Abstract
Participatory budgeting is described as a direct-democracy approach to resource allocation decision making. Theories assume it changes how public resources are spent by moving decisions from elected officials to citizens. The literature does not consider how earmarking—in which legislators direct parts of public budgets directly—might affect the impact of such policy devices. New York City’s participatory budgeting process which uses earmarks is analyzed to determine spending changes. Officials involved fund more projects at lower average amounts than those not involved but do not change the areas of funding, all of which is expected in systems of budgetary earmarks controlled by legislators.
Keywords
Generally speaking, participatory budgeting (“PB”) is a process that “allows the participation of non-elected citizens in the conception and/or allocation of public finances” (Sintomer et al., 2008, p. 168). This approach to citizen-engagement in the public budgeting and allocation process began in Brazilian cities, and has since spread to more than 3,000 municipalities (Su, 2017) in at least 15 different countries (Goldfrank, 2012). PB is increasingly popular as a concept among social reformers, subnational governmental elected officials, academics, and other researchers. Despite its popularity, Wampler (2012) notes that while PB can empower citizens, enhance democracy, and increase overall well-being, PB can also act as a way for governments to co-opt activists pushing for increased democratization.
Miller et al. (2017) note that scholars have labeled many diverse practices under the umbrella of “participatory budgeting,” which obscures differences between forms in various locations. Despite this definitional challenge, most scholars assume that PB influences how public resources are allocated because of the inclusion of private citizens in the budgeting process. In other words, the current literature assumes that PB fundamentally transforms public budgeting processes, at least for the resources subject to or included in the process. In its earliest implementation, this transformation redistributed the benefit of public resources from wealthier districts to poorer districts (de Sousa Santos, 1998).
This article tests this assumption that PB changes public spending. An alteration in public spending would signify a genuine transfer of significant decision-making power from elected officials to citizens through the PB process. Alternatively, PB might be a symbolic policy device that co-opts residents, deflecting their desire for budgetary influence through symbolic decision-making only. PB is analyzed through a quantitative case study of New York City using a natural experiment where PB for discretionary capital expenditures was implemented in stages across council districts beginning in 2011. This article also clarifies that budget reallocations can be of three types: size and number, locational, or functional—distinctions that have not been articulated in the empirical assessments of the effects of PB to date.
The empirical results indicate that districts adopting PB fund increased numbers of capital projects at smaller average amounts compared with districts that have not adopted PB. This finding is consistent with (but not exclusive to) the understanding of PB as a form of political patronage, in which elected officials spread public financed largesse more widely throughout their respective districts. This occurs because the PB money in New York City comes from, and is fundamentally controlled by, legislators. The PB process in New York City does not change this reality. Also, as a result of this design, funds are not redistributed between districts. Additional analysis finds that participatory budgeting in New York City (PBNYC) does not reallocate capital spending between functional categories; for example, it does not shift capital spending from parks to health. Because city councilmembers retain significant discretion in how capital spending is allocated even in the presence of PB in their districts, the New York City version of PB is best understood as a policy tool that lacks a functional allocative effect.
The rest of the article is organized as follows. The next section summarizes the literature relevant to how and why PB is expected to change spending in public budgets. We situate PB within the context of public budgeting, which fundamentally allocates resources. The existing literature primarily situates PB within the context of democracy instead. The section that follows analyzes this assumption of allocative effect and distinguishes between different types of budgetary allocative effects. Details of the New York City variant of PB process are then described, with specific testable research hypotheses articulated. The data used in the empirical analyses are then described, as are the research methods employed. The empirical results are then presented and results discussed. The article then concludes and offers recommendations that might lead New York City’s PB to a more functionally reallocative direct-democratic process.
PB and Allocative Changes to Public Budgets
The participation of citizens in governance is a defining feature of democratic regimes worldwide. However, this participation usually involves the selection of representatives who themselves make collective decisions. In fact, much of the current work on citizen participation focuses on direct democracy, in which individuals vote for or against specific laws or changes to existing laws (Hong & Cho, 2018).
One variant of citizen participation is PB, which has taken many forms in many settings. Broadly speaking, though, PB allows the participation of non-elected citizens in the conception and/or allocation of public finances. In principle, PB allows citizens or residents to vote for specific public projects, which governments then implement and often include a value-based component emphasizing social justice (Wampler & Hartz-Karp, 2012). The inclusion of citizens in the budgetary process introduces actors that may have different goals, such as the redistribution of public resources toward poor citizens, compared with politicians’ goals, such as maintaining control of budgetary processes and resources for patronage and clientelist purposes. Because of citizens’ differences in goals relative to elected officials, PB is widely expected and assumed to alter budgetary outcomes because citizen participants alter the traditional budgetary calculus and reduce the use of patronage in public budgets.
