Civil Service Adoption in America: The Political Influence of City Employees

t the turn of the twentieth century, most cities in America featured a patronage-based system of governance, but over the next few decades, patronage was replaced by civil service. Civil service restructured the relationship between elected officials and government employees, with employees benefiting from a variety of new protections. Yet in studying this change, scholars have largely ignored the role local employees themselves might have played in the transformation. We argue that city employees stood to benefit from civil service, and in places where they had agency and clout, they were important drivers of its adoption. We collected a dataset for more than 1,000 municipal governments, determining whether and when they adopted civil service and whether their employees were organized in an occupational organization. Our analysis of these new data shows the influence of city employees was an important contributor to the spread of civil service in American local government.


INTRODUCTION
T he twentieth century saw a major transformation of American local government, and at the heart of the change were government employees.In the early 1900s, most cities featured a patronage-based system of governance, in which elected officials had considerable control over local government employment.Over the next 100 years, patronage was replaced by civil service-a merit-based system of employment that regularized the hiring, firing, and promotion of municipal employees.Civil service restructured the relationship between elected officials and government employees, with employees benefiting from a variety of new protections.Today, those local employees and their unions-including unions of teachers, police officers, and firefightersare some of the most active groups in American politics, in local government and beyond.Thus, it is hard to imagine a development more central to American politics, and yet we lack a full understanding of its key dynamics.
Research on why and how this transition occurred is strikingly limited.A few studies have explored the politics of municipal civil service adoption, emphasizing the importance of Progressive Era municipal reformers, city demographics, political institutions, and party competition (e.g., Ruhil 2003;Ting et al. 2012), but have not considered the government employees themselves-those who were arguably the main beneficiaries of the reforms.Other research examines the effects of civil service, but again focuses on the effects on policies or political party fortunes (e.g., Folke, Hirano, and Snyder 2011), not on the conditions or organization of employees.And while DiSalvo (2015) and West (2009) point to civil service as a precondition of the shift to public-sector unionization, little empirical research investigates how this happened.The political contributors to the passage of state labor-management relations laws in the mid-twentieth century have received some attention (e.g., Hartney 2022;Moe 2011;Saltzman 1985), but the broader transformation seems to have begun much earlier with the shift to civil service and early advocacy of government employees, including firefighters, police, transit workers, and janitors (Slater 2004).There is still much to be learned about how government employees may have contributed to the adoption of civil service.
This article makes a start by examining patterns of municipal civil service adoption with a focus on the role of government employees.Our argument is simple: as a general matter, municipal employees sought the stability and autonomy that came with civil service, whereas elected officials preferred to retain discretion and control over the bureaucracy.In many places, because they were unprotected and subordinate to local politicians, city employees lacked agency.They were uncoordinated and politically weak, and thus, elected officials had little reason to give employees the security they wanted.In other places, however, groups of employees managed to coordinate and act cohesively in politics, and when they did, one of the main changes they pushed for was civil service.In cities where employees had agency and clout, they had the potential to change the political calculus of elected officials, increasing the likelihood that the city adopted civil service.While we do not propose this was the only pathway by which municipal civil service was attained, we argue that it was an important one for many cities and that city employees made contributions to the development of American cities that have not been fully appreciated.
To study this, we collected a dataset of whether and when more than 1,000 municipal governments adopted civil service.We augment this dataset with data on the local presence of early organizations of firefighters, which were (and are) one of the largest and bestorganized categories of municipal government employees. 1Our analysis shows that cities with early organizations of firefighters were significantly more likely to transition to civil service during the first decades of the twentieth century.Because it is difficult to isolate a particular causal pathway with these data, we take several approaches to bolster the evidence that city employee advocacy was a contributor to the adoption of civil service: (1) controlling for city demographic characteristics deemed important in the literature, (2) analyzing variation within cities and states over time, (3) testing for a differential role of employees by city size, (4) accounting for efforts of municipal reformers, (5) preliminary analysis of where and why early firefighters' organizations formed, and ( 6) presenting examples where employee groups were responsible for enactment.Together, our findings illuminate an understudied contributor to this transformation of local government: the political influence of city employees.

