In their influential 2009 study, Fernando Ferreira and Joseph Gyourko show that the identity of the party in power does not appear to matter for any fiscal outcomes in U.S. cities. This stands in contrast to findings from other countries where parties appear to be influential in shaping city‐level fiscal outcomes. The existing evidence suggests that in U.S. cities, Tiebout sorting (such as city residents’ voting with their feet) disciplines politicians to keep budgets balanced.
While the types of fiscal spending considered by Ferreira and Gyourko (total revenue and taxation, expenditure shares, city employment, and some crime statistics) are all highly visible to city residents, the political‐economy literature suggests fiscal profligacy will be more likely for types of expenditure where the cost is less salient or comprehensible to voters and the benefits go to a more narrowly defined constituency. My research shows that parties do matter in U.S. cities for exactly such types of fiscal expenditure, which Edward Glaeser and Giacomo Ponzetto label “shrouded benefits.”
My focus is on the biggest fiscal challenge that U.S. cities face in the coming decades: public‐sector pension obligations, many of which are severely underfunded. Pension underfunding is the result of benefit increases that are not accompanied by sufficient contribution increases. A general tendency for pension benefits to be underfunded is demonstrated in Figure 1, which plots the evolution of the average of per capita benefits and contributions over the past decades. Unlike federal social security, municipal and state pension benefits are legally binding commitments. It is therefore tomorrow’s taxpayers that are on the hook when these are unfunded. Where the funding gap is sufficiently large, it can culminate in municipal bankruptcy.
In theory, benefit increases lead actuarial accountants to recalculate the pension contributions required to finance the higher benefits. In practice, benefit increases may not result in appropriate contribution increases for any of several reasons. First, actuarially required contributions underadjust to certain benefit formula enhancements, especially those that incentivize earlier retirement and change effective retirement ages. Second, the transmission of benefit increases to actuarially required contributions can be neutralized by simultaneously increasing the actuarially assumed rate of return of pension funds, even if this rate is too high. Third, actual contributions paid can substantially fall behind the actuarially required contributions.
Pension underfunding is, at its core, a political‐economy problem: while voters may fail to account for the Ricardian equivalence of public debt more generally, this failure is more severe in the case of unfunded pension benefit increases because these are treated as if they were budget neutral. This misleading budget neutrality creates, in the words of James Buchanan and Richard Wagner, a “fiscal illusion” that makes pension benefit increases an attractive substitute to wage increases in politicians’ eyes. This is illustrated by an interview with the former mayor of Houston, Lee P. Brown. During a tight reelection campaign that he narrowly won with 51.7 percent of votes, he was instrumental in a large increase in municipal employees’ pension benefits. He later justified his decision to increase pension benefits by the fact that it was budget neutral, and that he did not “have the funds to give municipal employees the raises they deserved.”
In the long run, the municipal employer still needs to cover benefit increases through local taxes. One objection to the fiscal illusion argument is, therefore, that the budget neutrality of unfunded pensions should not matter because homebuyers capitalize future tax obligations into property values. However, the reality of the “shroudedness” of pension accounting means that even homebuyers who are aware of this future tax burden will be prone to underestimating it. That is because, aside from imperfectly observing official funding levels in local pensions, these official funding‐level figures can also be way off since the actuarial assumptions underlying them are settled in a highly politicized process. One insider summarizes this problem as follows: “consistent low‐balling of pension costs over the past two decades has made it easy for elected officials and union representatives to agree on very valuable benefits, for very much smaller current pay concessions.”
It is not clear that one should expect a partisan tilt in these political‐economy dynamics. There is no evidence that Democratic mayors are more fiscally profligate than Republican ones on other issues, and it is not clear that the Democratic Party is more fiscally profligate at any level of government. However, anecdotal evidence does suggest that Democratic politicians at the municipal level tend to be closer to (and more dependent on the political support of) public‐sector employees, and political‐economy theories that emphasize the importance of “bringing out the core” would thus predict a partisan tilt.
The primary source of data used in my analysis is the U.S. Census’s Annual Survey of Public Pensions, which, after linking to its historical versions, covers the years 1962–2016. My primary focus is on the evolution of a plan’s per capita benefit payments relative to per capita contributions. The unit of observation is a pension plan in a year; the mapping from municipal pension plans to cities is many to one because some cities have different plans for their major employee groups.
I analyze the evolution of per capita benefits relative to per capita contributions in municipal pension plans as a function of the party of the city mayor, using an extension of the mayoral election data produced by Ferreira and Gyourko. Many unobserved factors that can also influence pensions may determine whether city residents elect a Democratic or Republican mayor. To identify the effect of the mayor’s party, I therefore emphasize results from close elections; this controls for confounding factors that independently shape the outcome of interest.
The first empirical exercise is to confirm that none of the fiscal outcomes in Ferreira and Gyourko’s study respond to the party of the mayor in the slightly different data while using the slightly different measures I use. This means that the absence of partisanship holds true for all the highly visible major budget items in a city, and thus forms the benchmark of the analysis. My core finding is that changes in the political party of the mayor do, however, have a sizeable effect on the per capita benefits of a city’s pension plan. Having a Democratic mayor is associated with increases in annual per capita benefits of $2,000–$3,000 per person (expressed in constant 2010 dollars).
At the city level, pension benefits are either adjusted as part of collective bargaining or by statute. Since such adjustments take time, the core analysis studies changes in benefits four years out from the election (i.e., at the end of the narrowly won mayoral term). Varying this time horizon reveals that a statistical effect is discerned starting three years after the election, is economically and statistically strongest five years after the election, and then is less precise six years out.
Interestingly, plots of the results suggest neither a strong relationship between the Democratic Party vote share and pension benefits nor an across‐the‐board shift in pension benefits when the Democratic Party wins. Instead, the results are driven by a spike in pension benefit increases right around the close‐election cutoff. This suggests that, rather than reflecting preferences by the Democratic Party, the effect estimated by my approach primarily reflects commitments that are made to win close elections.
Investigation of other pension‐related outcomes provides evidence on the mechanisms linking short‐run benefit expansions to the long‐run erosion of funding levels: As benefits expand, the city begins to shirk part of the required contribution payments, active contribution‐paying members begin to retire, the number of pension recipients relative to paying contributors inside a plan begins to increase, and the plan’s asset base begins to erode.
As a final exercise, the data are sliced in a number of ways to provide further evidence on the political‐pork mechanism. First, slicing the data by plan type reveals that the effect is concentrated in pension plans for police and firefighters as opposed to general city employee plans. This suggests that pork barrel politics is aimed at the most well‐organized employee groups. Second, slicing the data into elections won by incumbents and challengers reveals that the effect is concentrated in elections where the challenger wins. This suggests that incumbents need to rely less on pork to win close elections.
This research brief is based on Christian Dippel, “Political Parties Do Matter in U.S. Cities … for Their Unfunded Pensions,” NBER Working Paper no. 25601, April 2019, http://www.nber.org/papers/w25601.