1. Introduction
“Although many people were important to Roosevelt’s political success, no single group of men was more important to him than the big city bosses,” according to historian Lyle Dorsett. “And in the final analysis no other man was destined to be so vital to the life and death of urban political machines as Franklin D. Roosevelt.”Footnote 1 Dorsett was not alone in describing a mutually beneficial relationship between Franklin D. Roosevelt (FDR) and urban machines. Historians have documented several accounts of urban party bosses attempting to leverage local control over federal New Deal funds to mobilize residents in support of not just local politicians but also state- and federal-level Democratic candidates.Footnote 2 In fact, during the 1930s, as journalists published stories of Works Progress Administration (WPA) workers being forced to donate their wages to Democratic campaigns,Footnote 3 it became almost a truism that the New Deal was both a social welfare program and a local-national patronage system that relied on urban machines to convert federal dollars into Democratic votes.Footnote 4
But while this argument is well-supported by archival evidence of urban bosses’ relationships with President Roosevelt and their efforts to secure him votes using New Deal funds, scholars have yet to assess whether they succeeded in doing so. This is the question that this paper takes up: to what extent did urban machines link together a local-national patronage system during the New Deal? To what extent were urban bosses able to leverage the influx of federal funds during the New Deal to mobilize their constituents in favor of the Democratic Party?
To answer these questions, we draw on federal government spending data compiled by economic historians Price V. Fishback and Shawn KantorFootnote 5 and categorizations of New Deal Democratic machines from secondary sourcesFootnote 6 to shed light on the relationship between local-, state-, and national-level patronage politics during the New Deal. We focus particularly on the case of the New Deal’s work relief programs, as they represented more than a third of federal grants made during the New Deal from 1934–1940,Footnote 7 and they were the programs that urban bosses felt were most ripe for patronage.Footnote 8
We explore the patronage relationship from both sides. First, we ask: to what extent did a patronage logic dictate the flow of work relief spending to U.S. cities and counties? Based on both a longitudinal comparison of work relief expansion in machine versus non-machine cities, and a cross-sectional analysis of the political and economic predictors of county-level work relief spending, we find no evidence that machines benefited disproportionately from these programs. Instead, the strongest predictors of work relief spending were economic need, although we also find evidence that swing counties received more money than otherwise similar counties.Footnote 9 What about the other side of the relationship? Even if urban machines did not get more than their fair share of the federal pie, bosses still had incentives to use the funds they did receive to generate support for federal Democrats: bosses did experience a surge of patronage jobs and federal funds in their cities, and they had no way of knowing that (controlling for need) they were getting no more than other, similar cities. In short, given the historical evidence, it is likely that many bosses thought federal New Dealers were holding up their side of the exchange relationship.Footnote 10
As such, we ask: as work relief programs expanded, were urban machines particularly capable of turning out Democrats in federal elections? We examine this answer with respect to both Democratic Senate candidates and FDR himself. We find clear evidence that Democratic Senators’ vote share increased in machine counties. We estimate that Senators saw a 4-percentage point larger increase in vote share in machine counties, relative to non-machine counties, for states with average growth in work relief spending between 1932 and 1936. However, FDR did not receive similar electoral benefits. Instead, in both county-level analyses of presidential vote returns and synthetic control analyses, we find little evidence that urban bosses were successful in using work relief funds to turn out voters for FDR.
These findings contribute to long-standing questions about the role of urban machines during the New Deal. Observers at the time were overwhelmingly concerned that Democratic politicians were using New Deal programs to benefit urban bosses and themselves. But unlike previous studies showing how patronage created entrenched party dominance at both the local- and state-levels in the United States,Footnote 11 we have found only limited evidence that machines were serving to connect voters on the street corner to federal Democrats using work relief programs. While these findings do not contradict compelling historical evidence that bosses like Edward Crump of Memphis worked tirelessly to elect FDR while also advocating for their piece of federal patronage, they do indicate that such politicians were not as successful in their efforts as has previously been thought.
More generally, these findings offer new considerations about the organizational and political bases of the New Deal Democratic coalition. The New Deal marked an unprecedented rearrangement of the relationship between the federal government and the American public. And yet, even as beneficiaries of the government’s novel largesse mobilized to support their own programs, there is little evidence that this mobilization drew recipients into the Democratic coalition.Footnote 12 Instead, scholars have argued that the national party required local organization to succeed: urban machines and organized labor, both of which connected Northern, urban workers, in some cities across ethnoracial lines, to the party.Footnote 13 But we find limited evidence that machines were better than similar, non-machines at serving this organizational function. This provides further indication that the New Deal realignment, and the centrality of Northern cities to its new coalition, were not produced through self-interest or quid pro quo alone—and suggests the importance of other forms of organization, including labor, in shaping urban workers’ experience of the Great Depression and the political lessons they drew from it.Footnote 14
2. Machine politics during the New Deal
Political machines in U.S. cities sprouted during Reconstruction, reached a height during the Gilded Age, and declined steadily until their eventual demise in the 1970s; but for the intervening years they existed in various forms throughout the United States.Footnote 15 Leading up to and during the New Deal, they were a semi-regular feature of American politics.
