Splitting Up or Dancing Together? Local Institutional Structure and the Performance of Urban Areas

Abstract

This article analyzes institutional changes in local governance structures as determinants of wage premium and innovation capacity of urban areas. By combining individual and metropolitan area data for the US, we study the role of institutional fragmentation, related to the number of local governments operating in an area, and institutional coordination, stemming from the creation of authorities fostering the collaboration of local governments. Our findings suggest that more fragmented institutional landmarks do not benefit the wage competitiveness and innovativeness of urban areas. If anything, they harm them. Conversely, stronger coordination among local governments boosts the productivity of functional regions by increasing their wage premium and improving their capacity to innovate. Coordination agreements between different counties or municipalities are especially relevant in the case of urban areas modifying their functional borders over time. These findings provide key insights into the economic effects of reforming the governance structure of metropolitan areas.

Key words:

• local governance
• US MSAs
• institutional fragmentation
• institutional coordination
• wage premium
• innovation

JEL codes:

• H70
• R12
• R23
• J3

Acknowledgments

We are grateful to the editor in charge and three anonymous referees for their constructive comments. We also thank Davide Fiaschi, Alessandra Faggian, William Kerr, and all participants to seminars at the ERSA Conference in Cork, University of Pisa, CRENOS workshop in Corte for useful comments and suggestions on earlier versions of this article. We thank Sergio Petralia for sharing data on US patents. All errors are our own.

Notes

1 To make some examples, a worker may experience an increase in the earned nominal wage of about 40 percent by moving from the MSA of Abilene, TX, to San Jose, CA. In terms of innovation, MSAs such as San Jose, CA, or New York, NY, produce over one thousand patents per year, while many other metropolitan areas produce very few or no patents at all (e.g., Laredo, TX).

2 A growing number of contributions have examined the importance of the quality of regional governments, intended as the effectiveness and accountability of local administrations, lack of corruption, and rule of law. Key elements influenced by regional government quality include the economic effectiveness of public policies (Crescenzi, Di Cataldo, and Rodríguez-Pose 2016), the capacity to innovate (Rodríguez-Pose and Di Cataldo 2015), and the inclusiveness of economic and labor market strategies (Iammarino, Rodríguez-Pose, and Storper 2019).

3 We adopt wage premium to describe productivity. Wages are usually proportional to (and not equal to) labor productivity by a factor that depends on the local monopsony power of the firm.

4 Our focus is on the effect these governance features have on metropolitan outcomes, but a developed literature exists also on the root of metropolitan areas’ institutional changes. While a detailed discussion of it is beyond the scope of this article, it is worth mentioning that some authors claim that metropolitan governance is systematically prone to red tape (Trounstine 2009), or to being manipulated by powerful interests (Logan and Molotch 1987), inevitably leading to overlapping and fragmented local structures. Storper (2014) proposes interpreting the process of metropolitan governance as a large-scale principal-agent problem.

5 A different, yet related, literature has focused on the issue of mergers of municipalities. This is an operation involving the creation of new, larger local governments, replacing the smaller ones. The literature evaluating the impact of such institutional transformation reports mixed evidence (e.g., Allers and Geertsema 2016; Luca and Modrego 2020).

6 In more detail, changes in the shapes of MSAs depend primarily on the evolution of spatial patterns of urbanization. During the second half of the last century, MSA expansions were generally linked to urban sprawl into low-density residential areas (Lopez 2014). This has not necessarily implied the incorporation of poorer economic areas, since it has often been the result of selective migration of better-off people to the suburbs combined with a deterioration of economic conditions of low-income groups who remain in the city core. More recently the picture has become more blurred, since in some metropolitan areas, the continued exurban expansion has been coupled with a move back to the city (Hanlon, Short, and Vicino 2009)

7 The upward jump from 1970 to 1980 can be explained, among other things, by the fact that several new MSAs have been created in the 1970s and only appear in our sample from 1980. These new MSAs have a smaller population on average. To be precise, the sample of MSAs increases from 227 in 1970 to 297 in 1980. On average, the 227 MSAs existing in 1970 have a population of 633,240 inhabitants, while the new 70 MSAs appearing in the data set in 1980 have an average population of 145,649 inhabitants (the average in the 297 MSAs of 1980 is 586,995). This jump is accounted for in our analysis as we exploit within-MSA variability.

8 The number of COGs soared in the 1960s and 1970s as a result of federal requirements and massive increases in federal aid to state and local governments between 1957 and 1977. Most regional planning commissions were converted to COGs during this period (Vlassis 2007).

9 We exclude workers in public administration in order to reduce endogeneity issues. As a robustness test, we replicate the analysis by excluding all occupational sectors in which there may be public employees (see “Empirical Analysis”).

10 The extent of these metropolitan areas varies with decennial MSAs’ definitions, except for 2010 for which metropolitan designations are based on 1999 OMB delineations rather than on later ones. The choice derives from the fact that more recent delineations use significantly revised standards that impair comparability with earlier ones (Ruggles et al. 2020).

11 A list of metropolitan coordination bodies used in the analysis is available from the authors upon request.

12 Given that we cannot control for individuals’ unobservable characteristics, such as ability, endogeneity may be an issue. As our main estimates are performed at the MSA level, we produce statistical checks to verify how serious this issue is in our data. The estimation of event studies (see “Governance Structure and Innovation”) is made also for this purpose. We discuss this caveat further in the conclusions of the article.

13 The model is $logwage{s}_{a,t}={\delta }_{1}G{f}_{a,t}+{\delta }_{2}G{c}_{a,t}+{\delta }_3{X}_{a,t}+{\varphi }_a+{\tau }_t+u_{a,t}$. The matrix X_a,t includes a broader set of controls as compared to the second step of two-step estimates.

14 This implies that, for instance, if a COG has been created in 1969, the dummy variable will take value 1 from 1980 onward, not from 1970, assuming that the effect takes more than one year to materialize. Conversely, if the COG has been created in 1965, the dummy takes value 1 in 1970 until the end of the period. Adopting different time lags does not significantly alter the main results, as shown in Table D5 in Appendix D in the online material.

15 This implies that if, for instance, COG X has been created in 1979, the dummy variable D_X,t−1 takes value 1 in 1970 and 0 otherwise; D_X,t−2 takes value 1 in 1960 and 0 otherwise; D_X,t−1 takes value 1 in 1970 and 0 otherwise; the dummy D_X,t+0 (Gc in the graphs) takes value 1 in 1980 and 0 otherwise; the dummy D_X,t+1 takes value 1 in 1990 and 0 otherwise, and so on. In other words, here we assume the possibility of an immediate effect.

16 In the case of one lag, the dummy, for example, takes value 1 from 1970 if the COG has been created in 1969 or any other year in the 1960s; with two lags it takes value 1 from 1970 if the COG has been created in 1968 or any other year in the 1960s; with ten lags, it takes value 1 from 1970 if the COG has been created in 1960, but it takes value 1 from 1980 if it has been created any other year of the 1960s.

17 The command was implemented by considering only the never-treated units in the control group, as suggested by Sun and Abraham (2021).

18 In order to construct this variable, we exploit the exact year of establishment of Consolidated city-counties. Lagging this variable by five years, as done for COGs, leaves the results unchanged.

19 It may as well be that Consolidated city-counties have an impact within the territory for which they are responsible, but this effect is not visible when looking at the aggregate of metropolitan areas. Testing for this goes beyond the scope of this article.