BACKGROUND AND LITERATURE ON MUNICIPAL CIVIL SERVICE
Prior to civil service reform, American politics was largely a patronage system in which incumbent politicians could hire and fire employees on the basis of political calculations.Victorious political party organizations filled government positions with their supporters, who then helped the party win future elections.From the perspective of government employees, the way to get a government job was to be in service to a political party or set of candidates that would win elections, and the path to job security was to work to keep those officials in office.This distorted democratic representation and accountability (Bryce 1888;Royko 1971;Steffens 1902;Trounstine 2008).
The Pendleton Act of 1883 introduced civil service to parts of the federal bureaucracy, and two key pillars of that reform were the recruitment of employees based on merit, through competitive examinations, and the prohibition of firing or demoting employees on the basis of politics.While bureaucracy scholars have probed how this change shaped the federal bureaucracy (e.g., Carpenter 2001;Gailmard and Patty 2007;Shefter 1983;Skowronek 1982) and the events that led up to it (e.g., Hoogenboom 1961;Van Riper 1958), others have puzzled over why the shift to civil service in state and local governments was slow, gradual, and piecemeal.Half a century after the Pendleton Act, only nine states had adopted state civil service, and there were no bursts of reform in cities until the 1910s and 1930s (Ruhil 2003;Ting et al. 2012;Ujhelyi 2014).Moreover, many early municipal civil service systems were not comprehensive but rather applied only to certain city departments, most commonly police and fire-which were often the largest departments in the city (as they are today).
In 1939-1940, changes in federal policy altered the environment for state and local governments.In 1939, an amendment to the Social Security Act required state and local governments to enact merit-based personnel systems for any employees administering funds related to social security, health, or unemployment compensation (Ruhil 2003;Ting et al. 2012;Ujhelyi 2014).Starting in 1940, then, many subnational governments likely adopted civil service to comply with this requirement.Notably, cities rarely eliminated civil service after adopting it (Ting et al. 2012;Tolbert and Zucker 1983).
Most research on state and local civil service has focused on its effects (e.g., Folke, Hirano, and Snyder 2011;Kuipers and Sahn 2023;Rauch 1995;Ujhelyi 2014).Research examining the contributors to civil service adoption has been more limited-more focused on state governments than local governments and focused on a small set of possible factors.Scholars have shown that the Pendleton Act spurred reforms in some places (Miller 2009;Schiesl 1977;Thelen 1972).More importantly, historians have described how civil service was a component of the agenda pushed by Progressive Era reformers who emphasized the importance of technocratic administration of government as a way to solve problems of municipal governance and bring down political machines (Frant 1993;Hoogenboom 1961;Kaplan 1937;Stewart 1929;Tolbert and Zucker 1983).But while the National Municipal League advocated for civil service in its model Municipal Program (Liazos 2020), Progressives regularly failed to achieve civil service enactment at the state and local levels (Berry 2009).To date, there has been no research that quantitatively measures the presence of reform organizations or their influence on civil service adoption.
Other theoretical accounts emphasize incumbent politicians' electoral incentives to adopt civil service.Focusing mainly on state governments, a model by Ting et al. (2012) highlights the importance of the competitiveness of elections for spurring politicians to adopt civil service.Ruhil (2003) emphasizes the costs and benefits of civil service versus patronage, arguing that the many changes in the late nineteenth and early twentieth centuries (immigration, shrinking opportunities for patronage at the federal level, changes to political institutions) altered city incumbent politicians' electoral calculus, making civil service more in their interest than it had been in the past.
Quantitative empirical studies have also identified some city-level correlates of adoption.Ruhil (2003) and Tolbert and Zucker (1983) analyze civil service adoption in the early decades of the twentieth century (in 167 cities of more than 50,000 residents and 252 cities of more than 30,000 residents, respectively) and find that larger cities were more likely to be early adopters.This is to be expected, scholars argue, because between 1900 and 1940 city populations grew rapidly.These larger populations generated more demands on government, and civil service was thought to be important for higher quality governance and building state capacity (Finegold 1995;Lowi 1964).Tolbert and Zucker (1983) also find that early adopters tended to have a larger middle class (measured as smaller proportions of manufacturing workers and "illiterates"), and Ruhil (2003) puts emphasis on city political institutions.
Finally, these studies find state and regional patterns.Ruhil (2003) shows cities in the South were less likely to adopt civil service, although the reason remains unclear.State policies on civil service also matter.Cities in states where statutes or constitutions required it were more likely to have civil service, which, in the early years of the twentieth century, was only the case in New York, Massachusetts, and Ohio (Tolbert and Zucker 1983).2Also, in states that had civil service requirements for state agencies, cities were more likely to have civil service (Ruhil 2003).

