The distributive politics of privately financed infrastructure agreements

Privately financed infrastructure agreements (PFIAs) are increasingly being used across the globe, bringing private money into the delivery of public goods. How does introducing private actors to such a process change how we think about distributive politics? I investigate this question using both quantitative and qualitative analyses, uncovering a relationship consistent with PFIAs being used as distributive goods and exploring how the credit-claim potential of PFIAs may affect their distributive use. My quantitative analyses (on 16 middle-income countries) present evidence suggestive of a relationship between electoral variables and the likelihood of a PFIA being present in a district. In districts aligned with the national ruling party, PFIAs are more likely to be concentrated in swing districts than core districts. I find that this relationship is more pronounced for PFIAs that are more directly attributable to the government. My qualitative press analysis provides insights into how politicians use various features of PFIAs to create credit-claiming opportunities.


| INTRODUCTION
The distributive politics literature has started to recognise that the type of good one studies matters.Patronage, understood as "the proffering of public resources […] by office holders in return for electoral support" (Stokes, 2007, p. 3), in short, will not always look the same across different types of distributive goods.Kramon and Posner (2013) make this point adeptly in their work, showing that "the conclusions one draws about who benefits from government allocation decisions can vary markedly depending on the outcome one happens to study" (p.461).I contribute to this emerging literature by exploring the degree to which incumbents are able to use a category of public goods that have not been studied in the distributive literature to date -infrastructure delivered with private financing -to further their electoral ends.
I study privately financed infrastructure agreements (PFIAs), which collect the financing, construction, and operation of assets into long-term contractual arrangements and are an appealing way for governments facing strict budget constraints to deliver public infrastructure.Since the 2008 financial crisis, such agreements between government and the private sector have become increasingly popular for high-, middle-and low-income countries as fiscal constraints have tightened (World Bank, 2016a).These agreements provide a way to finance large-scale projects that the government may otherwise be unable to provide (Yehoue et al., 2006); simultaneously allowing the government to push forward expenditures (Ball et al., 2001) and to claim credit ex ante for projects that may never be completed (Williams, 2017).Governments have been found to front-load investment obligations onto private partners in PFIAs, such that the government itself has essentially no up-front costs (Post, 2014).
PFIAs are an important means for delivering infrastructure in low-and middle-income countries. 1The World Bank counts over 8000 PFIA projects in such countries for the 1990-2020 period in it database of projects, representing investments of over US$2029 billion (World Bank, 2022).PFIAs have been estimated to represent around 10% of the approximately $1.3 trillion invested annually in infrastructure by developing countries (Fay et al., 2018). 2 Focusing on Latin America and the Caribbean (LAC) countries as they will represent a significant segment of the countries analysed here, it is estimated that between 2008 and 2016, commitments to infrastructure through PFIAs in LAC countries were around 0.6%-1% of GDP (World Bank, 2016b).Given that, on average in 2018 in LAC countries, governments spent roughly 30% of GDP (OECD, 2020) in total, up to 1/30 of that amount being spent on PFIA projects is quite substantial.
Infrastructure will be an essential part of low-and middle-income countries' ability to become more productive and enact meaningful economic and social change.It is argued that such countries must double current investment levels in infrastructure to have any transformational impact (UNCTAD, 2018).To date, PFIAs have enjoyed sustained support from international organisations, even if concerns about the fit of these privately financed projects with a global agenda on sustainable development are starting to emerge (UN DESA, 2016).Given that PFIAs are expected to be increasingly used in transitioning countries to deliver this essential infrastructure, a deeper understanding of the political determinants of the way in which PFIAs are distributed is required if these kinds of financing arrangements are to continue to be encouraged by the international community.PFIAs present a new theoretical challenge in that they possess features that differentiate them from other policy instruments used in distributive politics. 3When engaging in a PFIA, the government shares the risk and the cost of a project and brings private interests into its decision-making.The involvement of governmental procurement divisions with private firms means that the public rules of distribution are more likely to not be applied as they are intended (Stokes et al., 2013, p. 7).Extant literature shows both that there is significant scope for corruption in the various procurement stages of PFIAs (Bertelli, Mele, & Woodhouse, 2021;Iossa & Martimort, 2016;Klitgaard, 2012;Takano, 2017) and that private firms and governments are frequently able to renegotiate the terms of their contracts (Engel et al., 2009;Guasch & Straub, 2006).Both of these features mean that private firms have room to influence the way in which the public rules of distribution are enacted, shaping them to their preferences.By bringing a private party into the equation, PFIAs introduce a set of interests that do not necessarily align with those of political office holders.Frequently, business interests and political interests will not converge and, as such, one would expect to see less electoral targeting of PFIAs than projects delivered through full public provision.However, as my press analysis unveils, PFIAs can offer valuable credit-claiming opportunities for political office holders.First, incumbents appear to use the long duration of PFIAs to claim credit on multiple occasions for the same project and to do so in electorally important moments (such as just before an election).Second, incumbents adopt a rhetoric about efficiency gains through private involvement, associating themselves with positive concepts about growth, competitiveness, and economic wellbeing through their involvement in PFIAs.Third, they engage in these strategies before backlash occurs against the cost overruns and delays to which PFIAs are frequently victim (e.g., Angulo et al. (2020)).PFIAs, in short, represent valuable credit-claiming opportunities and my quantitative analyses suggest that incumbents successfully manage -despite diverging political and business interests -to exploit these opportunities and concentrate PFIAs in electorally important districts.
In what follows, I present a theory of PFIAs as distributive goods and establish how this builds on the existing literature.I outline how PFIAs can be used as credit-claiming opportunities by politicians through a qualitative press analysis.I then use quantitative analyses, on 16 middle-income countries, for which both PFIA distribution and district-level election data are available, to explore whether these credit-claiming behaviours translate into meaningful relationships between the allocation of PFIAs and district-level electoral variables in a systematic manner across countries.I provide the first exploratory examination of whether PFIAs are associated with electorally valuable districts and expand upon the scope of previous analyses by looking at the distributive politics of PFIAs on a cross-national level.As such, I build on extant theory and respond to direct calls for research into how public investment is shaped by credit-claiming incentives (Maskin & Tirole, 2008).
Whilst I cannot identify a causal relationship due to the nature of my research design, I find robust associations between the electoral features of a district and its likelihood of being allocated a PFIA and present evidence that such distributive patterns are stronger for more government-attributable PFIAs (i.e., with higher credit-claiming opportunities).In short, I make the case that PFIAs seem to follow the same distributive logics that affect goods delivered through full public provision, despite the private interests that shape them.This means that governments are, effectively, able to use private money with which to claim credit.