Shah (2007) defines PB as a “direct-democracy approach to budgeting” (p. 1) that has the benefits of (a) educating and engaging citizens on the processes of governance; (b) improving delivery of services by increasing transparency, accountability, and efficiency while also tackling corruption and clientelism; and (c) providing opportunity for marginalized groups to directly participate in budgeting decisions. This last characteristic is particularly relevant because it implies that one benefit of PB is a change (immediate or gradual) in where public moneys are spent because of a change in who makes significant budgetary decisions (residents vs. public officials).
Fung (2006) introduced a framework for understanding public participation in governance structures in general. The framework’s dimensions include the participants in the process, how communication and decisions are made in the structure, and how participation influences public policies (i.e., the amount of impact). Miller et al. (2017) further separate the communication and decision-making domains into two separate dimensions, because they can vary independently from each other. Nevertheless, Fung’s (2006) “Democracy Cube” is a three-dimensional space in which participatory mechanisms can be located and compared with each other and other public decision-making arrangements. The framework can be used to determine how effective or not different mechanisms are for addressing different democratic concerns.
The most relevant portion of Fung’s (2006) framework for this article is the domain focused on how public participation influences actual policies. Fung (2006) opines that the distinctive feature of PB is that excluded actors (poor people and other excluded groups) are now included in allocative decision making. However, the framework does not regard what is arguably the most important consideration related to PB as an institution—namely, does it change how public resources are allocated?
One of the reasons the first PB experiment in Porto Alegre, Brazil, was considered revolutionary was that it changed not just “where” public money was spent (spatially) but also on “what” the money was spent (functionally). While more money flowed toward poor neighborhoods, the Brazilian experiment also resulted in public spending reallocated to basic sanitation and health care from other public functions (Gonçalves, 2014). The process was an improvement from a social justice perspective compared with the prior clientelist system in which communities received projects largely as political rewards only (Souza, 2001). Boulding and Wampler (2010) analyze data from 220 Brazilian cities and observe that PB is associated with increased spending on health care and small decreases in poverty. The early Latin American experience with PB fundamentally shifted a significant share of public resources from serving the well-off to serving the poor.
Ganuza and Baiocchi (2012) distinguish between PB as a policy instrument in its early stage—a way to alter how citizens, civil society, and the government interact—and PB as a policy device in its later stage—a simple neutral initiative that may improve governance but not shift budget allocations. Baiocchi and Ganuza (2014) find that some processes in Latin America and Europe included direct decision making, while others were merely consultations (similar to Fung’s second domain); these distinctions in the process clearly influence the ability of PB to alter budget allocations by shifting decision making more directly to the people as conceived in Brazil. Allegretti and Herzberg (2004) and Ruesch and Wagner (2014) find little allocative effect from PB in Germany, perhaps because of the asymmetry in government officials’ versus citizens’ power as well as its intent to not redistribute toward more marginalized populations. Bassoli (2012) finds similar limited allocative effects from Italian PB.
These examples highlight what Su (2018) notes—the tension in direct-democratic experiments (such as PB) between empowerment of citizens and co-optation by politicians. Such co-optation might suggest the use of PB processes to fulfill the needs of elites and politicians rather than the citizen participants themselves (i.e., hinder or slow down the shift of public resources from the well-off to poor). Grillos (2017) notes that politicians in Indonesia use PB as a patronage strategy, in which small, dispersed projects are preferred over larger ones. This would suggest that Shah’s (2007) assumption that PB breaks the link between public budgeting and patronage is not universally valid.
However, other PB efforts have seemingly altered public spending. Hong and Cho (2018) focus on the outcome of PB in South Korea by examining whether the process alters the placement or location of surveillance cameras between neighborhoods. They find increased resources (cameras) allocated to poor communities when PB is used than when not used. Similarly, Shybalkina and Bifulco (2019) examine whether PBNYC changes the distribution of projects within city council districts. They find evidence that PB does shift funding to lower income census tracts (although not the lowest income census tracts) within New York City council districts. However, because PB operates within New York City council districts, this implies a shift of resources, at best, from between slightly different economic groups, whereas the early Latin American experience saw a more explicit shift from rich to poor.