GOVERNMENT EMPLOYEES
Strikingly absent from most of the literature on both the effects of and contributors to civil service adoption are the government employees themselves: the individuals who were directly impacted by the reforms.There is some political science literature on government employees and their organizations, but it almost exclusively examines patterns of their activity and influence since the late twentieth century and with an emphasis on collective bargaining and unionization (Anzia 2022;Anzia and Moe 2015;Moe 2011;2019;Hartney 2022;Saltzman 1985).Long before the shift to collective bargaining in the public sector in the late 1950s, however, there was a shift to civil service.DiSalvo (2015) and West (2009) argue that the ability of government employees to organize in the mid-twentieth century was aided by the earlier adoption of civil service and the erosion of patronage politics, which gave rise to greater opportunity for employees to advocate for their interests independent of public employers.Political scientists have yet to analyze how government employees might have helped to shape these institutional developments in American local government in the first place.
We argue that as of the early twentieth century, city employees had a strong interest in civil service and that when they had agency, they helped influence its adoption.Compared to patronage, civil service had clear, immediate benefits for government employees.Positions covered by civil service were regularized with job qualifications, open advertisements, and, frequently, examinations.Perhaps most importantly, civil service employees could not be removed from their positions for political reasons.While civil service examinations and hiring could be manipulated by enterprising politicians to ensure the hiring of their supporters, it was harder to manipulate dismissal without cause (e.g., Erie 1988).Civil service systems were widely viewed as removing the political yoke around public employees, offering them predictability and stability in employment.Thus, as a general matter, employees had a strong stake in civil service and stood to benefit from its adoption.
With patronage widespread (Erie 1988;Mayhew 1986;Sorauf 1960), however, city employees and others interested in civil service likely encountered powerful resistance, because these changes reduced elected officials' control over city bureaucracy and outlawed a primary means by which they built electoral support.Moreover, city employees across the country varied in whether they had agency-political capacity and coordination to advocate for institutional change.We propose this variation in local government employee agency affected the extent of bottom-up pressure politicians faced.We expect cities where the employees managed to coordinate and act collectively in politics were more likely to see civil service enacted.
While early local government employee agency could take a number of forms, including the formation of mutual benefit societies, active participation in elections, and informal lobbying (see Slater 2004), one of the most common and effective vehicles for exercising political voice-and one of the most visible and measurable-was the formation of employee unions and professional associations (see Banfield and Wilson 1965, 210-6).Early occupational organizations of government employees in the United States date back to the late nineteenth century, and by the early twentieth century, there were several local government employee organizations, mostly organized along craft lines.The National Education Association (NEA) was formed in 1870, and the American Federation of Teachers started in 1916 (Spero 1948, 314, 319).Police officers also formed local organizations-the first chapter of the Fraternal Order of Police (FOP) organized in 1915 in Pittsburgh (Walsh 1977)-but their organization was more controversial, especially the question of whether local police organizations should be allowed to join the labor movement.Shortly after the American Federation of Labor (AFL) agreed to issue charters to police unions in 1919, the Boston police strike set back the organization of police officers for many years (Slater 2004), and as a result, police unionization proceeded in a relatively fragmented and delayed manner.
In cities, firefighters were some of the earliest organized public employees.During the nineteenth century, the threat of conflagration loomed in America's cities: By the late 1800s, fire had decimated dozens of the nation's largest cities, and the problems grew worse as cities densified and buildings increased in height.Yet prior to the Civil War, firefighting was an all-volunteer operation, albeit with some contributions from city coffers (Tebeau 2003).By mid-century, property owners and fire insurance companies began to push for municipal fire departments staffed by paid firefighters.By 1900, nearly all large cities (those with populations over 25,000) had established fire departments with paid firefighters (Bureau of the Census 1905).However, Tebeau (2003, 239) explains that at this time, "politics tied firemen to ward leaders and structured their relationships to their local communities.In return for their assistance during elections, ward officials rewarded firemen and other municipal workers with employment." As early as the 1880s, firefighters organized mutual benefit societies and social clubs.In addition to providing death and illness benefits, they advocated for changes to local firefighting practices and personnel matters, like regular days off, higher wages, and rotation in shifts (New York Sun 1918;Department of Commerce 1918;Spero 1948, 228-44).In 1873, the National Association of Fire Chiefs was established (later becoming the International Association of Fire Engineers), and the International Association of Fire Fighters (IAFF) was formed in 1918, organizing rank and file firefighters to promote better working conditions. 3The IAFF considered itself part of the labor movement and identified as a union (and was chartered by the AFL), although as of 1930 it prohibited members from striking (Slater 2004, 35).
The historical record illustrates that the IAFF played a role in civil service adoption, revisions to civil service laws, and protection of the laws once they were in place.In its documentation of its organizational history, the IAFF website features a cartoon from the organization's magazine, dated 1934, showing an IAFF firefighter exterminating "politics" under a tree bearing the "fruits of civil service" (see Figure 1).The caption argues, "if used freely, satisfactory results guaranteed."4In 1932, in Salem, Oregon, firefighters dropped leaflets across the entire city promoting the adoption of a civil service ordinance (Statesman Journal 1932).The IAFF also worked to amend civil service laws and decisions that they felt undermined their members (e.g., Brooklyn Standard Union 1921).
We propose civil service was especially desirable to these nascent local employee groups because of the political and legal environment of the early twentieth century.At the time, formal collective bargaining was illegal in government.When the National Labor Relations Act (NLRA) sanctioned collective bargaining for private-sector workers in 1935, public-sector employees were excluded (Walker 2020).Many local governments sought to bar their employees from joining unions, including firefighters (Cincinnati Enquirer 1919;Sacramento Bee 1919;Louisville Courier-Journal 1933;Richmond Times Dispatch 1934;Slater 2004, ch. 3).During this era of hostility toward government employee organizing, employee groups struggled to maintain their organizations.In the face of these collective action problems, civil service promised greater continuity and predictability in employment and longer career horizons, thereby creating more stability in the groups' potential membership and enhancing the incentives for employees to invest in the organizations' efforts.
In addition, these early employee organizations advocated for policies that benefited their members, such as higher compensation, more regular and limited work hours, paid vacation, better working conditions, and regularized processes for handling grievances (e.g., Slater 2004), and they often worked within civil service systems to push for those policies.In 1943, for example, the St. Louis IAFF took out a quarter-page advertisement in the Globe Democrat advocating for higher wages, and the ad implored voters to speak out against the Civil Service Commission's Compensation Plan (St. Louis Globe Democrat 1943).In 1926, when Scranton's mayor threatened to dismiss firefighters to "make room" for his new hires, the Scranton IAFF asserted that it was "fully prepared to protect the firemen and the policemen and [would] also likely take some legal steps to prevent the mayor from venting political spleen under the guise of bringing the entire department up to civil service standards" (Scranton Times-Tribune 1926, 3).Thus, during this period when collective bargaining was a far-off goal, civil service offered government employee organizations an institutional foothold from which to advance policies they favored.
In sum, long before the 1960s and the push for state labor-management relations laws in the second half of the twentieth century, local public employees in many places had political agency.They were active in local and state politics and coordinated to advocate for policies that would benefit them as a group.A primary way they did this-and advanced their interests-was forming and maintaining employee organizations and unions.In municipal governments, firefighters' organizations and unions were especially prevalent and important.
We propose employee political agency was important to the adoption of civil service.As a general matter, elected officials had strong reasons to resist civil service, because it reduced their control over the bureaucracy and upended their ability to build electoral support through the provision of government jobs.The extent of bottom-up pressure from government employees and their organizations thus stands to be an important factor shaping when and whether cities transitioned to civil service.The central hypothesis of our article is that cities where the employees had agency-as shown by their organization-should have been more likely to adopt civil service in the years prior to the 1939-40 change in the federal policy landscape.