A machine is a hierarchical form of political party organization that relies on brokers to distribute goods, services, and patronage jobs in exchange for votes. Some machines were organized under one city boss who effectively controlled city politics, while others were neighborhood-based, with ward bosses sitting at the top.Footnote 16 Bosses fulfilled their side of these quid pro quo relationships because they had access to (or in the case of citywide bosses, controlled), the “local state,” which made it possible for them to award municipal, county, and party jobs to party members; pass laws and regulations that were favorable to business or criminal interests; and direct the behavior of the local police.Footnote 17
To ensure that voters upheld their end of the bargain, machines were highly centralized: from the precinct captain to the ward boss to the city-wide party leader(s), public resources poured downward to residents, and votes flowed back up again.Footnote 18 Through these relationships, people on the street corner were integrated into the party organization.Footnote 19
In such a world, parties were strongest where the economic needs were most compelling—where voters were most dependent on patronage for economic survival.Footnote 20 To maximize votes while minimizing costs, bosses sought a minimum of political integration. In cities highly segregated by ethnicity and race, machines funneled resources to specific, loyal neighborhoods, thereby drawing together a minimum winning coalition from different ethnic groups to gain and retain political office.Footnote 21
Machines were, in short, experts at using state resources to “maximize [their] electoral support”Footnote 22 or, at least, minimize their chances of losing. Because of this, scholars have argued that machines were well-suited to serving as connectors in a local-national Democratic Party machine during the New Deal, linking voters from the street corner to the presidency. Moreover, such a relationship would have been mutually beneficial: machines had an interest in using their organizations to benefit federal-level Democrats because the New Deal presented opportunities for consolidation and growth for many machines that had been in decay for decades.Footnote 23 This is because the federal government devolved control over New Deal social programs to local levels, expanding the very meaning of control over the local state: it came to include not just local resources from property taxes, fines, and assessments but also patronage jobs and federal funds from new relief programs.Footnote 24
3. Pathways to patronage in New Deal work relief programs
For this reason, local control was the key institutional feature of New Deal programs that made it possible for machine politicians to turn New Deal dollars into support for federal Democratic candidates.Footnote 25 And while all New Deal programs held the possibility for patronage—at the very least through hiring in the new federal jobs that were created to administer the programs—work relief programs were of the greatest interest to machines because they also provided millions of relief jobs.Footnote 26 As such, this section describes the opportunities for patronage in New Deal work relief programs as the locus of power over spending and patronage jobs shifted across different levels of government. These details of program design, as well as historical evidence of machines using work relief as patronage, lead us to several predictions about how machines served as connectors in a local-national quid pro quo within these programs.
3.1. FERA and the CWA, 1932–1935
The New Deal’s foray into work relief began on July 1, 1933, when Congress approved legislation that created the Federal Emergency Relief Administration (FERA) and authorized $500,000,000 to be made available for federal grants-in-aid to states. Harry Hopkins, the head of FERA, initially intended for funds to go primarily to help the “employable” rather than the “chronically dependent,” but the guidelines were sufficiently open to interpretation that funds were typically divided into two categories: work relief and direct relief.Footnote 27 FERA was, as its name suggests, an emergency measure designed to mitigate the worst harms of the Great Depression. As such, it relied on long-standing local relief apparatuses to distribute funds and determine applicants’ needs.Footnote 28
In an effort to curtail patronage, Hopkins deliberately bypassed Congress in administering FERA: instead, it was governors, including Republicans, who chose the state-level administrators that directed funds within the state. This decision rankled Democrats from county party organizations all the way up the Senate, many of whom wanted a piece of this new federal patronage and complained about Hopkins to James Farley, Chair of the Democratic National Committee.Footnote 29 Among several disgruntled Senators, Harry S. Truman, then a Senator from Missouri, wrote to Farley, “when the patronage was handed out the people who control things in the party in this State were not recognized … it is rather discouraging to say the least.”Footnote 30
Examples abound of Hopkins’ commitment to depoliticizing FERA. One of the few mechanisms that FERA had for controlling state and local relief agencies was granted in Section 3(b) of the 1933 authorization act, which allowed Hopkins to withhold grants or federalize the administration of state programs if he thought this would provide “more effective and efficient cooperation between the state and Federal authorities.”Footnote 31 In March 1935, FDR used this authority to grant Hopkins control of Ohio’s FERA administration, due to “evidence concerning corrupt political interference with relief.”Footnote 32 This decision was made despite the fact that Ohio’s Governor Davey was a Democrat who at least tepidly supported the New Deal, and later lost his re-election bid to a Republican.Footnote 33
And yet, despite Hopkins’ efforts to ensure professional, apolitical administration of FERA, it was the local relief agencies who were responsible for determining need, investigating cases, disbursing money, and recommending work projects.Footnote 34 This devolution ultimately provided ample opportunity for racial and gender discrimination as well as patronage.Footnote 35
Shortly after FERA began, Hopkins sought FDR’s approval to divert funds from the Public Works Administration, which was slow to start its large-scale infrastructure projects, to a new organization, the Civilian Works Administration (CWA), which would take on shorter-term construction projects now housed under the work relief components of FERA. Although the CWA was a purely federal program in that workers received pay directly from the federal government without the states acting as financial intermediaries, its administration and staffing was identical to FERA’s.Footnote 36 As such, it faced many of the same challenges—among them, the intransigency of local relief agencies, their discriminatory practices, and their ties with local politicians.Footnote 37 The CWA federal office received thousands of letters complaining of political favoritism.Footnote 38 Lorena Hickok, a journalist who served as one of Hopkins’ six envoys traveling across the country, reported on the graft in the local operations of the CWA. When visiting Tampa, a machine city dominated by a White supremacist party organization, she reported, “So far as CWA and relief are concerned, this state seems to be chock full of politics and petty graft. It seems to be worst in Tampa, which, I am told, could teach even New York City something about political control.”Footnote 39
While Hopkins’ own intentions in both FERA and the CWA were geared toward thwarting patronage, his efforts were at odds with those of various Democratic Senators, Congressional Representatives, and James Farley. But even as several Democrats fought for control over spending and hiring for New Deal programs, they all had different interests: urban bosses wanted local control of both the relief roles and the growing number of supervisorial and administrative positions for the new relief programs; Senators wanted to impose their powers of advice and consent over state-level administrators of New Deal programs; Congressional Representatives, depending on where they lived and to whom their loyalties lay, often took opposing stances.Footnote 40
3.2. The WPA, 1936–1940
The administrative structure of the WPA, which replaced the work relief components of the CWA and FERA beginning in 1936, was a compromise among many of these competing claims. FDR was authorized to establish the WPA under the Emergency Relief Appropriation Act of 1935. Unlike FERA, the WPA located funding discretion within the federal government.Footnote 41 But Hopkins’ crusade for a fully nationalized and apolitical administration was only partially successful. City, county, and state governments, as well as local relief agencies, remained essential to its administration. Local governments were the primary sponsors of WPA projects: in 1940, about half of sponsors were city, township, and county governments.Footnote 42 Those sponsors submitted proposals to their state WPA, which then sought approval from Washington, and if that approval was granted, the state WPA helped develop plans for how to administer the project. Local relief agencies were also responsible for determining applicants’ eligibility for WPA jobs.Footnote 43
But it was the WPA itself that set the states' employment quotas—a process described as shrouded in “mysticism” by Donald Howard in his early-1940s assessment of the WPA for the Russell Sage Foundation.Footnote 44 WPA administrators offered different formulas to Congress over the years—most notably, in 1940, then-administrator Colonel Harrington indicated during a hearing held by the U.S. House of Representatives that the agency employed a 40-40-20 formula, in which 40 percent of the quota was determined by a state's population, 40 percent by their unemployment, and 20 percent left to WPA discretion based on recommendations from regional and state administrators. This formula, apparently new that year, appeared to have been modified by 1941 and, when probed by Howard in 1942, “high federal officials … declared that it was not being used.”Footnote 45
There is no doubt that there was greater federal discretion over the distribution of WPA spending than what was provided for under FERA, CWA, and the Social Security Act’s means-tested programs. But scholars have offered competing interpretations of what this meant for patronage politics. For example, Fishback et al. and Wallis et al. argue that federal control over job quotas in each state gave Hopkins greater authority to limit patronage.Footnote 46 That Hopkins was typically opposed to patronage infiltrating his agencies is relatively undisputed, but the idea that federal control meant greater control for Hopkins himself is somewhat misleading.Footnote 47 In reality, the interests of various federal actors were multiple and varied, as described above.