Analyzing Whether PB Alters Public Spending
The extant literature provides mixed results about whether or not PB actually affects the allocations of public budgets. The empirical evidence shows variation by location (country), process (consultative vs. direct democracy), and amount of budget subject to citizen participation (large vs. trivial). Hence, whether PB affects actual budget allocations seems to depend upon these various factors and should not be assumed a priori. Heterogeneity of PB forms seems perhaps more relevant for predicting outcomes than whether or not PB simply exists in some governance structure of some locality.
Much of the current literature does not consider existing public budgeting systems when predicting expected outcomes. However, PB processes inevitably involve allocating resources derived from citizens (from taxes, fees, and other means) and is almost universally included in routine public budgeting processes. Specifically, the current literature does not consider how the earmarking process—in which legislators direct parts of the public budget directly—might affect the impact of PB. Wu and Williams (2015, 2017) differentiate between programmatic earmarks for a purpose from a specific revenue source and legislative earmarks in which legislators direct money within their districts. Most literature on state and local earmarking relates to the first form; however, PBNYC and many other areas is linked to the second form. Within the context of PB, whether funds for projects derive from a revenue source with no political actor influencing the outcome or from earmarks controlled by legislators is a potential unexplored mechanism for understanding the budgetary effects of PB. This is the gap in the literature we inform.
Furthermore, different studies analyze different types of reallocations. The extant literature suggests at least three types of public budgeting allocation change possibilities resulting from PB:
Reallocation by size and number of projects—in which resources are distributed more broadly but in smaller amounts;
Reallocation by location—in which the sites of public spending change;
Reallocation by function—in which resources are shifted between specific programs (e.g., shifting public resources from parks to health as in the early Brazilian experiment).
The first type of allocative change may result from a fixed amount of public money being split into smaller grants; these grants can then be distributed to more beneficiaries. While this may provide funding to more beneficiaries (thus reflecting an empowerment objective), it may also reflect use of PB by elected officials and elites to fine-tune their use of patronage to gain votes, which is contrary to the stated anti-clientelistic purposes of participatory process. The second type of allocative change is spatial and may be focused on the empowerment objective of PB, by redirecting funds from wealthier to poorer areas. The third type of allocative change might be caused by citizen preferences that differ from elected officials, like early PB initiatives, in which spending toward basic infrastructure was increased. Of importance, the first two types of allocation would be expected from a PB process from legislative earmarks, while the third type would be expected from a process that derives from a more apolitical one. These distinctions based on budgeting processes are absent from the current literature.
The Case of PBNYC
This article analyzes New York City’s PB system and develops specific testable hypotheses about whether PB results in smaller but numerous grants to beneficiaries (which is consistent with earmarking and patronage because the funds are controlled by legislatures), or whether it shifts public spending by function (which is consistent with programmatic earmarking). New York City is one of the only municipalities that has a large and expanding PB system that also makes its data transparent and available, permitting empirical analysis. 1 Single cases such as this are the norm in PB research due to variation between municipalities’ systems as well (see Hong & Cho, 2018; Shybalkina & Bifulco, 2019; Su, 2018, as recent examples). Nevertheless, results from these studies may not generalize to other municipalities.
History of PBNYC
Four city councilmembers initiated PBNYC in the fall of 2011 during planning for fiscal year 2013. Councilmember participation in PB has since expanded; in the 2016–2017 cycle, 31 of 51 members committed a total of US$40 million (New York City Council, 2017b). The committed amount represents the amount allocated in the capital budgeting process, not the amount spent during the fiscal year. The management of the process has also evolved over time as well. Initially, the process was coordinated through a cooperative agreement between civil society organizations and the participating councilmembers, resulting in a citywide steering committee. This arrangement extended ownership of the process beyond elected officials to include civil society (Jabola-Carolus, 2017). After the election of a new speaker in 2014, the council devoted more staff and resources to citywide coordination and support of PBNYC, which had two effects: increasing resources for coordination and replacing civil society participation with city government staff. Gilman (2016) finds that many of the more than 40 original organizations participating in the steering committee dropped out as the council took more direct control. Hence, PBNYC is a tightly controlled, top-down process unlike most of the early versions in developing nations.