DATA
To better understand the relationship between early municipal employee agency and civil service adoption, we assembled an original dataset from tables in the Municipal Yearbooks, which are annual volumes of city statistics and activities compiled by the International City/County Management Association (ICMA, formerly the International City Managers' Association, see Ridley, Nolting, and Arnold [1934]). 5In select years, these Yearbooks contain statistical tables with information on the status of cities' civil service provisions as well as the existence of municipal employee organizations-our primary measure of the employees' political agency.While the information provided and number of cities vary from year to year, this is a rich source of data on personnel matters in American municipal governments in the early and mid-twentieth century.
We started by using tables from the 1940 to 1944 Municipal Yearbooks that identify when cities adopted civil service, back to the late 1800s.Starting in 1945, the relevant tables in the Yearbooks contain information on whether the city had civil service but not the year of adoption.We digitized a select set of tables to generate an indicator of whether a city had civil service each year.Our description of civil service adoption below presents data up through 1962 (the last year for which the Yearbooks contain civil service information for the full set of cities with a population greater than 10,000), but because of The resulting dataset has 1,674 cities for which we have some information on the status of civil service in the city during this time period.For 583, we identified the date of civil service adoption directly from the 1940-1944 Yearbooks.Another 456 cities had not adopted civil service by the last year they appear in the Yearbooks' civil service tracking (usually 1962).304 cities are shown as having civil service for all years in which data are available but for which the Yearbooks do not contain a date of adoption; these cities are excluded from our analysis because we do not know their civil service status for years before 1940.Finally, 331 cities changed to civil service between 1945 and 1963.For these, we drew on all available information in the Yearbooks and coded them as having civil service for every year following the first indication of civil service in the Yearbooks.We also test the robustness of our results to excluding cities for which we approximated the date of adoption (Supplementary Table A2).Using all this information, we created a panel dataset of the cities from 1883 to 1962 with the binary indicator Civil Service.As others note, cities rarely eliminate civil service once it is adopted, so we code cities as having civil service for every year following initial adoption.
We also use information from the Yearbooks on early municipal employee organizations.It is important to note that comprehensive data on employee organizations in US cities do not exist even today; there are no datasets on whether US municipal governments have unions, let alone detailed data on the specific organizations (Freeman and Ichniowski 1988).Thus, while the information on early employee organizations in the Yearbooks is not perfect, it is a valuable and underutilized resource and one that sheds light on early city employee organizing.
Specifically, the 1938-1940 Yearbooks include lists of municipal employee organizations and, for each organization, the year it was established.The ICMA assembled these lists by acquiring directories of a few prominent national-level employee organizations, including the IAFF and the American Federation of State, County and Municipal Employees (AFSCME), and then listing their local chapters in each state.It is not a comprehensive list of all local employee organizations at the time, because if the ICMA did not obtain a directory from a particular national organization, or if there were local organizations unaffiliated with a national organization, they were not included.Notably, police organizations are absent from the list. 6or our analysis, then, we focus on the IAFF, which is a good proxy for employee agency for several reasons.First, it was the organization with the broadest reach into American municipal governments in the early twentieth century.As of 1940, it had been around for decades and counted the largest number of locals-in over 400 cities (Ridley, Nolting, and Arnold 1940, 141-3).AFSCME, by comparison, started later, in 1935.Second, firefighters were a large share of city employees.Third, unlike police or teachers, virtually every local firefighters' union is affiliated with the IAFF, which means the ICMA data are most likely a comprehensive list of local firefighters' unions at the time.
Finally, while we have data going up to the 1960s, we focus on years up to 1940 because we are interested in the relationship between local employees' organizations and local civil service-and national circumstances changed in 1940 when new federal legislation encouraged the adoption of local civil service (Folke, Hirano, and Snyder 2011;West 2009).The impact of federal legislation makes it more difficult to disentangle local motives for civil service adoption from changing incentives created by national legislation.