In fact, those Democratic Senators who had complained that the New Deal was not abiding by the principle of “to the victor goes the spoils” saw greater federal control of WPA as a victory.Footnote 48 In particular, they fought for and won a role for the Senate in approving appointments of state- and regional-level administrators, as well as anyone earning more than $5,000 per year for the agency.Footnote 49 This provided a direct route for Democratic Senators to control the flow of patronage jobs within their states because the WPA leaders played a key role in endorsing projects, determining the number of workers each project needed, and deciding how to staff the administrative personnel of the agency.Footnote 50
That Senate confirmation of WPA administrators was a route to patronage politics was well-understood at the time.Footnote 51 The WPA was reauthorized annually in a series of Emergency Relief Appropriation Acts, and after Senate confirmation was “inadvertently” left out of the 1938 act, it was reintroduced in 1939 amidst vehement opposition in the House. As Representative Woodrum of Virginia, a Democrat with an increasingly oppositional stance toward the New DealFootnote 52 stated in a floor speech in June 1939, Senate confirmation would “throw the WPA right square into the middle of local politics in every State in the Union.… There is considerable opinion to the effect that when we require State WPA administrators to have political endorsements for appointment we aggravate the political interference in the program. I certainly think that.”Footnote 53
Although Senators saw the WPA as a victory for patronage, it also gave expanded control to the Executive Branch: it was the administrators, appointed by FDR, who determined the “mystic” employment allocation formulas. Whereas Hopkins under FERA had to federalize a state’s program to take control, WPA administrators could shuffle funds from state to state without much explanation—a fact that bolstered claims about FDR using the WPA for political opportunism.Footnote 54
As such, the WPA created for the first time a direct and significant financial link between urban bosses—who could propose projects to state-level administrators—and the Executive Branch, which housed the WPA federal office that would ultimately approve those projects. As Erie writes of this novelty: “For the bosses, the WPA represented a unique public works grant-in-aid— a federally financed, locally picked labor supply for machine initiated neighborhood projects…By judicious choice of the type, location, and even timing of WPA projects, machines could build neighborhood electoral support.”Footnote 55
4. Machines as links in a local-national patronage system
The historical record thus makes clear that Democratic politicians advocated for various administrative changes—a combination of local influence and Senate confirmation—that would allow them to leverage New Deal work relief dollars for patronage purposes. In part for these reasons, and of course for purely political purposes, New Deal work relief programs were subject to relentless charges of “playing politics.” And importantly, critics did not just argue that these programs were propping up local Democratic machines; they also believed they benefited FDR himself.
By 1938, that criticism—aimed primarily at the WPA—had reached a peak, as observers noted a massive uptick in the employment quotas for key states in September and October of 1936 and 1938, just before the elections.Footnote 56 Although the FDR administration offered several rebuttals and explanations for these apparent inconsistencies, observers cast doubts on the president’s intentions for a variety of reasons, including the administration’s failure to demystify state employment quotas and the Senate’s role in WPA administrators’ confirmation. As journalist Stanley High, writing in Current History in 1939, proclaimed of Senatorial confirmation: “Through that ample opening the bosses moved in. For most of the four years since, the WPA has been their happy hunting ground. From the White House down to the precinct captain, this preserve has been protected…”Footnote 57 Journalists like High indicated that they knew exactly how those extra, federally funded jobs in the fall of 1936 and 1938 had benefited FDR and Democratic candidates up-and-down the ballot: interviews with angry WPA workers suggested they were forced to donate portions of their paychecks to the local Democratic Party, spend their “work hours” for campaign-related activities, and even erect billboard “shrines” and roadside advertisements for FDR and other Democrats.Footnote 58, Footnote 59
Drawing on these accounts, as well as letters to James Farley, Harry Hopkins, and the various New Deal agencies, historians have described a pattern of patronage that led straight from the federal government, through states, and all the way to the precinct captains, then back up again.Footnote 60 As Erie notes, FDR ran a risk in trying to use work relief funds to his political advantage—negative coverage like High’s could dissuade non-WPA workers from voting for the president. For this reason, Erie argues that “Partisan conversion of the WPA required discreet middlemen—the bosses.”Footnote 61 These hierarchical party organizations were the ideal mediators between federal New Dealers and work relief recipients, relying on an existing apparatus of precinct committee members and ward leaders to convert work relief into votes. For example, amidst the rise of the new Democratic machine in Pittsburgh, one committee member explained his job: “The secret of a good committeeman is to help people, grab the opportunity and help.… In that manner you obligate the voters […] It can work in reverse. One family voted against FDR because they disliked me!”Footnote 62 In other words, local party members saw their role as binding urban voters to not just the local, but the national Democratic Party, and they used federal patronage to accomplish this. Such accounts are not limited to Pittsburgh, but also appear in historical investigations of well-known machine cities like Chicago, Kansas City, Memphis, and Jersey City.Footnote 63
This discussion leads to three predictions about the role of urban machines in sustaining a local-national quid pro quo using New Deal work relief programs. (1) First, despite Hopkins' efforts during FERA and the CWA, all work relief programs provided some opportunity for patronage. We therefore expect that machines received more work relief spending per capita than otherwise similar non-machine cities. But machines had no way of verifying whether they received more than their fair share of these New Deal programs, meaning that they were unable to hold federal New Dealers accountable if they reneged on their side of the quid pro quo. Moreover, bosses’ experience of work relief programs was characterized by an influx of federal work relief funds, which scholars argue were large enough in absolute terms to instigate a local-national patronage exchange, regardless of their size relative to non-machine cities.Footnote 64 As such, we expect that, given the same increases in state-level work relief expenditures per capita, machines were better able than non-machine cities to leverage those funds to generate votes for (2) Democratic Senatorial candidates, who were sometimes the key links between urban machines and FDR and wielded important power over the distribution of WPA funds, and (3) President Roosevelt.