| Distributive goods
I understand a distributive policy to be a "political decision that concentrates benefits in a specific geographic constituency and finances expenditures through generalized taxation" (Weingast et al., 1981, p. 644).There are two models that have been explored at great length in the debate surrounding how distributive goods are allocated: the core voter model (Cox & McCubbins, 1986) and the swing voter model (Lindbeck & Weibull, 1987).The former argues that vote-maximising parties will concentrate distributive goods in districts where their core voters are, whereas the latter argues that they will allocate such goods to districts with high concentrations of political moderates who are likely to switch their votes according to the particularistic goods offered by the parties (swing voters).Dixit and Londregan (1996) combine these two models, arguing that if parties are equally able to allocate redistributive goods to all voting groups, then they will adopt the same tactic and target those groups with a high concentration of swing voters.Whereas, if the parties differ in their ability to allocate such goods according to their knowledge of a given group, then a machine politics outcome will obtain.
It is widely accepted that public works programmes are used as electoral tools.The nonprogrammatic distribution of public works projects can lead to either the over-provision or the under-provision of public goods due to rent-seeking, 4 both of which create inefficiencies to the disservice of the citizenry.Thus, although some types of nonprogrammatic distribution of public works projects can benefit citizens, normatively one would prefer programmatic distribution.

| PFIAs as distributive goods
There is an emerging literature that considers PFIAs as distributive goods (Bertelli, 2018;Maskin & Tirole, 2008;Post et al., 2017) and careful work has been done that considers typologies of hybrid forms of public good provision (Post et al., 2017), how the nature of a good affects the political logic of its distribution (Albertus, 2013), and how national leaders limit the supply of patronage goods to electorally valuable areas, using different logics at different levels of government.I add to this literature by bringing new actors into the theory of distributive politics and presenting evidence consistent with governments being able to use private money with which to claim credit.
I apply a distributive lens to the use of PFIAs to explore if they are distributed programmatically or otherwise.One would expect PFIAs to display less pronounced distributive patterns than public goods provided through full public provision as private interests must also be satisfied in this model of delivery and political interests may often not align with those of the business community.PFIAs are vulnerable to principal-agent problems in that the government and private contractor have "diverging incentives for both project cost and service quality in PPPs" (Siqueira et al., 2018, p. 2).The government must cater to private partners' profit-maximising interests in order to attract finance.
In a context of full public delivery, political elites face a trade-off between rent-seeking opportunities and gaining electoral support through the improvement of public goods (Bohlken, 2019).This trade-off is less heavily weighted towards private interests than in PFIAs where there is evidence that "the PFI tail wags the planning dog" 5 (Shaoul, 2005).That is to say, private interests have the opportunity to shape projects so that they fit the bill to be delivered via private finance (rather than full public delivery), thus potentially engineering highly profitable projects for themselves (Hodge & Greve, 2007).These risks become increasingly salient as the market for private partners becomes less competitive, leading to situations where private interests take precedence over those of the public (European Court of Auditors, 2018).
My hypotheses seek to explore whether the private involvement in PFIAs prevents incumbents from using them as electoral tools.However, they are limited by the data I have at my disposal (see Section 4).As such, they focus on electoral variables (alignment with national ruling party, margin of victory).The allocation of PFIAs can, of course, also be shaped by other variables which go beyond this paper, including the private sector environment in a given locality.For example, the composition and strength of networks of private partners (Bertelli, Woodhouse, et al., 2021) can affect the competition environment and the bargaining power of the government vis-à-vis the private sector.Thus, where a few firms dominate the market they will be able to dictate the terms of PFIAs more strongly than where there is plentiful supply from the private sector and in such settings we would expect allocational patterns not to follow electoral logics as incumbents are less able to impose their preferences on private actors.
My first set of hypotheses are derived from the literature on core and swing voter targeting, discussed above.The first hypothesis probes whether incumbents are able to concentrate PFIAs in those districts where they are in power.The idea here is to test, in as simple a manner as possible, whether incumbents are able to shape where PFIAs are allocated despite the competing private interests that they must contend with: H1. if an electoral district is won by the same party as the national governing party, it is more likely that a PFIA will be present there.
This hypothesis rests on the assumption that the party with the majority of votes -the governing party (or parties in the case of a coalition government) -has more influence over where public goods are allocated and that this party will endeavour to reward those districts that voted for it.
The second hypothesis examines how the margin of victory with which a district is won affects the distribution of PFIAs: H2. the margin of victory with which a district is won affects the likelihood of seeing a PFIA there.
This hypothesis is 'directionless' in that there are several ways in which marginality can interact with whether a district is aligned with the national governing party.In one scenario, office holders may seek to reward core districts where they win -and win historically -with a large margin.In another scenario, office holders may focus on rewarding swing districts where they only just won.What I am interested in is whether or not these electoral variables are systematically associated with the concentration of PFIAs across countries.