ANALYSIS
We begin by displaying the number of municipal civil service adoptions by year in Figure 2. Our dataset includes more cities and spans a longer time period than the data analyzed by Ruhil (2003), but we see patterns similar to what he shows: (1) an initial phase of municipal civil service adoption immediately following the Pendleton Act of 1883, most of which were in New York and Massachusetts, (2) a second surge of adoptions between 1910 and 1919, after which they slowed; (3) a third surge of adoptions between the late 1920s and 1935; and then (4) a massive increase between 1935 and 1940.In addition, many municipalities did not adopt civil service until after 1945: there was a large increase in adoptions after World War II and a continued trend of new adoptions during the 1950s.Moreover, hundreds of cities had not adopted civil service by 1962.
In Figure 3, we show the years in which cities first established local IAFF chapters by 1940 (for those that had).It shows that a number of cities established an IAFF local just before 1920.As with civil service adoption, there was a slowdown in IAFF local creation in the early 1920s, followed by a second surge of new local IAFF organizations starting in the late 1920s, with adoptions per year continuing to mount through 1940.That our dataset records more than 400 IAFF chapters by 1940 is notable, because it was more than double the number of chapters in the next largest local employee organization: AFSCME.
Two points about Figures 2 and 3 are worth underscoring.First, patterns evident in Figure 3 align with how others have described the advances of the early labor movement.As Slater (2004Slater ( ) documents, 1916Slater ( to 1919 were especially active for employee organizations in both the public and private sectors.There were many strikes in these years, with 3,600 strikes including 4 million workers throughout the United States in 1919 alone (30). Moreover, between 1915 and1921, the share of public-sector employees in unions increased from 4.8% to 7.2%, even as total government employment rose from nearly 1.9 million to nearly 2.4 million (18).Then, after the 1919 Boston police strike, the 1920s saw a weakened labor movement, with many cities and states prohibiting public-sector workers from joining unions.Both of these patterns are reflected in our data in Figure 3 on IAFF local chapter establishments.The second pattern of note is that municipal civil service adoption slowed during the 1920s just as labor organization and activity slowed as well.This is suggestive that employee organizations were a contributor to civil service.
To examine the relationship between employee organizations and civil service, we begin by analyzing whether cities that had early IAFF locals were more likely to have adopted civil service by 1940.For that analysis, we need to account for city-level correlates of municipal civil service adoption that have been identified in the literature, such as city size, population diversity, and state-level civil service requirements.
To do this, we encoded data from the decennial Census of Population and Housing for 1930.
We have data on the civil service status of 1,371 cities as of 1940, and at that time, 38% had civil service.When we combine these cities with four city-level Census variables-population and the shares of the population that were illiterate, foreign-born, and Black-our dataset is reduced to 1,211 cities.In Table 1, we present the averages of these variables, separately for those with 1 8 9 0 1 8 9 5 1 9 0 0 1 9 0 5 1 9 1 0 1 9 1 5 1 9 2 0 1 9 2 5 1 9 3 0 1 9 3 5 1 9 4 0 1 9 4 5 1 9 5 0 1 9 5 5 1 9 6 0 and without civil service by 1940.As others have found, cities that adopted civil service during this early period were typically larger and had higher shares of foreignborn residents than those that had not yet adopted it.In addition, on average, the early civil service adopters had smaller shares of Black residents.
We also examine the share of cities that had civil service by 1940 in each of the four major regions of the United States and find variation across regions.It was most prevalent in the Northeast, where 54% of the cities had civil service by 1940.In that same year, 47% of cities in the Midwest had civil service and 42% in the West.The South stands out for its low rate: Only 21% of its municipalities had civil service by 1940.
Moreover, of the cities that had adopted civil service by 1940, 51% had an IAFF.That number is only 19% for the cities that did not have civil service.When we consider whether a city had any known municipal employee organization by 1940 (as documented in the Yearbooks)-whether it was an IAFF chapter or otherwise-the numbers change little: 21% of the cities without civil service had a documented employee organization by 1940, compared to 55% of the cities with civil service.
We next use OLS to regress the civil service indicator (for each city as of 1940) on the indicator for IAFF locals, controlling for the city demographic variables shown in Table 1 and clustering standard errors by state.7 Column 1 of Table 2 presents those estimates.We include indicators for three of the four Census regions to account for regional patterns in civil service adoption, and we also include an indicator for whether the city's state had a civil service law that covered state workers by 1940 (coded according to Ting et al. [2012]).The estimates show that cities that had IAFF locals by 1940 were more likely to have civil service by that time as well, even accounting for these other city characteristics.On average, cities with IAFF locals were about 25 percentage points more likely to have adopted civil service during this early period.In column 2, we replace the IAFF indicator with an indicator for whether there was any municipal employee organization in the city by 1940, and our estimates are substantively the same.
In column 3, we combine the indicator of IAFF locals with another measure of local firefighters' agency: whether the city had a firefighters' benevolent association as of 1917, which we created using a list of such associations provided in a 1918 Department of Commerce report called "Statistics of Fire Departments of Cities" (Department of Commerce 1918, 22).The main independent variable in column 3 equals 1 if a city either had an IAFF local by 1940 or is listed as having had a firefighters' benevolent association in 1917; it equals 0 otherwise.Again, the estimated coefficient is positive: on average, cities with firefighters' organizations were about 25 percentage points more likely to have civil service by 1940.
In column 4, we return to the model from column 1 and include state fixed effects instead of region fixed effects.This accounts for state-level characteristics associated with municipal civil service and local IAFF organizations, such as state personnel laws.In addition, while many cities adopted civil service locally, state legislatures played an especially large role in governing municipalities during this period, and some cities acquired civil service as a result of state legislation.While the Yearbooks do not include information about how civil service was adopted for each of these cities, with the inclusion of state fixed effects, we can evaluate whether within states, cities with IAFF locals were more likely to get civil service.We find that they were indeed more likely to get it.The coefficient estimate on IAFF is reduced-to about 14 percentage points-but  still reveals a meaningful positive relationship.Thus, there is an association between early organization of employees and civil service adoption, as we expect.Moreover, we find that larger cities were more likely to adopt civil service early, as were cities in states that had civil service provisions for state government.Cities with larger Black populations were less likely to adopt civil service, which fits with reports that in some cities, white individuals opposed civil service adoption because Black individuals would be able to apply for jobs (Liazos 2020, 105).
We next take advantage of the time-series nature of the data on civil service adoption and IAFF presence, regressing the civil service indicator for each city-year on the presence or absence of an IAFF local in that cityyear.We limit the analysis to years from 1900 to 1940 because IAFF organization did not begin until the twentieth century. 8Although Census data are available for many of our cities, there is a great deal of missingness in these Census variables over time.So, here we exclude the demographic controls and add fixed effects for cities and years, which account for characteristics of cities that were constant over time as well as national yearly trends likely to affect all cities, such as the Great Depression.As is typically the case with observational data, and as we discuss below, there remain threats to causal inference, 9 but these models allow for additional assessment of whether employee organizations were involved in the push for civil service.
The results in column 1 of Table 3 again suggest that employee organizations were associated with civil service adoption.Partialing out the effects of secular time trends and time-constant city characteristics associated with both IAFF organization and civil service, we find that on average, cities with organized firefighters were 21 percentage points more likely to adopt civil service.
In column 2, we add an indicator for whether the cities had an AFSCME local in a given year.AFSCME began to organize close to two decades later than IAFF, so we do not necessarily expect the presence of an AFSCME local to be as strongly related to the adoption of civil service during this period, but this model serves as a test of whether it was really government employee agency that made a difference-as opposed to something else about cities where firefighters organized.The results show both types of local employee organizations are significantly associated with a greater likelihood of civil service adoption: a 21-point increase for IAFF and a 7-point increase for AFSCME.
Next, we add total population (logged) to our model, which we assembled from decennial records from 1890 to 1940 using Census of Population state reports and interpolated within cities for the between-census years. 10Cities were generally increasing in population during this period, and one possibility is that the estimated coefficients on employee organization in columns 1 and 2 could reflect a pattern of faster-growing cities being more likely to adopt civil service and establish employee organizations.In column 3, we add logged city population to the model from column 1 but exclude year fixed effects.We find greater population was indeed associated with greater likelihood of civil service adoption, and also, the coefficient estimate on IAFF remains positive and statistically significant.In column 4, we add year fixed effects and find cities were about 20 percentage points more likely to adopt civil service when they had IAFF locals.Finally, in column 5, we add the indicator for an AFSCME local, and we find both IAFF and AFSCME locals are associated with a significantly greater likelihood of civil service adoption.Thus, even accounting for secular time trends, time-constant features of cities, and city- 8 In the Supplementary Material, we use different time windows (A5) and evaluate whether the association varies by decade (A7). 9In the Supplementary Material, we discuss the parallel trends assumption and newer estimators for empirical applications with staggered treatment adoption and heterogeneous effects (see review by de Chaisemartin and D'Haultfoeuille {2023]).While the two-way fixed-effects model allows us to account for time-invariant city characteristics and annual trends, we cannot interpret these estimates as the causal effect of IAFF organization on civil service adoption.However, they are consistent with our theory. 10We hand-entered data for incorporated places from the following:

CAUSAL PATHWAY: ADDITIONAL EVIDENCE IN FAVOR OF CITY EMPLOYEE AGENCY
Our analysis so far demonstrates a powerful link between employee agency and civil service adoption.There remains some question, however, about whether employee organizations led to civil service or civil service in some cases could have made it easier for government employees to organize.As the policy feedback literature emphasizes, the political activity of interest groups can lead to policies favorable to those groups, but policies can also create incentives for interest groups to form, grow, and engage in politics (Hacker and Pierson 2014;Pierson 1993).With our quantitative data, the sequence is somewhat difficult to parse out, and the reality is that both processes were probably in play.To offer additional evidence in support of our proposed time ordering (i.e., IAFF chapters were commonly established in advance of the adoption of civil service), we replicate the model in column 1 of Table 3 with one-, two-, and five-year lags of the IAFF variable.The results, shown in Supplementary Table A6, show a strong association between IAFF establishment and civil service adoption in subsequent years.This further suggests that municipal employees were actively recreating the terms of their employment during this period in a way that scholars have not previously recognized.Another potential concern with the results above is that perhaps there may have been other time-varying characteristics of cities associated with local firefighters formally organizing an IAFF chapter and cities adopting civil service.We do not assert that employee agency was the only contributor to civil service adoption, or that it was the main reason for civil service reform in every state and city.We also cannot feasibly account for all the potentially relevant characteristics of over a thousand cities a century ago.To move closer to an understanding of the role employees played, however, in this section, we carry out additional analyses that consider other factors that may have also contributed to the spread of civil service.