5. Data sources and variable construction
To test these expectations, we draw on several different sources of data on voting, urban machines, and federal government expenditures during the period of the New Deal through the onset of World War II (1932–1940). We carry out four main analyses at different levels of aggregation to arrive at our conclusions. First, we examine whether machine cities saw larger increases in work relief expenditures than non-machine cities from 1933–1940. Second, we estimate the predictors of total county-level work relief spending from 1933–1939. Next, we look at the other side of the quid pro quo relationship. Our third analysis is a two-way fixed effects model, comparing Democratic vote share in machine and non-machine counties, in states with the same increases in work relief spending. We find that machines were able to turn out their constituents for Democratic Senators, but not for FDR. Given that this latter finding does not conform to the expectations laid out above, our final analysis tests the robustness of this null result through synthetic control analyses using Chicago and Pittsburgh as cases. The historical record consistently argues that both machines used New Deal dollars to build their operations and turn out votes for FDR.Footnote 65 And yet, these analyses verify our earlier findings—neither city produced a higher Democratic vote share after 1936 than the comparable synthetic control made up of non-machine cities.
5.1. Machine cities/counties
Several political scientists, historians, and economists have offered their own definitions of urban machines during the nineteenth and twentieth centuries, often including the New Deal era. Following Trounstine, we began our search by limiting the scope to the 100 largest cities in 1930, according to Gibson.Footnote 66 To determine whether or not a city contained a Democratic machine during the New Deal era, we then consulted several secondary sources that had already compiled similar lists.Footnote 67 Trounstine’s Political Monopolies in American Cities provided the most detailed and extensive account, including dates when machines had consolidated into political monopolies and had sufficiently biased electoral rules in their favor, as well as dates for machine demise. However, because we do not require that a machine had become a local monopoly according to Trounstine’s definition, we also consider five other sources that cover multiple machines: Boulay and DiGaetano’s accounting of the timing and demise of urban machines; Dorsett’s detailed treatment of seven machines during the New Deal; Erie’s work on Irish political machines; Mayhew’s discussion of machine organizations during the reform era and the New Deal; and Menes’ work on machines during the early twentieth century. If two of these sources found evidence that a machine was active during two or more of FDR’s presidential elections and run by Democrats, we took their definition. If only one source noted a machine, or if there was disagreement among the sources, we consulted additional secondary sources and newspaper articles to independently verify these definitions. These additional sources were most often case studies of state- or city-level political developments, usually produced for state historical associations. Table 1 provides a list of the machine cities included in our analyses, the sources we used to identify them, and further information about how our list compares to other secondary sources.
Table 1. New Deal Machines, 1932–1940
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20240829034811717-0857:S0898588X23000081:S0898588X23000081_tab1.png?pub-status=live)
Note: We include Utica as a machine city, although only Menes identifies it as a Republican machine. Bean (1994) refers to Utica as a machine that switched to the Democrats in the late 1920s. Utica’s machine turned out voters for FDR’s gubernatorial campaign in 1928 and received patronage from him when he was governor. With respect to San Antonio, Trounstine considers the machine monopoly as ending in 1939, the year when the machine mayor (Quinn) lost election to a reform candidate; but because Quinn returned to office in 1941, Booth and Johnson argue that the machine staved off real reform until the late-1940s.
a Members of the Denver and Tampa party organizations had ties to the Ku Klux Klan and other white supremacist groups, and the San Antonio bosses worked to suppress and channel a growing Mexican American voting bloc. As such, local control in these cities served the dual purpose of perpetuating the racially unequal distribution of New Deal resources and maintaining machine power.
Boston was the only case in which we had to adjudicate disagreements among our main sources as to whether a city had a political machine. While several treatments of urban machines during the New Deal include Boston (Dorsett, Erie, Boulay and DiGaetano), there is debate as to whether Mayor James Michael Curley was a true “boss.” Much of this debate centers around Curley’s external relationships: because FDR did not trust Curley’s ability to bring him votes, he apparently cut Curley out of New Deal patronage and worked instead with Massachusetts’ Democratic Senator.Footnote 68 As such, we include Boston so as not to sample on our dependent variables (whether Boston received more work relief and whether the local party used it to benefit FDR and other New Dealers). Moreover, Curley appears to have used city jobs to build the local party apparatus of a typical machine.Footnote 69 As an additional robustness check, we find that removing Boston from our list of machine cities does not change results.
As the table indicates, it is possible that several machines began or ended during the New Deal; but given the difficulty of determining the precise start or end dates of complex party organizations, our “machine” measure is a time-invariant binary indicator, taking the value of 1 if a city/county is listed in Table 1, and 0 otherwise. In Table A5 of the Appendix, we replicate our longitudinal analyses using a time-varying definition of machines, based on the start and end dates shown in Table 1, and find nearly identical results.
Although city and county boundaries rarely overlap directly, we believe that analyses using the county-level indicator is a reasonable conceptualization of the machine as a political organization because, while machines were often centered on mayors in major cities, they were typically linked closely to the lowest-level of political party operation in the United States—that is, the county political party. For example, as Dorsett and Erie describe, some of the most infamous urban machines of the New Deal era were really county-wide operations, with power located in the hands of the mayor, the chair of the County Democratic Party, or shared between the two.Footnote 70 This includes the Pendergast machine in Kansas City/Jackson County, MO; the Kelly–Nash machine in Chicago/Cook County, IL; the Crump machine in Memphis/Shelby County, TN; the Hague machine in Jersey City/Hudson County, NJ; the combination of the Pawtucket and Providence machines in Providence County, RI; the San Antonio machine in Bexar County; the O’Connell machine in Albany/Albany County, NY; and the Lawrence/Guffey machine in Pittsburgh/Allegheny County. It was often through these county-level parties that machines were linked to state-wide political parties and wielded power in state-level nominating conventions. As such, the county is an important and useful level of analysis.