| Credit claiming and PFIAs
The distributive politics literature has shown that particularistic and targeted public expenditure is a method used by incumbent parties and legislators to build electoral support both in national and local contexts (Levitt & Snyder, 1997).Extant literature also provides evidence that allocating more resources to a district is causally linked to increases in constituent support for incumbent parties (Lazarus & Reilly, 2010).These models are based on the assumption that voters are able to associate public spending with the incumbent and then reward the latter for the expenditure directed to them.This mechanism relies upon political parties using credit-claiming messages to gain acknowledgement for spending (or bringing private investment to the area (Jensen & Malesky, 2018)), with legislators constructing messages that are designed to generate a belief that they were responsible for money being directed to a particular constituency.Some politicians are better campaigners, while others may also lack the resources or organisational capacity to inform voters of their pork-barrelling activities.The distinction between obtaining constituency resources and being recognised for attracting those funds is not merely semantic.Voters' decisions are based on individual perceptions that seem to depend less on how many resources the candidate actually generates and more on whether the candidate succeeds in providing information about her achievements.It is important to note that, when considering distributive goods, what matters for political agents is not necessarily the provision of the public good itself, but the credit-claiming opportunity that it represents.Events such as highway demonstration projects (Evans, 2004) where politicians can take credit for and ownership of a project are essential to show constituents their efficacy in bringing public goods to their district.
Unlike 'traditional' distributive goods, such as earmarked water projects with full public financing (Ferejohn, 1974;Maass, 1951), PFIAs are moderated by the bureaucratic bodies that manage them and the private financiers who make them possible.Such intermediary forces can make a difference to how goods are allocated (e.g., Bertelli and Grose (2009)).There are theoretical reasons to expect that different kinds of PFIAs will offer differing degrees of credit-claiming opportunity.Some infrastructure projects are more salient to voters than others, one need only think of the media buzz around the building of a new hospital and how that would compare with the updating or renovating of a waste management site.There are a plethora of reasons for which a PFIA may be more or less important to a certain group of voters: the jobs it brings to an area (e.g., a new local hospital), there being some controversy surrounding a certain type of infrastructure (e.g., building a nuclear energy plant), there being a particularly urgent need for a particular good (e.g., Covid-19 vaccine development and delivery).However, I focus on a specific feature of a project's salience that is comparable across national contexts, namely how attributable the project is to the government. 6I posit that the more 'public' or directly linked to the government a project is, the more valuable it is to an incumbent as a credit-claiming opportunity (Bertelli, 2018).This leads to my final hypothesis that: H3.PFIA projects that are more attributable to the government are more likely to display distributive patterns than less attributable PFIAs.