PBNYC Funding
PBNYC funding is directed through city legislators. All city councilmembers receive an annual fixed amount of money which they are permitted to allocate at their own discretion. 2 This earmarked amount is US$5 million annually and generally limited to capital spending only. These discretionary earmarks are not part of the operating budget but are instead considered add-ons to the larger capital budgeting process. All councilmembers who participate in the PB process commit to using US$1 million (or 20%) for PB and retain the remaining US$4 million of funding for their own discretionary allocations. The final amount allocated to PB might be slightly above or below the US$1 million threshold depending upon which projects are selected by the public within the council district.
Expectations
It is anticipated that the councilmember’s allocative preferences are across the entire US$5 million discretionary earmark amount. Therefore, it is expected that any redirection arising from PB that does not reflect the councilmember’s allocative preference will be balanced out using the remaining funds available to the member. This provides councilmembers the opportunity to balance capital spending in their districts to reflect their own preferences regardless of the programmatic focus selected by their constituents through the PBNYC process. By holding back a large share of funds, councilmembers are able to balance their total allocations to maintain their overall preferences and retain significant decision-making about public spending. 3 This leads to the principal empirical question examined here: “To what degree do councilmembers use their earmarks—including PB—as part of a political strategy?” In this particular context, councilmembers might spread “good will” to more engaged constituents by funding multiple small projects rather than a limited number of larger projects—that is, reallocate as a patronage strategy. This hypothesis would suggest that (a) the average size of the member-directed capital expenditures will be smaller for members engaged in PB compared with those who are not, and (b) the number of funded projects will be larger compared with member districts not participating in PBNYC. These smaller and more numerous projects allow the councilmember to provide funding to an increased number of local not-for-profit providers or an increased number of local public agencies, resulting in more widely distributed publicly financed patronage for his or her district. Such distribution might be because politicians and local agencies build and maintain patronage relationships within council districts (Marwell, 2004). Alternately, councilmembers might become more aware of community needs through the PB process.
On the contrary, if the PB funds are actually independent from the legislative earmarking process once a councilmember commits the funds to it, then we would expect to see a reallocation of the third type—reallocation by function—because the new budgetary actors brought in through the PB process are likely to have preferences that differ from elected officials. Carnes and Lupu (2015), for example, demonstrate that in Latin America, lawmakers from different classes have different attitudes about economic priorities. Corvalan et al. (2016) state legislatures with more class heterogeneity are correlated with increased government spending in welfare and education. In India, Pande (2003) finds that including disadvantaged minorities leads to increased public spending on programs from which these minorities might benefit; similarly, Bhalotra and Clots-Figueras (2014) determine that increased female representation leads to increased spending on programs that benefit women. In a similar vein, increasing participation in budgeting might shift allocative funding.
The focus here is fundamentally an examination of whether PBNYC shifts budgetary decision making from elected officials to residents directly or not, whereas the first hypothesis assumes elected officials retain control of the funds despite citizen input through PBNYC. Across all city council officials, roughly half of their assigned discretionary capital spending is directed toward education and nearly one fifth toward parks. Six other categories receive only about 2% (New York City Council, 2017a). For this analysis, a functional change refers to a change in shares of expenditures assigned to these specific categories.
Data and Method
The data for our main dependent variable—council district capital appropriations—come from NYC Open Data, which is a portal that gives access to municipal agencies’ data. 4 Data from budget adoption years 2008 through 2016 on all 51 city council districts were collected. The raw dataset contains 13,007 observations and key variables such as the sponsors’ names, budget year, adoption year, project title, award recipient (i.e., the name of the organization or agency), the amount of the final grant, category, and so on.
Each observation represents a capital project that sometimes has multiple sponsors and, therefore, multiple council districts involved. This presents challenges because the unit of analysis is a council district, not a capital project, which necessitates separating multi-sponsor projects into individual sponsors so that each sponsor can be linked to a district. To that end, capital projects that have no more than two individual council sponsors are identified. Projects co-sponsored by groups such as the Jewish caucus or the lesbian, gay, bisexual, and transgender (LGBT) caucus are difficult to link to any specific district(s) and are therefore excluded from the analysis. The sample is further restricted to projects with at most two sponsors because when two members co-sponsor a project, they are likely (but perhaps not always) to be equal partners in the effort. With that assumption, the project amount is equally divided between the two sponsors. With three or more sponsors, it is difficult to assign the total project amount to individual districts (sponsors) with reasonable accuracy, and advice from council staff indicates that these projects reflect use of funds controlled by the speaker.
