on Retirement Security (NIRS) that cites its own earlier survey and comes to some seemingly thought-provoking conclusions at first glance that lack support upon further scrutiny. For example, the paper reports that “67 percent of Americans indicate they would be willing to take less in pay increases in exchange for guaranteed income in retirement.” Also, “87 percent of Americans say pensions are a good way to recruit and retain qualified teachers, police officers and firefighters.”

However, the report has significant shortcomings. For example, it does not explain the protocol of the study nor provide the survey instrument. We do not know their precise demographics, selection process, allocation strategy, etc. More importantly, the original questions of the survey are not disclosed, which makes it impossible to be used for any other study or review of its overall research value and objectivity.

Without these important parts, it is hard, if not impossible, to judge the merit and impact of the findings. For example, were the study participants told how much less in pay they will have to take in order to get bigger benefits in the future? Explaining this could make a difference in their response.

Also, how useful is the finding that 87 percent of subjects think that a pension is a good recruitment tool? Unless those people were employed or had some sort of professional or personal connection to those types of plans, their opinions are not telling us much about choices of the population that would be considering employment in public sector. (Note: Another similar survey that could be of interest to a reader is the Reason-Rupe poll conducted in 2015. In contrast to the study by NIRS, it discloses the study protocol and questionnaire.)

What could be more useful is interviewing a sample of teachers, police officers, and firefighters themselves, with a wide range of age groups represented. The study could have benefited from asking more specific questions that would have determined a marginal rate of substitution between current pay versus future benefits—the current findings suggest no sense of tradeoffs. They could have also asked current or new teachers about the relative importance of pension benefits for them when they were choosing their job, or deciding to remain in it.

And although this particular NIRS study has significant shortcomings and is unreliable, surveys could be a very useful tool for determining attitudes towards pension benefit changes. In general, it would be valuable to conduct a regular survey of new hires in the public sector regarding their motivations. This information would make it possible for employers to adjust the parts of the compensation package that are most relevant for new employees.

Another recent study published in the academic journal Educational Evaluation and Policy Analysis looked at an individual’s risk aversion as a choice predictor for retirement compensation. They compare teachers’ preferences under FAS (final average salary) defined benefit pension plans to CB (cash balance) plans. Both plans have the same expected value, but different risk. With CB plans, earnings are more evenly distributed over the whole lifecycle, whereas with FAS plans, earnings tend to be unevenly distributed with lower earnings in the beginning and upward spike towards the end of the lifecycle.

The authors find that ‘risk averse’ teachers prefer CB plans to FAS plans (the main category of defined benefit retirement plans). This type of research design is relatively new for the applied labor research field, in that it uses existing empirical data and relies on the same mathematical foundations used in traditional regression analyses, while improving upon traditional applied econometrics research by using behavioral concepts—in this case, risk aversion.

This is especially important for the retirement benefit debate because, in many cases, changes to existing plans are seen as cuts. However, even when the benefit structure is changing, the individual’s overall compensation over the whole lifecycle is not necessarily lower. It could simply be that the compensation is distributed differently across the lifecycle. This research shows that such a change is not necessarily deterring people from entering a profession. In fact, it shows that there is a pattern that predicts a preference for alternative plans that would allow teachers to earn benefits more evenly throughout their careers.

A complicating factor is that risk aversion is a variable unique to the individual; so, it would be hard to design a plan that fits everyone’s needs. This does, however, bring up an important point that it may be valuable to provide employees with a choice between the two types of plans, as we’ve seen in recent state-level pension reforms in Michigan, Arizona, Pennsylvania and elsewhere.

Conclusion

Taken together, these three recent research papers used different research approaches, all attempting to improve our understanding of the effects of changing retirement benefits on labor market participation. While they offer some interesting insights, they do not provide us with any confidence in whether pension reforms will deter or encourage talent recruitment in the public sector.

There are, however, several lessons learned from reviewing this literature:

• First, it would be useful to conduct surveys of new entrants to learn their motivations in entering the public sector, and how much retirement benefits play into those decisions. Without knowing what public sector employees think, it is much harder, although not impossible, to determine the magnitude of any potential changes to the benefit structure.
• Second, none of the recent literature investigated/examined the recent pension reforms and their effects on recruitment. We are now in a unique situation where we can observe the recruitment patterns before and after most recent reforms, which could reveal important information about the effects of any changes.
• Third, pension policy research could borrow more heavily from both the latest behavioral economics and traditional labor economics fields. Almost none of the policy papers reviewed take advantage of the vast amount of recent academic research in economics that could be applied to pension policy.

While this review is in no way exhaustive, it does suggest that there has not yet emerged a clear understanding of how changes to the retirement benefit structure can alter labor force participation in public sector employment. Accordingly, definitive statements to the contrary should be taken with a grain of salt. With the growing amounts of unfunded public pension liabilities across the country, it is important to understand and address the research gap to better understand the implications of different reform approaches.