| QUALITATIVE EVIDENCE: THE CREDIT-CLAIMING POTENTIAL OF PFIAS
Before exploring whether there are aggregate patterns that are consistent with PFIAs being used as distributive goods (Section 4), I must first establish how they can be used as credit-claiming opportunities by politicians.To do this, I conduct a review of the available news coverage of the types of PFIAs in the countries covered by my sample.The goal here is to document possible ways in which politicians strategically use PFIAs to claim credit.This is important given that the credit-claiming opportunities offered by PFIAs have not been explored in the literature to date.This qualitative evidence paints a picture of how PFIAs can be used by politicians as credit-claiming opportunities. 7Note that the examples I present here are intended to illustrate possible ways in which PFIAs can be used as credit-claiming opportunities, that is not to suggest that alternative explanations for these media events are not at work or are unimportant.
My sample of countries for the quantitative analyses consists of those middle-income countries for which both PFIA distribution and district-level election data are available (more details: Section 4.1).For my qualitative analyses, I was unable to conduct a press analysis for all the countries in the sample due to language, time, and capacity constraints.As a result, I chose to focus on the Spanish language press, from Mexico, Costa Rica, and Colombia in particular, as Spanish-speaking countries constitute almost 40% of my sample.I also undertook a review of press coverage of PFIAs in Albania in order to present evidence from an Eastern bloc country as these countries also feature prominently in my sample (another 40% of the sample) and I wanted to ascertain whether the types of behaviours I found in Spanish-speaking countries were also found in these countries. 8 A first observation pertains to how politicians claim credit for PFIAs.They use inauguration ceremonies, press conferences, government visits, plant inspections, and public speeches to tie themselves to a project and make their contribution to bringing about the project known to the public.Moreover, high-level politicians are often involved in these credit-claiming events with presidents frequently being in attendance at inauguration ceremonies alongside multiple ministers.A compelling example is the case of a 'government visit' to the new Daniel Oduber Quirós airport terminal in Guanacaste, Costa Rica in 2012, where the Vice President of the Republic, the Public Works Minister, and the Tourism Minister were all in attendance (La Nación, 2012).Moreover, national politicians are often involved in such credit-claiming events even if they occur at the local level.For example, the Colombian press, commenting on Vice President Germán Vargas Lleras' credit-claiming behaviour, describes it as being "no longer surprising that the Vice President […] accompanies the Department's Governor […] to most events in Cundinamarca's municipalities.He was present at the signing of the petition for the Bogotá-Funza highway expansion; at the presentation of the Public Private Alliance (APP) […] at the inauguration of the first section of the Ruta del Sol, and a few days ago, at the opening of the Mancilla Reservoir.This week he announced, with 'great fanfare', a $2 billion investment to improve the Department's roads" (El Espectador, 2015).
An additional feature of how incumbents claim credit for PFIAs pertains to their use of language.There is a rhetoric about 'economic growth', 'market competitiveness', and 'efficiency' that is often employed when discussing PFIAs that can play into politicians' credit-claiming strategies.They can make claims about both relations with the private sector and efficiency gains that would not be possible with traditional procurement.A good example comes from Colombia in 2000, where the Transport Minister, Gustavo Canal Mora, highlighted the role of road concession PFIAs, such as the 'Vías para la Paz' programme, in "creating jobs" and "better conditions of competitiveness" for the country (La Republica, 2000).Grand claims are also possible given the large scale of many PFIAs that also allow politicians to signal the stability of their country or region to foreign investors.For example, the Costa Rican President, Luis Guillermo Solís, when inaugurating a hydroelectric plant PFIA in 2015, made claims about it contributing to creating "stable conditions for investors" and in "boosting economic well-being which generates jobs and enables us to work on the reduction of poverty in a transparent governmental exercise and in an open fight against corruption" (Agencia EFE -Servicio Económico, 2015).
Second, I observe that PFIAs offer various intertemporal opportunities to claim credit, thanks to the different stages of negotiation and contracting that they involve and their long duration.As such, a single project can provide multiple credit-claiming opportunities akin to the ribbon-cutting or ground-breaking ceremonies that politicians find so appealing in traditional public service delivery (Hirano et al., 2009).For example, the Prime Minister of Albania used the contract signing for the PFIA to restore Korça's old bazaar -at a high profile meeting where the Deputy Prime Minister, Mayor of Korça, Minister of Agriculture, were also present -as an opportunity to credit-claim in 2014 (Shekulli, 2014) and then also organised a site visit to inspect the ongoing works the following year, where the Deputy Prime Minister was in attendance and called the (unfinished) project "[a] promise kept by the Albanian government" (Shekulli, 2015).This type of behaviour occurs across PFIAs that cover a range of different sectors and geographies, but all have in common that the politicians in question claim credit for the projects before any construction work has begun, on the occasion of the signing of the PFIA contract.PFIAs are valuable electorally in that by signing a contract with a private partner they represent a more credible promise to deliver a public good or service than a simple campaign promise.A third observation involves the timing of the multiple credit-claiming events that politicians are able to engineer around PFIAs.The long duration of PFIAs can give incumbents some flexibility in terms of when they decide to claim credit for a project.For example, President Arias of Costa Rica decided to inaugurate the Santamaria airport in Alajuela just before finishing his mandate as President and at only 82% completion of the new terminal (Agencia EFE -Servicio Económico, 2003;La Nación, 2004, 2010).Press coverage bluntly states that "Arias didn't want to leave the presidency without a partial inauguration of the Juan Santamaría international airport" (La Nación, 2010).The journalists covering PFIAs are all too aware of the cycle of promises at the beginning of a mandate with regards to the delivery of essential infrastructure and how this frequently ends in disaster, with huge delays and cost overruns, but too late for the initiating politician to be held to account.Indeed, Arias name-checked three PFIAs in his final speech as president.He claimed to be leaving the country with less poverty and healthier public finances, adding that "concession tools [were] what allowed us to build the new San José-Caldera highway, expand the Juan Santamaría Airport and improve radically the efficiency of the port of Caldera" (La Nación, 2010).
The timing of PFIAs, thus, is an important feature of how they can be used as credit-claiming opportunities by incumbents.However, it is important to note that their long duration can also make it difficult to pinpoint exactly how some features of the credit-claiming process play out in reality.For example, one government may approve a particular PFIA, but not be in office when it comes to completion, meaning another party can claim credit for it.Similarly, trying to disentangle whether it is the promise of a PFIA or the completion of the project itself that wins over voters is also difficult to assess.In this paper, I seek to establish the various ways in which PFIAs can be used as distributive goods, but it is beyond the scope of my analyses to make these types of distinctions about the exact timing mechanisms at work in credit claiming for PFIAs.Here, I measure the distribution of PFIAs when the projects reach financial closure -that is, when the contract is signed between the government and the private parties -which is a critical moment as closure means that it is very difficult for the government to back out of the agreement at a later date. 9This means that future governments are locked into the project.In line with this logic, many of the credit-claiming behaviours that my qualitative findings speak to occur at the early stages of PFIAs, being centred around the presentation of projects, the fact of having attracted private finance, and the signing of project contracts.
PFIAs vary in the types and intensity of credit-claiming opportunities that they present.They offer multiple intertemporal opportunities for claiming credit, many types of symbolic ceremonial moments that can be created around the different stages of the contracting process, and the chance for politicians to present themselves as effective policy-makers through efficiency-centred narratives about the method for delivering infrastructure.