5.2. New Deal expenditures
To construct measures of New Deal expenditures, we draw on federal government spending data at the city-, county-, and state-levels. All data are from economic historians Price V. Fishback and Shawn Kantor who digitized records from the U.S. Office of Government Reports (OGR) showing total county-level federal spending on New Deal programs, aggregated and averaged over the period 1933-1939, and state-level federal spending for each state and year from 1918-1941.Footnote 71 The OGR did not report county-by-year data. Fishback and Kantor also digitized records on public and private spending in 116 U.S. cities over the period 1929-1940.Footnote 72 Each source contains data on the work relief programs of interest to us: the FERA, Civil Works Administration (CWA), and WPA. City-level data further breaks down the direct relief and work relief components of FERA spending, such that we are able to remove the funds that were directed toward the “chronically dependent” for sustenance rather than for work.
Unfortunately, none of these data are perfectly suited to understanding the relationship between New Deal expenditures, urban machines, and Democratic mobilization. Ideally, we would have county-by-year data for the entire country, spanning at least the 1932–1940 time period. Short of this, we triangulate among all three data sets—state- and city-level data for the full period and county-level data for only a portion—to paint the most detailed picture possible of these relationships. This is particularly important because previous studies of New Deal spending have reached contradictory conclusions regarding the economic or political logic of how money was distributed, based on whether they examined county- or state-level data.Footnote 73
6. Analytic strategy
We undertake four analyses to answer our two research questions. We begin by exploring whether there is support for the idea that urban machines received more federal patronage than other cities. To do so, we first assess whether the rise in work relief spending per capita was greater in machine cities versus non-machine cities using the longitudinal data on public spending in 116 cities. We show that machines did see larger increases in work relief spending than non-machine cities between 1933 and 1936, but first differences analyses show that this difference was not statistically distinguishable from zero.
While these longitudinal data are rich and detailed, they do not allow us to assess what did predict the local distribution of work relief, and what role (if any) politics played in that distribution. To do so, we next predict county-level per capita spending across all work relief programs (FERA, CWA, and WPA) from 1933 to 1939 according to the following model:
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20240829034811717-0857:S0898588X23000081:S0898588X23000081_eqn1.png?pub-status=live)
Where ${\text{WorkReliefPerCapit}}{{\text{a}}_{\it t}}$ is the average total work relief expenditure in a county over the 1933–1939 period;
${{{\alpha }}_{\it r}}$ represents regional fixed effects; and
${\epsilon _{\textit i}}$ represents the idiosyncratic error term. Although we attempted to measure the direct relationship between machine counties and work relief spending, we are limited by a small number of observations in this cross-sectional analysis and thus lack statistical power. As such, we use this model to provide evidence as to the underlying political and economic logics of work relief spending. The vector
${\it{Politic{s_t}}}$ includes several indicators drawn from existing literature. When it comes to government programs like New Deal work relief, which lack established, public rules for the distribution of funds (“nonprogrammatic” distribution),Footnote 74 scholars have identified several ways that politics might affect “who gets what:” first, FDR and Congressional Democrats may have used their influence over the new programs to reward reliably Democratic counties, home to their “core” voters;Footnote 75 second, they may instead have targeted swing voters, trying to tip the balance and persuade them to vote for Democrats;Footnote 76 and finally, given FDR’s control over WPA appointments and the opaque allocation work relief resources, FDR may have sought to use those resources to benefit members of his own party, strengthening the Democratic ticket down the ballot.Footnote 77 As such, our political indicators include: the average two-party vote share for Senate, governor, and president during all elections from 1922–1928 (a measure of Democratic support) and the absolute value of the difference between Democratic support and 50 percent (a measure of the extent to which it is a “swing” county”);Footnote 78 an indicator that tallies each county’s total number of Democratic Party representatives in the House of Representatives, Senate, and Governor;Footnote 79 and an adaptation of Ansolabehere, Gerber, and Snyder’s measure of malapportionment at the state-house level to the U.S. Congressional level.Footnote 80 We further include a variety of economic indicators from Fishback et al. and Wallis et al., represented by vector
${\it Economic{s_t}}$, that may have dictated work relief spending, as well as several demographic variables from the 1930 and 1940 Census.Footnote 81 Table A1 of the Appendix shows descriptive statistics for all political, economic, and demographic covariates.
After evaluating whether machines received more work relief funds than non-machine cities, and whether a political or economic logic dictated the flow of work relief spending, we turn to our second question: did machines generate more votes for federal Democratic candidates, given the same increases in work relief funds? To answer this question, we first compare changes in Democratic vote share in both senatorial and presidential elections, between states with more and less work relief spending per capita, and between counties that are or are not dominated by a political machine. Given the timing of New Deal work relief programs, which began in 1933, our analysis extends from 1932-1940. More specifically, we estimate the following model:
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20240829034811717-0857:S0898588X23000081:S0898588X23000081_eqn2.png?pub-status=live)
Where ${\text{DemVoteShar}}{{\text{e}}_{it}}{ }\,$is the Democratic vote share for president or Senate in county i in year t, from the ICPSR. We focus on these two branches of government because, as described above, Senators were seen as key mediators in the process of local-national political patronage—both for their own benefit, and for the benefit of the president.
${{{\alpha }}_{\it i}}$ represents county fixed effects, which control for observed and unobserved time-invariant differences across counties;
${{{\gamma }}_{\it t}}$ represents year fixed effects, which account for any shocks particular to certain election years; and
${{\text{p}}_r}$ represents region fixed effects for each of the nine Census-designated regions.
${\epsilon _{\it it}}$ represents the idiosyncratic error term, which is clustered at the state level.
${\text{MachineCount}}{{\text{y}}_{\it i}}$ is a binary variable taking a 1 if a county contained a machine, 0 otherwise. Because this is a time-invariant measure, we can only estimate the partial interaction.
${\text{WorkReliefPerCa}}{{\text{p}}_{\it st}}$ is the amount of federal work relief spending per capita in state s in year t. The work relief programs operative in each year vary: for the presidential vote share models, the measure includes only WPA spending because it was the only work relief program in place during presidential elections; and for the Senate models it is a combination of FERA and CWA prior to 1936 and just the WPA thereafter. Our estimate of interest is
${\beta _2}$, which we can interpret as the average difference in the change in Democratic vote share, between machine and non-machine counties with the same change in state-level work relief per capita.