City Size
We first evaluate the contributions of two alternative pathways to civil service emphasized in the literature: city size and Progressive Era municipal reform efforts.Beginning with city size, the existing literature suggests that civil service was adopted to fulfill a functional need (e.g., Tolbert and Zucker 1983). 11In large, urban communities where the demands of service provision grew exponentially in the early twentieth century, it could be that elected politicians had incentives to build state capacity and professionalized their municipal workforces to handle these new demands, regardless of any pressure from employees, and irrespective of what the employees wanted (Finegold 1995;1964).In our earlier analysis, we accounted for logged city population in our models, and in the additional analysis shown in the Supplementary Material, we estimate the model from column 1 of Table 2 semiparametrically, allowing logged population to enter nonlinearly, and we still estimate a large, positive coefficient on IAFF.Even so, these aggregate results could mask a relationship that varies by city size.If the existing scholarship is correct, employee agency could have mattered more for smaller to mid-sized cities than for the largest cities.
Figure 4 shows the proportion of cities in each of four population bins with civil service as of 1940 (mediumgray bars).We also show this proportion separately for cities with and without IAFF locals in each population category.The light-gray bars in the middle show the proportion of cities without IAFF locals that had civil service, and the black bars show the proportion of cities with IAFF locals that had civil service.The black bar is higher than the light-gray bar in all cases, showing that for all population categories, a larger share of cities with IAFF locals had civil service than cities without IAFF.However, there is also some indication that employee organization mattered less for the largest set of cities than for small and mid-size cities: The gap between the lightgray and black bars is larger for the first three size categories than for the largest cities.Moreover, in the Supplementary Material, we re-estimate the model from column 1 of Table 2, except that we interact the IAFF indicator with logged city population (centered around its mean).We still find that larger cities and cities with an IAFF local were more likely to adopt civil service, but the coefficient on the interaction term is negative.This suggests the IAFF was less important to the adoption of civil service in large cities like Seattle and more important to smaller and mid-sized cities, such as Aberdeen and Bellingham, Washington.Understanding institutional development in these smaller to mid-sized cities is important, moreover, because while city population growth was rapid proportionally speaking, many turn-ofthe-century cities remained quite small.In 1940, the Census counted 3,646 urban places (cities with populations greater than 2,500 people), which were home to approximately 57% of the US population, but nearly 70% of those places had fewer than 10,000 residents (Bureau of the Census 1941).Another account of civil service adoption in the literature emphasizes the power of Progressive Era municipal reformers.As we have explained, during the first decades of the twentieth century, municipal employees were largely hired and fired through patronage systems, and in about 30% of American cities, patronage systems formed the foundation for strong political organizations that dominated politics for a decade or more-commonly known as political machines (Trounstine 2008).Even where stable political machines did not dominate local government, the use of patronage was widespread and generated a high degree of turnover among municipal employees.Municipal reformers opposed the "spoils system" because it was believed to ensure the victory of machine politicians and because, with it, "efficient administration is impossible" (Boston Evening Transcript 1897, 8).Civil service was listed among reform goals, alongside home rule charters, city manager structures, nonpartisan elections, the abolition of district-based representation, and other institutional changes intended to address governance challenges (Stewart 2022).Municipal leagues (local reform organizations) proposed that "if our large municipalities are to be no longer cesspools of corruption, if our municipal governments are to be made honest and businesslike, if our police forces are to be kept clear of thugs and thieves, the appointments to places in the municipal service must be withdrawn from the influence of party bosses and ward heelers and must be strictly governed by the merit system" (New York Times 1894, 16).
Further, there are cases in which reformers clearly drove the push for civil service.In Iowa, for instance, Des Moines city leaders vigorously campaigned for a state-level law permitting cities to adopt charters that included nonpartisan elections, direct democracy, and civil service (Shambaugh 1911).The Des Moines Plan was implemented by the Iowa legislature in 1907 and then, after a failed attempt, adopted in Des Moines in 1908 (Des Moines Daily Tribune 1907, 1; Vilas County News 1908, 8).Our dataset shows many other Iowa cities adopted civil service before their IAFF chapters were established.
Clearly, then, municipal reformers supported civil service laws and hoped to weaken political machines through their implementation.Less well understood is the extent to which their efforts generally contributed to civil service adoption in US municipal governments.There are reasons to question whether the adoption of these laws can be attributed to the presence of reform organizations: For one, civil service became a broadly popular reform. 13Additionally, there is evidence that cities with council-manager governments (favored by reformers) are actually less likely to have civil service (Frant 1993).While we assert that city employees contributed to civil service adoption in many places, we do not claim that municipal reformers were uninvolved.Still, one might be concerned that the relationship we have found could be mainly due to the efforts of reformers.As a way of accounting for the presence of reform or machine organizations, we gathered data from a table in the 1940 Municipal Yearbook that codes cities as having a city manager or commission-style government, specifies the share of the city council elected at large, and notes whether municipal elections were nonpartisan.In Supplementary Table A10, we show that adding these variables to our base model does not change our conclusions.Furthermore, these variables are not significant positive predictors of cities getting civil service.We also look at places where reformers had more difficulty achieving their goals: cities where machine organizations dominated.We draw on data from Trounstine (2008), which includes codes for the presence of political machines for 190 cities in our dataset.As shown in Supplementary Table A10, adding an indicator for a dominant machine negatively predicts the adoption of civil service, but the coefficient on IAFF is strong and positive.