After dropping observations with more than two co-sponsors and those with any number of non-individual sponsors, the final dataset contains 10,711 observations on capital projects—more than 82% of the total original projects. Because the aim is to construct panel data on council districts, each capital project observation then needed to be linked with a particular council district, and this link was absent from the raw capital projects dataset. Using Google search and sources like Wikipedia, councilmembers’ websites, and the New York City Council website, a list of councilmembers by district for each year from 2009 through 2017 was assembled, and then merged with the main dataset on capital projects. Each of the 10,711 capital project observations were thus linked to a council district.
The original dataset classifies projects into 29 discrete spending categories that were consolidated into nine functional areas by collapsing similar types. The nine categories created were as follows: social services, education, culture and library, transportation, health and environment, housing, public safety, parks and recreation, and economic development. The original dataset also has one category labeled “public buildings” that does not belong to any of the nine created categories because “public buildings” is not a function. Therefore, 278 “public buildings” observations were manually reclassified into one of the nine functional categories using project description information contained in the raw data. Table 1 presents a summary of old and new categories, and displays the percentage of observations in each category. As discussed in more detail below, the share of spending on these categories, the total spending in each category, average size of projects, and number of projects are used as dependent variables in the empirical analyses.
Functional Categories of Capital Spending and Share of Observations.
CUNY = The City University of New York.
The main explanatory variable of interest is a dummy variable indicating participation in PB. The variable is set to “1” if a district’s representative participated in the PB process in a particular fiscal year, and “0” otherwise. The “participatory budgeting projects” dataset from NYC Open Data is used, which provides details on the project title, the amount involved, council district, year, and so on. 5 The vote year and council district variables are used to create the explanatory dummy variable. The dataset reports only those districts in each year that had at least one project voted on through the PBNYC process. The dummy variable takes the value “1” for all district-years reported in the dataset, and “0” otherwise. Because the first cycle of PBNYC was completed in 2012, the dummy variable is set to “0” for all districts from 2008 to 2011.
Additional variables are included in the empirical models to control for non-PB-related factors that may influence capital spending decisions within council districts. Population density is included to control for potential heterogeneous PB voting behavior and preferences between thickly and thinly populated districts. Population data are obtained from Infoshare.org, which is managed by Community Studies of New York, Inc., a not-for-profit organization based in New York City. Infoshare.org provides population data extracted from the American Community Surveys (ACS) which are ongoing surveys conducted by the U.S. Census Bureau reporting “period estimates” on many demographic variables. Infoshare.org summarizes data at the NYC council district level using the ACS 5-year estimates for 2005–2009, 2006–2010, and so on, through 2011–2015. A 2008–2013 panel of population estimates by council district is constructed by assuming that each 5-year estimate best represents the middle year in the period. Specifically, it is assumed that 2006–2010 estimates best represent the year 2008, 2007–2011 represents the year 2009, and so on. Then, the population estimates are divided by area in square miles (excluding water) of each council district (obtained from NYC open data 6 ) to calculate population density. An interrupted time series of these demographic data from 2008 to 2016 is created by linearly extrapolating the 2010–2013 trend all the way to 2016.
Demographic variables on council districts related to race, the share of seniors, median household income, and homeownership rates are also included to control for potential differences in PB voting between groups as well as heterogeneous preferences for capital spending. 7 It is worth pointing out there is no omitted category of race because our race variables are continuous rather than dummy variables. When the unit of analysis is a person, race must be coded as dummy or categorical variables because an individual can, for example, either be Hispanic or Asian, which necessitates an omitted or reference category. Our unit of analysis is a council district which can be part Hispanic and part Asian, and thus an omitted category is unnecessary.
The data for all demographic control variables come from the ACS 5-year estimates via Infoshare.org. Similar to the approach with the population variable, a 2008–2013 panel is constructed by assuming each 5-year estimate best represents the middle year in the period. In other words, the ACS 2006–2010 estimates for the 2008 panel are used, 2007–2011 estimate for the 2009 panel, and so on. An interrupted time series from 2008 to 2016 is created by linearly extrapolating the 2010–2013 trend to 2016.
Political party affiliations are also included to control for potential heterogeneous preferences across political ideologies related to capital spending. The data come from the website of the City Board of Elections which provides district-wide summary of voter registrations by political party from 2011 through 2017. 8 The percentage of Democrats and Republicans are included in the models as continuous variables which is why there is no omitted category. Again, an interrupted time series from 2008 to 2016 is created by linearly extrapolating the 2011–2016 trend back to 2008. All these variables are finally merged into a single dataset by matching on council district and year. Then, observations that have the same year and council district are combined together. The final data are reduced to 459 district-year observations, which is the product of 51 council districts over 9 years (from 2009 to 2017). Financial data are adjusted for inflation to reflect constant dollars.