To interpret ${\beta _2}$ as an unbiased estimate of this difference, we assume that, if there were any state-level changes from 1932 to 1940 that affected Democratic voting other than work relief spending, those changes affected voting behavior in machine- and non-machine cities similarly. If, for example, states that saw the largest increases in work relief per capita also saw the largest increases in unemployment, and that unemployment was concentrated in machine rather than non-machine cities in those states, and those needy voters were turning toward the Democratic Party for reasons apart from work relief and machine organization, this could violate our modified parallel trends assumption. Although we observe only small differences in our measures of economic need across machine and non-machine counties (per Table A1), we address these concerns in several ways. First, our preferred model includes several time-varying covariates at the county level, shown above in vector
${X_{\it it}}$; a state-level measure of all other New Deal grants per capita, by year; and fixed-effects for the Census' nine regions.Footnote 82 Table A1 of the Appendix shows the descriptive statistics for the machine indicators, outcome variables, and controls. Second, we replicate Model (2) with time-varying state trends, in Table A6, and find substantively similar results.
And finally, our last analysis provides a robustness check for the findings from Model (2). In particular, we probe a set of surprising null results, namely, that machine counties did not generate more votes for FDR than non-machine counties. We consider two case studies where we have a particularly strong expectation of finding evidence that machines used New Deal work relief funds to support FDR: Pittsburgh and Chicago. This is because both Democratic machines emerged during the New Deal, and both new bosses received favors from FDR because FDR believed they were providing him votes. As such, if there were any machines that might have succeeded in using New Deal dollars to generate votes for FDR, it was them. To evaluate whether this is the case, we created synthetic controls, drawing on demographic, political, and spending data from non-machine counties to create estimated counterfactuals of what the voting trajectories of Chicago and Pittsburgh would have been had they not consolidated machines in 1936.Footnote 83 The synthetic control in each case is the “weighted average of all potential comparison units that best resembles the characteristics of the case of interest.”Footnote 84 In this way, synthetic control analyses are similar to matching where a treated unit (a machine county) is matched with an untreated unit (a non-machine county). Synthetic control methods are particularly useful in small N analyses because they can reveal whether exemplary cases, like Pittsburgh and Chicago, perform as our two-way fixed effects model predicts. Here, we create the synthetic controls using several indicators from before 1936, the year that Chicago and Pittsburgh each consolidated machines: 1930 county-level demographics (the same used in the models in Table 2); Democratic presidential vote share from 1928 and 1932; and average work relief spending per capita from 1933-1935 (all of the “pre-treatment” years). For the last piece, we matched the city-level public spending data from 1933-1936 to counties. This is not a perfect match, but as noted above, there is no county-by-year data on New Deal spending; as such, we rely on the city-level data to capture most of the county-level variation in work relief spending during the period. That said, we also ran the analyses leaving out the spending data and found nearly identical results.
Table 2. Work Relief Spending and Democratic Vote Share in Machine Versus Non-Machine Counties, 1932–1940
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20240829034811717-0857:S0898588X23000081:S0898588X23000081_tab2.png?pub-status=live)
Note: 95% confidence interval shown in parentheses. Standard errors are clustered at the state level
* p < 0.05; **p < 0.01; ***p < 0.001.
7. Findings
7.1. Patronage from the Top-Down: Work relief spending and Urban Machines
We begin by examining work relief spending in machine versus non-machine cities. Drawing on the city-level public spending data available for 77 of the largest 100 cities, Figure 1 charts the differences in per capita expenditures from the work relief programs of FERA, the CWA, and the WPA in machine and non-machine cities between 1933-1940. Although the sample does not include two machine cities (Camden, NJ and Tampa, FL), the data provide the most granular view of public spending in cities during the New Deal. As such, the figure provides initial indications of whether urban machines received more federal patronage than other cities.
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20240829034811717-0857:S0898588X23000081:S0898588X23000081_fig1.png?pub-status=live)
Figure 1. New Deal Spending Per Capita, Machine Versus Non-Machine Cities.
As we can see, while machine and non-machine cities received almost the same amount of federal funding per capita during the FERA and CWA years of work relief (1933-1935), this changed with the beginning of the WPA: in total between 1933-1936, work relief spending grew by almost $2 more per capita in machine cities versus non-machine cities. This is consistent with complaints lodged against the FDR administration at the time.Footnote 85 But as we show in Table A2 of the Appendix, these differences in work relief expansion between machine and non-machine cities are not statistically significant. Moreover, machine cities saw larger declines in spending between 1936-1940, such that when we examine the entire period, 1933-1940, work relief actually expanded slightly less in machine cities than in non-machine cities.
But the city-level data shown in Figure 1 do not allow us to observe the factors that did predict work relief spending because we do not have economic and demographic data at the city-level; as such, it is still possible that political logics were dictating spending more than economic need. To assess this, we predict county-level work relief expenditures averaged over the 1933–1939 period, for the counties containing the 100 largest cities, using the political, economic, and demographic indicators described above in Model (1). Figure 2 reports the statistically significant coefficients from the model, and Table A3 of the Appendix reports the full regression results.
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20240829034811717-0857:S0898588X23000081:S0898588X23000081_fig2.png?pub-status=live)
Figure 2. Predictors of New Deal Spending Per Capita.
As Figure 2 shows, work relief spending was dictated largely by a needs-based logic: it was greatest where unemployment was high in 1930 and increased most from 1930–1940. These findings largely corroborate Fishback, Kantor, and Wallis’ results.Footnote 86 The only other explanatory variable that reaches significance is our measure of electoral competitiveness: it appears that, rather than targeting Democratic counties in an attempt to mobilize core supporters, New Dealers targeted swing counties in an effort to persuade voters to their side.Footnote 87 This points away from machines as the main beneficiaries of work relief spending: they had much higher rates of prior Democratic Party voting but lower measures of electoral competitiveness than non-machine counties (per Table A1). In Figure A1 we plot the model residuals, which provide further evidence that there is no apparent, undetected relationship between machines and work relief spending that is not explained by political and economic differences. In short, while machines did receive large influxes of work relief spending because of their high levels of economic need, these results provide little evidence that they received more than their fair share of the work relief pie.