| Data and methods
My data covering infrastructure projects are drawn from the World Bank's Private Participation in Infrastructure Projects Database (World Bank, 2022).Whilst this database covers only low-and middle-income countries, 10 which might pose issues in terms of the generalisability of my findings, these countries are theoretically relevant.First, these countries are expected to 14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License increasingly rely upon PFIAs to deliver essential infrastructure over the coming decades (as laid out in Section 1).Second, the middle-income countries in my sample can be expected to mark a lower bound in terms of identifying patterns of distributive politics, as parties are likely to increasingly rely upon credit-claiming in order to influence elections as the country continues to democratise and politicians have greater incentives to target goods as electoral competition increases (Gainza et al., 2021;Shmuel, 2020). 11Further to this, as countries develop economically they accumulate the basic infrastructure for the delivery of essential goods and services and, as they do, incumbents are able to focus their attentions on distributing infrastructure according to electoral logics rather than logics of need (which are so crucial when in the earlier stages of development).I geo-code each of the PFIAs in the World Bank dataset and link them to the electoral district in which they are located (more details: Appendix, Section 2.2).Thus, I am able to match each project with its electoral district and the relevant district-level variables, such as which party won there and with what margin.My unit of analysis is the district-year.Infrastructure projects (made up of a "bundle" of agreements) are observed in the year of financial closure that is, the year in which the government and private partner entered into the PFIA.Summary statistics are provided in Tables 1 and 2.
Once geo-coded and mapped to electoral districts, I am left with 191 infrastructure projects 12 in 16 countries (Table A1, Appendix), spanning 30 years , covering 54 separate elections, and distributed across 968 separate districts.I am unable to undertake the electoral district mapping exercise for all of the 137 countries covered by the World Bank dataset as I am constrained by the availability of district level election data.Comparable district level election data are, unfortunately, limited which is what restricts the scope of my sample.Whilst this certainly limits the generalisability of my findings, the types of countries included in my sample are varied according to numerous factors from GDP (from relatively richer Mexico and Turkey, to low-GDP Solomon Islands and Mauritius 13 ), to geography (from Eastern Bloc countries such as Albania, Lithuania and Romania, to Latin American countries such as Bolivia, Costa Rica and, Mexico and Venezuela), to level of democratisation (from relatively high-ranking Mauritius and Costa Rica, to low-ranking Turkey and Venezuela (EIU, 2010)).Moreover, they were selected purely according to district level data (Brancati, 2018) availability criteria (rather than any criteria that might be associated with the likelihood of adopting PFIAs, for example).As such, I have no reason to believe that the sample should be skewed in a particular direction as it pertains to the distributive use of PFIAs.By geo-locating projects across the electoral maps of multiple countries and combining these data with detailed data on PFIAs, I am able to explore distributive politics on a large, cross-national scale, probing associations between electoral variables and PFIAs' geographic distribution.By analysing the distributive patterns associated with PFIAs across this sample of very different countries, I can be more confident that the relationships I uncover are not an artefact of a particular context (and its specific configuration of institutions, history, culture and economy), but are rather part of a broader phenomenon that affects many different types of countries.Moreover, by leveraging cross-country micro data, I am able to address broader concerns in the public policy and administration literatures about the lack of generalizability of many findings about public management that are drawn from single-country studies or cross-country averages (e.g., Meyer-Sahling et al. ( 2021)).