The regression model used to analyze whether PB shifts capital spending allocations is
where Dependent Variable n represents (a) average expenditure per project, (b) count of projects, (c) the share of spending on each project functional category for district i in year t, and (d) total expenditures for a functional category for district i in year t. 9 The first two variables capture the earmarking patronage hypothesis, and the second two the shift in budgetary decision-making hypothesis. Control variables are lagged by 1 year because capital projects are selected in the prior year for funding in the current year. Fiscal year effects are also included to control for unobserved trends that equally affect all council districts. This research design uses a time series of average spending per project, the number of projects in a district, and the functional allocation of spending to determine whether councilmembers participating and non-participating in PBNYC show differences in their earmarking.
Table 2 lists the variables and definitions used in the empirical analyses.
List of Variables and Definitions.
Empirical Results and Discussion
Descriptive Statistics
Table 3 reports summary statistics for the variables before and after districts implemented PB. The table reports that the data contain 31 council districts that adopted PB during the study period. The important variables to examine are project count, average amount, and the percentages spent on particular functional capital spending categories. Before PB, districts averaged about 22 projects funded by a city councilmember’s discretionary grants, and this increases to over 27 after adoption of PB. At the same time, the average size of the grant declined from nearly US$319,000 on average to nearly US$230,000 on average.
Summary Statistics: Only Districts That Implemented PB.
Note. PB = participatory budgeting.
In terms of capital spending categories, the average council district allocated only about 1% of capital spending to public safety before PB, while the largest share of capital spending was devoted to education on average—at about 34%. Following adoption of PB, these districts average about 38% of capital spending on education. Parks and recreation spending is relatively unchanged before and after adopting PB.
Model Results
Table 4 reports the results addressing the primary research question about whether PBNYC, because it is part of the earmarking process of the budget, reflects a patronage strategy by politicians. As hypothesized, PBNYC might permit councilmembers to reduce the average size of PBNYC-funded projects and increase the number of funded projects. As shown, the average size of capital projects declines by more than US$61,000 following a council district’s adoption of PBNYC. The results further indicate that districts increase the number of funded capital projects by three per district per year following the adoption of PBNYC. PB does seem to alter the number and size of capital projects in districts per year. Whether this reflects legislators’ seeking to spread largesse throughout their districts to help with future elections or whether it is a response to newly discovered needs through the PB process (i.e., empowerment) remains an open question. Interestingly, the results provide evidence that increased enrollment with the two major political parties results in more capital projects in a council district, perhaps because of better group coordination. However, future research may want to investigate the role of political parties and citizen enrollment on capital spending decisions.
Effect of Participatory Budgeting on Average Capital Expenditure per Project (U.S. Dollar in Thousands) and Number of Capital Projects.
Note. Robust t statistics in parentheses.
p < .1. **p < .05. ***p < .01.
Table 5 reports the results testing whether PBNYC altered functional spending categories. The coefficients are not statistically significant and find no evidence that participating councilmembers change their overall discretionary capital expenditures between categories. 8
Effect of Participatory Budgeting on the Share of Functional Categories in Total Capital Spending.
Note. Robust t statistics in parentheses. Constant term included but not reported.
p < .1. **p < .05. ***p < .01.
The results in Table 5 also show that as a district’s share of seniors increases, capital spending increases for education (about 1 percentage point for each percentage increase in seniors), while spending on parks and recreation decreases. Furthermore, as the share of owners in a district increases, capital spending on housing in the district is reduced; hence, the results seem to indicate that owner-occupied housing is associated with less public housing investment in a district. The coefficients on the various race and ethnicity measures provide marginal evidence that increases in each category (while holding others constant) are associated with declines in capital spending in education and increases in cultural and library spending; the coefficients on percent White and Black also provide marginal evidence of increased parks spending. The coefficients on the two major political party enrollment variables provide evidence that districts with more of either major party shift capital spending from parks to housing. The results in Table 5, while not finding evidence of PB shifting capital spending functionally, do suggest heterogeneity for capital demand between various constituencies, which might be worth future consideration in the public budgeting literature.