7.2. Votes from the Bottom-Up: Work relief spending and democratic vote share
But urban bosses had no way of verifying whether this was the case. Their experience, as recounted by Dorsett, Erie, and Mayhew, was an incredible influx of federal work relief funds and patronage positions to fill. As such, even if they were not receiving more than their fair share of the pie, they may still have been using that influx to assist federal Democrats based on their assumption that federal Democrats were assisting them. So, were they better than other county-level parties at using work relief funds to marshal votes for federal Democratic candidates? Figure 3 provides a prima facie view of this. Drawing on our categorization of machines and county-level voting data from 1924-1940, it charts the differences in Democratic Party vote share in machine and non-machine counties containing the 100 largest cities for presidential elections (panel (a)) and for Senate elections (panel (b)). The dotted vertical line marks the year FDR was first elected, in 1932. As the figure shows, without considering the effect of New Deal spending and other confounding variables, machine counties were, on average, more Democratic than non-machine counties in presidential elections—and this was true for the most part both before and after the New Deal. In presidential elections, the difference between machine and non-machine counties appears to hold constant throughout period, while in Senate elections the difference fluctuates over time but appears to have grown larger, on average, from 1932 onward (with the exception of 1936).
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20240829034811717-0857:S0898588X23000081:S0898588X23000081_fig3.png?pub-status=live)
Figure 3. Democratic Party Voting, Machine Versus Non-Machine Counties. (a) President, (b) Senate.
To assess whether these changes were the result of machines' particular ability to marshal Democratic votes using work relief funds, we turn to the results from Model (2), which compares Democratic voter share over time in senatorial and presidential elections, between states with more and less work relief spending per capita, and also between counties that are/are not dominated by a political machine. Table 2 shows the results of these analyses. The models predict county-level Democratic vote share in presidential elections (columns 1 and 2) and Senate elections (columns 3 and 4) for counties containing the 100 largest cities. All models include county and year fixed effects, and models in columns (2) and (4) also include fixed effects for region and the time-varying covariates described above. For the sake of readability, we report only the coefficients of interest on works grants and works grants x machine county in the main text. Table A4 of the Appendix reports estimates for the full set of coefficients.
Columns (1) and (2) provide little evidence that machines were able to use the expansion of work relief programs to increase FDR’s vote share better than non-machine counties. In fact, the negative point estimates suggest that, if anything, machines were worse at using work relief to funnel votes to FDR than non-machine counties (although this estimate is only statistically significant at the 10 percent level). The opposite is true for Senate races: here, as columns (3) and (4) indicate, machine counties saw larger increases in Democratic vote share for Senate than non-machine counties, given similar levels of state work relief expansion. Moreover, machine versus non-machine differences in Democratic vote share for Senate races were substantial. In states with an average increase in state work relief expenditures per capita between 1932 and 1936, machines saw a 4-percentage point larger increase in Democratic vote share for Senators, relative to their non-machine counterparts. In Table A6, we include state-level time trends and reach the same substantive conclusions; however, with very few degrees of freedom the coefficient for works grants x machine county is only statistically significant at p < 0.1. Moreover, as expected, we only observe these results for work relief programs: when we replicate Model (2) with all other New Deal grants instead of work relief grants, we find that machines were unable to increase Democratic vote share for either FDR or Democratic Senators.
7.3. Case studies: Chicago and Pittsburgh
Taken together, our findings thus far only partially corroborate the account of urban machines using New Deal work relief programs to link together a local-national patronage system within the Democratic Party: while they were better than non-machine counties in using federal funds to increase Democratic Senate candidates' vote shares, we have not found evidence that they were able to do the same for FDR. Given that this last result contradicts our expectations, we probe this finding further below. In particular, given the small group of machine counties and the relatively large standard errors on our estimates for presidential elections in Table 2, it is possible that machines were working successfully to FDR’s benefit, but we are unable to detect this with our data. As such, we turn to two synthetic control analyses of cases where we have a particularly strong expectation of finding “machine effects” on Democratic presidential vote share: Pittsburgh and Chicago are both Democratic machines that emerged during the New Deal. And according to existing accounts, these new machines—the David Lawrence/Joe Guffey machine in Pittsburgh and the Ed Kelly/Patrick Nash machine in Chicago—were able to consolidate because they received favors from FDR who, in turn, believed they were providing him votes.Footnote 88
Journalists Alsop and Kintner produced an account of Senator Joe Guffey’s rising political star for The Saturday Evening Post in 1938. According to them, Guffey—along with his ally David Lawrence, who was chair of the Allegheny County Democrats—built his Democratic machine in Pittsburgh almost entirely on the back of the New Deal. Guffey, they explained, had no real policy preferences beyond an ingrained sense of loyalty to the Democratic Party and an ingrained belief in quid pro quo. And when it came to the New Deal, he saw an opportunity to build a Democratic machine in a long-standing Republican state.Footnote 89 As Alsop and Kintner wrote in the second piece in their series on Guffey’s politics:
He made the President’s gratitude for the band wagon’s start pay dividends in patronage and pap. He used the patronage and pap to build the machine. And then he gathered in the converts of the New Deal, trained them, organized them and used his machine to vote them […] The captain and his committeemen, who are the serfs of Guffeyland, cultivate their precinct with great industry, hobnobbing with the voters, distributing small favors, and, at election times, leading their people to the polls.Footnote 90
Perhaps most consequential for national politics was Guffey’s effort, working alongside Robert Vann, the editor of the Pittsburgh Courier, one of the largest Black newspapers in the country, to bring Pittsburgh’s Black voters into the Democratic fold.Footnote 91 In return Guffey received control over the distribution of key patronage jobs for his clients.Footnote 92
There is evidence that a similar process took place within Chicago’s Democratic machine. As historian Rita Gordon notes, Black migrants to Chicago in the 1920s found an allegiance to the national Republican Party reinforced by a local Republican boss, “Big Bill” Thompson, who courted their votes by supplying patronage.Footnote 93 Even amidst the depths of the Depression, Chicago’s Black voters stuck with Hoover in 1932. But when Anton Cermak rode the national Democratic wave to the Chicago mayor’s office in 1932, and was then replaced by Edward Kelly after his assassination in 1933, the new Chicago Democratic machine began courting Black voters by increasing “the patronage, both petty and important.”Footnote 94 FDR’s vote share among Chicago’s growing Black community also swelled after 1932, even as White voters in the city tempered their enthusiasm starting in 1936.Footnote 95 FDR, for his part, viewed the Chicago and Illinois machines as a great boon to his electoral fortunes.Footnote 96
In sum, both the bosses and FDR thought that there was a mutually beneficial relationship taking place: New Deal social programs not only strengthened an existing machine, but according to accounts at the time, they supported the creation of a new Democratic machine. Local bosses sought to leverage work relief funds to bring voters, particularly new constituencies among their Black residents, into the Democratic fold both locally and nationally, and FDR felt the bosses were successful enough to reward them with further control over patronage opportunities. In a word, if there were any two cities where machines were able to leverage work relief programs to FDR’s benefit, we might expect it to be these two.