Private-Sector Labor Organizing
Another approach to evaluating the possibility of omitted variable bias (and reverse causality) involves explaining variation in city employee agency.Our measurement and description of the presence of these early employee organizations is itself a significant contribution and opens up a host of new research questions, including why city employees had agency in some places and not others; what forms their political activity took, and thus, how they had influence; and what other policies and institutions they may have helped to shape.We cannot provide comprehensive answers to all such questions in this article, but here we provide an exploratory analysis of the early years of firefighters' unions in a preliminary effort to understand the variation in our main independent variable.
First, we evaluate whether the presence of early IAFF locals varies with city size, demographic characteristics, or region-the independent variables included in our earlier models.The results in Supplementary Material (A11) show that larger cities were more likely to have an IAFF local by 1940 than smaller cities and that the likelihood of a municipality having an IAFF local was relatively higher in the Midwest and lower in the Northeast.
We also propose and consider a second possibility: that other labor unions, especially those of workers in the private sector, helped to spur the formation of these early firefighters' unions in US cities.Most accounts of the labor movement in the United States focus on private-sector labor (see McCartin 2006), and the studies that do discuss the history of public-sector unions emphasize their rise in the second half of the twentieth century (e.g., Saltzman 1985).Our data, however, show that union locals formed in many cities early in the twentieth century.Moreover, even though they were government employees, early firefighters' locals were chartered and supported by the AFL.In addition, as we show in the Supplementary Material, the surges and slowdowns in IAFF establishments shown in Figure 3 coincided with surges and declines in work stoppages in the United States, and the latter were almost all strikes by private-sector unions.Thus, even though the passage of the NLRA in 1935 gave private-and publicsector unions different legal statuses and rights, there are signs that the early development and organization of labor unions in the two sectors may have been interconnected.
As a preliminary exploration of this, we draw on the work of Holmes (2006), who presents evidence that unionism in large mining and steel establishments in the mid-twentieth century spilled over to other industries in the same geographic area, such as healthcare and grocery stores.We extend this logic and consider whether unions in mines and steel mills also helped to inspire the agency and organization of city employees such as firefighters.It is reasonable to think that firefighters, miners, and steelworkers had overlapping social networks in some of these cities and faced similarly dangerous occupational conditions that might be ameliorated with organization.One possibility, then, is that some firefighters organized unions thanks to their proximity to and solidarity with unionized mining and steelworkers.
We begin by considering the nine cities that had firefighters' locals by 1916 (all of which were affiliated with the IAFF when it was formed in 1918).Every one of those cities was close to significant mining or steel production at the time, including not only large cities like Pittsburgh and Chicago but also Pueblo, Colorado; Great Falls, Montana; and Wheeling, West Virginia.In Figure 5, we explore this visually by mapping countylevel data on mining employment from the 1940 Census.We shade counties according to the log of one plus the number of mining workers in 1940, with darker shades indicating counties with more mining workers.Cities that had established an IAFF local by 1920 are denoted with points on the map.The map makes clear that many of the earliest IAFF locals were in or close to areas with greater mining employment.
Next, in Table 4, we regress the indicator for whether a city had an IAFF local by 1920 on logged mining employment (using the 1940 county-level data) and logged city population in 1920.In column 1, the estimated coefficient on logged mining presence is positive and statistically significant.In column 2, we add regional fixed effects, and in column 3, we add state fixed effects.Across these three models, we find a positive relationship between mining in the county and IAFF formation by 1920.However, after the 1920s, the spread of the IAFF appears to have had less to do with mining employment.In column 4, we replace the dependent variable denoting IAFF formation by 1920 with one denoting formation by 1940 (and include logged population from 1940), and the coefficient is substantially smaller.
There is much more research to be done on this question, and our analysis here is preliminary.Still, the results suggest that firefighters may have been influenced by private-sector unions in mining and steel establishments and that this may partially explain the geographic pattern of early firefighter unionization.To the extent private-sector union strength helps explain variation in our main independent variable-the presence of IAFF locals in cities-it is an explanatory factor that would almost certainly have a weak correlation with civil service adoption otherwise.One would not expect miners' and steelworkers' unions to have much (or any) stake in municipal civil service, suggesting that IAFF organizations developed first and then pushed for the adoption of civil service in some set of cities.

Examples of City Employee Organizations Influencing Civil Service Adoption
Perhaps the most direct evidence in support of the proposed mechanism comes from cases in which organized employees advocated for civil service and claimed victory when it was adopted.The link between employee organizations and civil service is apparent at all levels of government and for many kinds of government employees.In 1936, the AFL chartered AFSCME-the first national union of state and local public-sector workers-and a key pillar of AFSCME's agenda was expanding and strengthening civil service.AFSCME's constitution declared as a primary goal "the extension of the merit system to all non-policy determining positions of all governmental jurisdictions" (Kramer 1962, 27).At the federal level, the AFL chartered the National Federation of Federal Employees in 1917, and it went on to devote considerable resources to enforcing and bolstering the federal service system and warding off a return of the spoils system (Johnson 1940;Johnson and Libecap 1994).The Building Service Employees International Union (BSEIU, later renamed SEIU) used a variety of political strategies-including "behind-thescenes deals with officials, lobbying, appeals to the public, and other kinds of informal activities"-in its successful attempts to get its members covered by local civil service laws in the late 1920s and 1930s (Slater 2004, 8).
In Wyoming, the four cities in our dataset all had IAFF locals by 1930, and after the Wyoming state legislature passed its civil service law for cities in 1933, the IAFF local from Casper, Wyoming, claimed "our civil service bill took the hurdles against some pretty strong opposition…there may be faults in organization, but failure to accomplish beneficial results is not one of them" (Casper Tribune-Herald 1935, 5).In Washington State, many cities, including the largest, adopted civil service locally, but eventually, in 1935, the state legislature passed a law requiring municipal civil service for firefighters.(The state passed a separate law covering police officers in 1937.)The IAFF notes that state law in Washington had been patterned after the IAFF's own Model Civil Service Bill. 14hile our quantitative dataset does not cover police organizations before 1940, police organizations were also involved.A speech made by a former FOP Grand President in 1939 indicated that the FOP had engaged in the fight for civil service, saying that "We spent a lot of money to do it…" and that now they had "job security."He also highlighted how valuable the shift from the spoils system to civil service had been for police officers.In the past, he said, "a policeman's job depended on the corner saloon keeper… [and] when the Mayor took office half the police force went out" (Walsh 1977, 115).According to Reading, Pennsylvania's FOP Lodge President, this was counterproductive for the force and for taxpayers since it took at least "three years to make an officer a good officer" only for that officer to be fired a few months later for "political reasons" (Reading News-Times 1921, 5).In West Virginia, in 1937, the state passed a law requiring civil service for police departments in cities with more than 5,000 in population, and the head of the state's FOP was credited as a major force behind the new law: "Successful enactment of this measure, after two years of work towards the objective, is in a large measure due to the effort of a local man… Henry B. Squires, assistant chief of police here and president of both state and national orders of the Fraternal Order of Police…In his capacity of state F. O. P. president, he has led this latest drive for civil service regulations…" (Beckley Post-Herald 1937, 4).
Thus, there are clear examples where municipal employee organizations actively pushed for civil service-and were credited with success when it was adopted.In fact, contemporary economist John Commons called municipal employee organizations "the most important political contribution that has been made to civil service reform in a democratic government" (Commons 1913, 111).