| Dependent variable
For the main analyses, my Dependent Variable (DV) is a binary variable, PFIA Present, indicating whether the district in question has had a new PFIA granted (1) or not (0) in a given election cycle.9% of district-years in my estimation sample are granted a new PFIA.

| Theoretical covariates
The Aligned District variable (mean = .54,S.D. = .50)is a binary variable that indicates whether the party that won in the district in question is the same party -or coalition of parties -that won at the national level (1) or another party (0).The volatility of an electoral district is measured by the absolute difference between the vote share of the winning party in the district and the vote share of the opposition (defined as the second largest party in that district contest).This gives me my continuous Margin of Victory variable (mean = .14,S.D. = .12),which provides a measure of the closeness of the race.Figure 1 presents a histogram and kernel density plot for the Margin of Victory measure.
These measures are drawn from the Global Elections Database (Brancati, 2018), which provides information on the results of both national and subnational elections around the world from 1980 to 2010 (more details: Appendix, Section 2.2).

| Statistical model
My preferred model is a linear probability model (LPM) with district fixed effects (to account for unobserved district level heterogeneity) and a linear time trend. 14The LPM makes interpretation of results simpler than alternative link function regressions (I use robust standard errors to address the LPM's violation of the heteroskedasticity assumption).In Table 3, I present my results investigating whether key electoral variables influence PFIAs' distribution across districts.The main regressors are whether or not a district was won by the same party that won at the national level or another party (Aligned District), the margin of victory with which a district was won (Margin of Victory), and their interaction.As such, the final specification is a district level fixed effects model with PFIA Present as its dependent variable that includes as independent variables Aligned District, Margin of Victory and their interaction, as well as a linear time variable to control for trends in PFIA usage.

| RESULTS
I find support for hypotheses H1, H2, and H3, providing associational evidence that PFIAs may be used as distributive goods and are not distributed in the same way when they are less attributable to the government.First, PFIAs are more likely to be found in districts where the national governing party won (Aligned District), as can be seen in Table 3, Column 1.This supports hypothesis H1 and provides a first piece of evidence consistent with PFIAs being distributed nonprogrammatically.According to my specification of interest (Table 3, Column 1), being in an Aligned District is associated with a 5% (0.048) increase in probability of being allocated a PFIA (as compared to a non-aligned district).Second, the Margin of Victory in isolation has a non-statistically significant effect on the likelihood of receiving a PFIA.Third, focusing on the interaction effect of interest -how the Margin of Victory in a district interacts with whether it is Aligned with the national government -we see that despite the positive coefficient associated with Aligned District, the interaction between the two is negative.The interaction effect is strongly statistically significant (a 28% decreased probability), telling us that the effect of being in an Aligned District is not the same for all levels of Margin of Victory.
Given that I am dealing here with an interaction effect, to interpret the coefficients in a way that makes substantive sense, I calculate marginal effects at meaningful values of my variables.For example, when one compares a non-aligned swing district to an aligned swing district ('swing' here defined as a district that was won by a margin of 1%), the predicted probability of seeing a PFIA in that district moves from 10% to 14%.Compare this with the change from a non-aligned to an aligned core district ('core' here defined as a winning margin of 23%, the 75th percentile) where the change is of a much smaller magnitude (from 8% to 7%).We see, then, that the effect is driven by the difference in likelihood of seeing a PFIA in Aligned Districts. 15This can be seen, graphically, in Figure 2.
In a single-member district setting, these results could be understood as incumbents seeking to concentrate PFIAs in swing districts where they need to pull out all the stops to win votes, whereas in core districts they use other -perhaps less costly -ways to reward voters.In a multi-member district setting, instead, these results could be understood as the national ruling party seeing the reduction in Margin of Victory in a given district over time as a signal that they are losing their lead and targeting PFIAs there to convince voters to return to their party. 16 As I have noted, my research design does not permit the identification of a causal relationship between the distribution of PFIAs and incumbents' electoral strategising.That is, we cannot be sure that the estimates reflect government reactions to past voting patterns.There are other factors that could be driving the results, such as variation in need for infrastructure or in the distribution of business interests across a given country.One may also worry that the presence of PFIAs themselves might affect voting patterns.This seems unlikely, however, given that I find a positive association in non-aligned districts which implies that opposition Margin of Victory increases when there have recently been PFIAs allocated to a given district.
That being said, there are a number of reasons for which the relationships I uncover could plausibly be causal.First, PFIAs are supposed to be instruments used to provide essential infrastructure and value for money for citizens and should not, in theory, be affected by political logics.Given the difficulty of accurately measuring need for infrastructure across countries, I assume that there should be no reason for more marginal districts to also be the most needy in terms of infrastructure (systematically and across countries).This is a reasonable assumption, I posit, because even if more marginal districts were more likely to receive traditional distributive goods -due to strategic targeting on the part of governments -this would, if anything, bias downwards my estimates in that any PFIA targeting effect would come in addition to existing traditional distributive goods.If poorer districts were systematically more likely to be won by a particular party (e.g., left-leaning parties), one might expect a correlation to exist between the marginality of the district and its neediness.However, given that I look across different countries and election cycles these effects are likely to cancel each other out. 17 Second, business interests and political interests frequently will not converge and as such one would expect to see less electoral targeting of PFIAs than projects delivered through full public provision.This is because incumbents have to convince private partners to invest in a given project and accept its related risk, unlike in full public provision where the government simply decides where and what a project should look like and will contract a firm to deliver such a good.The phenomenon whereby incumbents award construction projects to firms to 'build for building's sake' is also unlikely with PFIAs as construction and management are bundled such that the private partner needs the asset to be profitable even after it's been constructed.This feature of PFIAs, again, should bias my estimates downwards in that private interests should constrain the ability of incumbents to direct PFIAs to electorally strategic localities.