With this in mind, we turn to the result of our synthetic control analyses, which match Chicago and Pittsburgh, respectively, to a weighted average of all non-machine counties based on their pre-1936 demographic and political characteristics, as described above. Figure 4 shows the results of these. Panel (a) shows Pittsburgh’s vote share trajectory from 1928–1940 as a solid line, compared to its synthetic control in a dashed line, while panel (b) shows the same for Chicago. The dotted vertical line marks the year 1936. Neither Pittsburgh nor Chicago had a Democratic machine prior to the New Deal. This means that we can conceptualize a kind of bundled treatment beginning in 1936: the first presidential election during which both cities had a Democratic machine in place, and the first one during the New Deal.
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20240829034811717-0857:S0898588X23000081:S0898588X23000081_fig4.png?pub-status=live)
Figure 4. Synthetic Control Analyses of Democratic Presidential Vote Share. (a) Pittsburgh, (b) Chicago.
We can think of the dotted line marking the synthetic control as the best approximation, based on observable pre-treatment characteristics, of the unobservable counterfactual: what would have been the trajectory in Democratic presidential vote share in Pittsburgh and Chicago had they not consolidated machines in 1936?
As we can see in both panels, there is little evidence that either machine city outperformed in Democratic vote share, relative to a similar, non-machine city. In fact, both machines actually performed worse than the comparison, non-machine control in 1936 and 1940—just as the negative coefficient estimates in Table 2 would predict. Taken together, the analyses shown in Table, along with Figure 4, suggest that there is very little evidence that urban bosses used federal work relief programs to generate votes for FDR. These findings also suggest a reconciliation of different interpretations as to what federal control over work relief meant for patronage politics. The New Deal’s critics, and later historians, pointed to Senators' victory in winning the right to appoint WPA administrators as the key institutional feature of the new program that allowed for local-national political patronage. But as Figure 4 indicates, this did not benefit FDR in machine cities.
8. Discussion and conclusion
These results contribute to a decades-old debate about the role of urban machines in stitching together a local-national patronage system during the New Deal. Historians have found plenty of evidence that FDR rewarded urban bosses who he believed were leveraging work relief spending to benefit him come election time: he sent work relief jobs to urban bosses, and the bosses used their machines to convert those jobs into votes. And yet, we find only limited evidence that urban machines indeed played this role of linchpin in a local-national patronage arrangement: machines did not see larger increases in work relief funds than non-machine cities, and work relief in general was sent towards the most economically needy counties and those that were the most electorally competitive. Despite this, machines still succeeded in using those funds to bring out votes for Democratic Senate candidates, but not for FDR, the national architect of these policies.
This is true, on average, despite the fact that some urban bosses had contentious relationships with their Democratic Senators. This was certainly true of Mayor Curley in Boston, who found himself at odds with both the Democratic Governor and Senator in Massachusetts. But in other places these relationships were close. For example, in Missouri, Senator Truman consulted with Pendergast before recommending a work relief director,Footnote 97 and in Pennsylvania, Senator Guffey built the Pittsburgh machine in cooperation with county boss David Lawrence. As a result of these relationships, even as bosses did not receive more of the work relief pie, they used the funds they did receive to help deliver votes to Democratic Senators.
How can we make sense of the fact that machines were most successful in helping Senators rather than the president? It is possible that FDR’s early discomfort with patronage—and his visible condemnation of graft in cases like Ohio—led local machines to focus on procuring votes for Senators, often their long-standing patrons, rather than the President, an unknown outsider. Although there is plenty evidence of Hopkins’ technocratic goals in the early days of FERA and the CWA, it is also clear that the haphazard approach to relief and the administration’s reliance on local agencies in carrying out left plenty of opportunity for patronage. Even if machines shifted their efforts to include votes for FDR during the WPA, they may not have been persuasive enough to control the local vote at a time when the New Deal and the President were suffering from growing criticism.
In short, while urban machines were ideally suited to turn federal funds into Democratic votes for FDR, they turned out to have limited success in doing so. This is not to say that machines were unimportant: on the local level, there is clear evidence that machines perpetuated an unequal distribution of public funds.Footnote 98 But given every opportunity, they were no more successful than other, similar cities in creating a local-national Democratic Party machine.
This suggests a reinterpretation of the organizational bases of Democratic Party support during the New Deal. Mayhew argued that the New Deal Democratic Party was based in urban machines and CIO unions, later joined by liberal activist groups during the Great Society era.Footnote 99 This created something of a distinction between the local and national Democratic Party, as Mayhew writes, echoing Gosnell’s observations in Chicago: “The upward transformative effect here is quite interesting. Evidently through the medium of exchange relations, party organizations geared on the whole to relatively inactive government at the bottom [machines] came to give vital support to hyperactive government building a welfare state at the top.”Footnote 100 But our findings suggest that these exchange relations were perhaps not as central in building support for the expansion of a federal welfare state, at least insofar as FDR was the driver of that expansion. Given that other research finds even less evidence that social programs generated support for FDR when they were not doled out by machines, this suggests that the link between New Deal programs, machine apparatuses, and the Democrats' New Deal Party was not a straightforward relationship of self-interest and quid pro quo.Footnote 101 Future scholarship should continue excavating the alternative forms of local organization that linked voters to the transformative politics of the New Deal.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0898588X23000081.
Acknowledgements
The authors would like to thank Dan Carpenter, along with the participants of the University of Chicago’s Urban Governance Working Group and Politics, History, and Society Workshop for their helpful feedback.
Funding statement
No sources of funding were provided to complete this research.
Competing interests
The authors declare none.