Spending per category measured in dollars (rather than as a share) is also analyzed. If the amount of capital spending is changing over time, shares may remain constant even though spending in specific categories is changing in absolute terms. The results of this analysis are presented in Table 6. PARTICIPATE is significant and negative, which provides evidence that districts reduce capital spending after adopting PBNYC; however, the individual categories are generally not significant which is evidence that PBNYC does not affect spending. 10 The coefficient on housing is negative but only significant at the 10% level. Taken with the results in Table 5, this might suggest that capital spending on housing is reduced following the adoption of PB, but the change is relatively small and not enough to significantly alter the overall functional spending categories. The results in Tables 5 and 6 together provide evidence that PBNYC does not change capital expenditure decisions at the district level—as a share of spending or in dollar terms (generally). Councilmembers may fund capital projects as determined by PBNYC participants, but other projects desired by the elected official do not suffer or change as a result. PBNYC does not have an allocative effect (in terms of changing which functional categories receive public spending) in New York City unlike in some other cities.
Effect of Participatory Budgeting on Total Capital Expenditures by Category (U.S. Dollar in Thousands).
Note. Robust t statistics in parentheses.
p < .1. **p < .05. ***p < .01.
The results in Table 6 support many but not all of the findings in Table 5. Districts with greater share of seniors devote more capital spending to education (US$83,000 more on average for each percentage increase) and less on parks and recreation (US$62,000 less on average for each percentage increase). Greater concentrations of owners also lead to more capital spending devoted to education and less to public housing. These findings are consistent with the link between housing and local education quality (or perceived quality through increased capital spending). However, the coefficients on the variables associated with racial and ethnic categories are not significant. Furthermore, the coefficients on the two major political parties’ variables find evidence of increased capital spending on housing; the coefficient on percentage Republican is also marginally significant and negative on parks capital spending, as in Table 6.
Robustness Checks
An important empirical concern in pooled ordinary least squares (OLS) models is that the effect from some omitted variable is attributed to the decision to adopt PB by a councilmember. Concerns about omitted variable bias may be alleviated through the use of a model that employs council district fixed effects to control for observed and unobserved characteristics. Such a model is estimated, and results are included in Table 7. The results are comparable with those in Table 4 in which size of projects is reduced while number of projects increases. Employing a fixed effects model to analyze functional allocations leads to qualitative similar results to those reported in Table 5.
Effect of Participatory Budgeting on Average Capital Expenditure per Project (U.S. Dollar in Thousands) and Number of Capital Projects.
Note. Robust t statistics in parentheses.
p < .05. ***p < .01.
One possibility is that the functional allocative effect from PB (Tables 5 and 6) only occurs the longer a district uses PB. The PARTICIPATE variable is redefined as the number of fiscal years since PBNYC was first adopted in a particular district. The equation is then re-estimated using the functional spending categories as dependent variables. The results are qualitatively unchanged.
Finally, the unit of analysis is changed from district-year to simply district-pre-PBNYC and district-post-PBNYC by collapsing the data into means. This approach would suggest that PB influences council districts marginally over longer time periods rather than annually. This argument would suggest that PB has only small incremental budgetary effects that take many fiscal years to be observed. The results are unchanged.
Discussion and Conclusion
PB holds the promise of increasing input from residents marginalized from official budget decision making, thereby increasing their roles in determining how public money is allocated. Fundamentally, PB was envisioned as a means to shift decision making downward, which would change how governments spend money. Much of the extant PB literature assumes that including regular citizens in this process will lead to allocative effects because of differences between the demands of citizens and elected officials. In this article, we address a gap in the PB literature by considering whether legislative earmarking in the PB process might counter this assumption, reducing the potential influence of the public. Furthermore, such a consideration might actually lead to PB becoming a patronage strategy in its own right, as legislators distribute an increased number of smaller projects across their districts.
Furthermore, the current literature does not distinguish between different types of budgetary allocative effects. PB might affect how many projects are funded, the locations of funded projects, or to which functional categories public spending is devoted. While the earliest advocates of PB believed that it was a means to decrease the clientelism and patronage that has accompanied traditional public budgeting in the past, more recent analyses have begun to consider whether elected officials and the elites might co-opt the PB process for their own purposes.