| Credit claiming and PFIAs: Empirical results
I now present my results exploring the credit claim potential of PFIAs (Columns 2-4, Table 3).The extent to which a citizen can attribute a PFIA to the government depends on a series of factors.PFIAs can be more or less 'visible' to voters, with a major physical project such as a new road or energy plant being more salient to a voter as opposed to a telecommunications project  that would likewise bring jobs to a district, but would represent less of a daily, physical reminder to the voter of the politician's efficacy in providing her district with public goods.Measuring this visibility directly, however, is impossible without being able to track the whereabouts of individual voters with respect to each piece of infrastructure.As such, I have to use features of the projects that can proxy for their probable salience to voters in a given district.Such a source of variation is the sector of a PFIA.
More visible infrastructure projects such as bridges, roads or water treatment plants will provide ample credit-claiming opportunity for parties and politicians.The physical presence of such projects and the fact that they are strongly geographically rooted in one place, provides the political agent with plentiful opportunities to remind constituents of her effectiveness in working for the district.This is what I understand as 'visibility' for the voter.As opposed to projects such as cellular networks or information computer technology (ICT) systems that -despite also creating jobs and serving the local population -are much less visible to the voter and are less electorally salient.Additionally, projects from sectors such as telecommunications where new technologies are more likely to be used can make contracting more difficult due to the complexity of the project.Indeed, Iossa et al. (2007) argue that "when the project is highly complex, it may be better to save on design costs and write a less complete contract" [p.62]where the roles of the government and private partner(s) are less clearly delineated.This can lead to reduced emphasis on the role of government within the project and, thus, a weakening of the credit-claiming opportunity.
Thus, I create two variables, Visible Project and Invisible Project, classifying energy (e.g., gas or electricity plants), transport (e.g., roads or airports) and water/sewage (e.g., water treatment or utility plant) projects as visible and information and ICT (e.g., telephone or Internet network) sector PPPs as invisible.An example of an energy sector (visible) project, is the Miravalles III Geothermal Power Plant in Guanacaste, Costa Rica (for more information see Vallejos Ruiz (2018)).Whereas an example of an ICT sector (invisible) project, is Bulgaria's Orbitel, a wireline telecommunications and Internet service provider headquartered in Sofia.
I find associational support that for visible sector projects, the relationship found in the main results is stable (Figure 3).Whereas invisible sector projects have a different relationship with marginality entirely (Figure 3).Coefficients are presented in Table 3 (Columns 2 and 3), where we see that, like for the main results, for visible sector projects, Aligned District, individually, has a positive coefficient, yet the interaction term between it and Margin of Victory is negative.Whereas for invisible sector projects, no statistically significant relationship can be identified.Column 4 presents results using another proxy for attributability (International PFIA, discussed in the Appendix, Section 3) where the main result persists.These results support hypothesis H3, where more attributable PFIAs (from visible sectors) follow distributive patterns, whereas less attributable projects follow different patterns.In the Appendix (Table A2), I present results showing that my interaction effect of interest is robust to being modelled in various ways.