DISCUSSION
Over the course of the twentieth century, American government underwent a fundamental transformation: from a spoils system in which government employees were subordinate to political party leaders to one in which employee associations and unions are not only independent from employers but also highly influential in American politics, especially in state and local governments.However, political scientists have yet to thoroughly investigate how and why it occurred.This article takes an important step by examining the politics of municipal civil service adoption in the first half of the twentieth century.We draw on quantitative and qualitative evidence to show that government employees in many American cities advocated for and secured the adoption of municipal civil service laws that changed the structures governing their employment.Cities that had organizations of firefighters-one of the largest groups of municipal employees-were more likely to adopt civil service early in the twentieth century.
Our qualitative data show that firefighters and other public employees were oftentimes organized quite early and actively pushed for civil service-and that they had success.The examples also suggest that employees used a variety of strategies to exert pressure on policymakers, including lobbying elected officials, engaging in electoral politics, and appealing to the public.Moreover, the quantitative dataset we have collected and used in this analysis is a major advance over what was previously available-and sheds light on patterns that have barely been studied.Our indicator of civil service covers a much larger number and greater diversity of cities than in previous research.Even more important are our new indicators of early city employee organizations, which make possible this first-ever quantitative analysis of the relationship between public employee agency and civil service adoption.
These findings open up a host of questions that deserve greater attention in future research.First, scholars should investigate the extent to which other groups of municipal employees-most notably, police officers-were also engaged in the push for civil service.While the ICMA Yearbooks do not track the early organizations of police officers, there were many local FOP lodges during the 1920s and 1930s, and early FOP conferences placed heavy emphasis on securing civil service laws in local communities.Studying the role of early police organizations is difficult, both because of the lack of data and because many police unions are not affiliated with the FOP.Additionally, it may well be that police officers were not as organized and influential as firefighters during this period (Walsh 1977).Even so, we consider the role of police officers in advocating for civil service an important area for future research.
Second, and more generally, our research shows how widespread public employee organizing and activism were during the first half of the twentieth century.A promising next step will be to do more to evaluate the forms their activity took and how that translated into policy change, including in-depth case studies like Slater's (2004, ch. 4) account of the BSEIU.Furthermore, our focus has been on how they influenced the adoption of civil service, but these early employee groups also advocated for other policies desired by their members, including wage and salary increases, pensions, and shorter and more predictable work hours.The extent of early public employee organizing has barely been studied by political scientists even though it stands to have had major impacts on the development of modern American government.In the future, more research should focus on when, why, and how government employees managed to organize during this period, the role of civil service in aiding that organization, whether and when employees influenced local policies and political institutions, and how policies diffused.The ICMA data we have gathered could serve as a foundation for future work.
There is also a need to better understand the role government employee organizations played in the passage of state public-sector collective bargaining laws during the 1960s, 1970s, and 1980s-another critical juncture in this broad transformation of government.While there is some existing research that examines this (e.g., Hartney 2022;Saltzman 1985), it primarily focuses on teachers and the development of teachers' unions.Our data show that some organizations of local government employees, such as the IAFF and AFSCME, were well organized (and considered themselves unions) long before the NEA shifted its position in favor of unionization and collective bargaining.Slater (2004) documents that some of these early unions were instrumental in securing the passage of the nation's first state public-sector collective bargaining law in Wisconsin in 1959.Moreover, as early as the 1940s, employee organizations in many cities managed to secure better wages and working conditions through informal negotiations and verbal agreements with government employers (Slater 2004).That government employers sometimes made and upheld verbal agreements with them even when collective bargaining was illegal suggests that government employee organizations had considerable political clout in many cities well before the 1960s.
As it stands, though, the key takeaway from our study is that early in this transformation, civil service provisions were extremely important to municipal employees, and an understanding of the shift to civil service is incomplete without considering their role.Going forward, there are several promising directions for future research.Scholars should further examine the role of these early interest groups in shaping the terrain of American politics and policymaking.Especially critical is more research on the policy and institutional contributors to the transition of the United States from a patronage-based system to one with an independent, influential bureaucracy-including in the nation's tens of thousands of local governments.

FIGURE 1 .
FIGURE 1. 1934 Political Cartoon Featured in the International Fire Fighter Magazine

FIGURE
FIGURE 2. Year of Municipal Civil Service Adoption

FIGURE
FIGURE 3. Year of Municipal IAFF Establishment

FIGURE 4 .
FIGURE 4. Proportion of Cities with Civil Service by 1940, by Population Size and IAFF

TABLE 2 .
Municipal Civil Service by 1940

TABLE 1 .
Descriptive Statistics, by Civil Service Status

TABLE 4 .
Mining Employment and Early IAFF Organization