This article draws upon these existing theories and develops testable research hypotheses in the context of New York City’s PB and overall budgetary process, and asks whether PB as earmarking (i.e., as spending directed by legislators) will affect the distribution of public funds. Much of the current literature assumes PB will alter these allocations because of a shift in decision making in the budget process from elites to citizens. We find that councilmembers use PBNYC primarily as a way to distribute an increased number of smaller projects in their council districts, but that functional spending is left unchanged. That is, rather than PB resulting in functional allocative changes from direct democracy, members simply allocate only a small fraction of discretionary capital spending to PB, fund more projects with smaller amounts than before, and leave overall spending categories unchanged. These findings suggest that PB does not eliminate patronage in public budgeting by public officials; rather, PB alters how this patronage may occur.
The analyses here can definitively argue that PB does not affect what category of spending is funded in the average district each year. However, Shybalkina and Bifulco (2019) do find that PBNYC influences the spatial distribution of projects within districts (i.e., where projects are funded). Therefore, while PB in NYC may not have an allocative effect on capital spending between functional categories, it may reallocate projects within districts by location. In fact, their finding in conjunction with support for the patronage findings in this article suggest that councilmembers who use PBNYC may seek to curry favor with voters and other important district stakeholders not by altering what is funded, but where it is funded and how many projects are funded. Future research should continue to disentangle these different effects from PB so that we can better predict what shortcomings in current public budgeting practices PB can address, and which it cannot. Future research should analyze different forms of PB to determine whether these differences lead to outcomes that vary from the New York City experience. Furthermore, the data do not allow visibility into whether PB changed spending within (rather than between) functional spending categories. For example, one possible outcome from PB is that “education spending” might be redirected between purposes. However, the data made public by the city cannot provide this detail and require additional data. Recent research finds that New York City public schools have increasingly shifted spending toward facility spending rather than classroom spending (Rothbart, 2016), and it would be important to determine whether PB had a role in this shift.
This article focuses on the changes to capital spending resulting from PB. Next steps in this analysis might involve determining whether smaller but more numerous capital projects meaningfully improve public service delivery or citizen satisfaction with these services. Additional research might look at what the changes found here in number and size of projects have any effects on previously marginalized populations. Such research would also inform the general public budgeting literature by linking outcomes with inputs.
The analyses here are mostly suggestive of the patronage role in the New York City PB process. That is, the results presented here are consistent with those predicted by patronage. However, the empirical tests and results cannot definitely rule out the empowerment function of PB. At the very least, the analyses presented here question the dominant story about PB empowering citizens. Future research with different and more nuanced about the project selection process than that currently available is required to further refine knowledge about the New York City PB process.
Given the findings here, a natural question to ask is what would need to change for PB to operate as a clearly transformative rather than patronage strategy in New York City’s budget process? Most basically, the current system lacks transparency around how projects are selected for the ballot and who is involved. It is difficult to obtain information about the times and locations of meetings about these. Delegates who cull these projects for the ballot are selected from volunteers by councilmembers. The process could be made more transparent by making these delegates more representative of the districts they represent, and making the process of project selection and inclusion clear to the public. Some proponents of PB state that even if PB has no real effect on resource allocations, it serves a public service through education of the public about the budgeting process. However, for this to really be meaningful, civic engagement needs to be earlier in the process and not just on voting for projects approved by the legislators or their proxies. If PB could be made truly representative and transparent, then a case might be made to expand PB and not restrict it to such an immaterial portion of the budget. Recently, PB was expanded into New York City high schools (with extremely small amounts of money at stake); participating schools are given US$2,000 annually for students to allocate (Lerner, 2018). The mayor touts this effort as a way of encouraging civic participation, but measuring whether PB accomplishes this goal will remain elusive. In 2018, a ballot referendum established a new citywide PB process under the guidance of a citywide commission beginning during calendar year 2020; thus, these recommendations may need to be adjusted as the new process is implemented.
Wampler et al. (2018) argue that PB is best understood as a policy tool that can lead to increased transparency, more accountability from elected officials, stronger civil society, and better outcomes. To the extent that NYC’s PB efforts do not shift capital spending toward marginalized communities’ needs, that elected officials permit only a tiny fraction of spending to be allocated this way, and that council members appear to use PB almost as a marketing tool within their districts, such desired changes are unlikely to emerge from the adoption of PB itself. Rather, the systems and institutions of PB seem key for unlocking the potential of PB across heterogeneous communities. Expansion of PB may avoid some of the potential patronage pitfalls discussed here if it is not under legislative control and is instead managed by an independent body.
Footnotes
Authors’ Note
The authors wish to thank Samuli Harju who contributed to an earlier version of this article.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