| DISCUSSION
My press analysis explores the credit-claiming opportunities represented by PFIAs, looking in practice at how politicians leverage PFIAs to demonstrate their effectiveness in bringing valuable public goods to their districts.The analysis uncovers a number of ways in which politicians endeavour to tie themselves to projects at various stages of their development and to take advantage of different features of PFIAs (such as their long time horizons or their purported efficiency) to highlight their involvement in projects and their efficacy in delivering public goods.The analysis also provides insights into the language employed by politicians when claiming credit for PFIAs, which are portrayed in grand terms, being touted as game-changing, highly welfare-enhancing projects.
My quantitative findings are consistent with incumbents being able to use private money, through PFIAs, to pursue political ends.I find that PFIAs are consistently more likely to be found in those electoral districts where the governing party won and won with a small margin.I also find that PFIAs that are attributable to the government follow the same pattern, whereas less attributable projects seem to be distributed according to a different logic, lending support to the argument that incumbents are sensitive to the credit-claim opportunity presented by PFIAs.As such, PFIAs seem to be used as windfalls by incumbents; they use an instrument that should be about providing fundamental infrastructure to serve their political ends.In this way, my findings shed light on another facet of how distributive politics can feed into the incumbency advantage (Albertus, 2013;Stokes et al., 2013).
The strength of my data is that they allow me to explore the distributive use of PFIAs across a range of very different middle-income countries -from Albania to Venezuela -which indicates that this is a powerful relationship that merits more attention, given that it emerges despite many national differences.Through district-level fixed effects I am able to control for permanent features such as geography (which would dictate the feasibility of certain PFIAs), but I cannot explore the effects of features such as industrial composition or ethnic diversity that change over time and that may affect how districts are viewed by incumbents.Neither can I explore F I G U R E 3 Predicted probability of visible (left) and invisible PFIA Present (right) (LPM, FE) (legend as in Figure 2).
14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License heterogeneous effects across these margins.Nor can I speak to the various motivations and logics that individual politicians may use when deciding where to distribute PFIAs or the specific bureaucratic landscape they face when making these decisions.This presents both a limitation to my findings and opens a promising avenue for future research.Future empirical studies, both quantitative and qualitative, could use smaller samples or within-country research designs to home in on these features and further clarify how features of PFIAs interact with district characteristics to shape their political use.For example, by incorporating intertemporal dynamics into analyses to understand how receiving a PFIA in time t affects the likelihood of receiving one in time t + 1, or to understand how electoral system changes or the relationship of the executive with the bureaucracy implementing PFIAs can impact the political value of PFIAs, or, in a within-country setting, combining data on PFIAs and full public provision projects to directly compare their distributive uses.

| CONCLUSION
Research into the political determinants of PFIAs as a way to provide infrastructure is growing (e.g., Bertelli (2018)).However, little is currently known about the distributive use of these projects or how politicians operationalise them for political gain across different institutional contexts.My analysis sheds light on both the political use of PFIAs and features of these projects that can shape their political attractiveness as credit-claiming opportunities.I build on existing literature by presenting evidence that is consistent with incumbents being able to use private money with which to claim credit.This adds to our understanding of how distributive politics works in that we see private actors entering the field, yet relationships consistent with electoral targeting remain.
PFIAs come with a degree of slippage in terms of the degree to which a government can both freely decide the type of public good to be built and how it should be delivered and, subsequently, directly claim credit for it.This slippage should theoretically reduce the likelihood of observing distributive patterns in how PFIAs are positioned across the electoral map, given that (potentially competing) private and political interests need to converge for the project to be realised.I have employed a cumulative approach to present a body of evidence that is consistent with PFIAs being used as distributive goods, despite the involvement of non-governmental actors.
I argue that the credit claim potential of PFIAs -proxied in a variety of ways -changes how they are used by political actors: PFIAs that are directly attributable to the government follow more pronounced distributive patterns.My qualitative evidence explores some of the potential mechanisms that may be behind such a phenomenon, investigating how politicians claim credit for PFIAs in a selection of the countries in my sample.This press analysis reveals numerous ways in which politicians use PFIAs to their strategic advantage, claiming credit for projects over the course of their negotiation and implementation and well before the public goods are delivered.
In a context like today's where policy delivery is increasingly complex and takes many forms, a plethora of questions emerge about what this means for democracy.I have presented evidence that suggests that hybrid governance solutions such as PFIAs do not sidestep traditional distributive politics.It seems that incumbents may be able to use PFIAs in the same way that they use other particularistic goods, yet they come at a cost in terms of the extent to which voters can hold politicians to account.My findings highlight the need for a closer examination of how the ever more complex ways in which public goods are delivered affect distributive practices.
14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Summary statistics, main model.Summary statistics: Proxies for attributability.14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

Main Visible Sector Invisible Sector International PFIA
14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Histogram and kernel density plot of Margin of Victory.The table reports results from LPM time series estimations.DV is PFIA Present for Column 1. DVs for Columns 2-4 described in column headings.Robust standard errors in parentheses., 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License F I G U R E 1Note: *p < 0.1; **p < 0.05; ***p < 0.01.T A B L E 3 Aligned district and political competitiveness effects on PFIA Distribution (LPM).DVs: PFIA Present, Visible Sector, Invisible Sector, and International PFIA.14680491 14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 14680491, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gove.12824 by University College London UCL Library Services, Wiley Online Library on [19/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License