Abstract
Political dynasties exist in practically every type of democracy, but take different forms in different places. Yet the types of dynastic structures have remained unexplored. We argue that horizontal dynasties—multiple members from the same political family holding different political offices concurrently—affect policymaking by replacing potential political rivals, who may oppose an incumbent’s policy choices, with a member of the family. But in developing countries, the policy change that accrues from dynastic status may not lead to higher levels of economic development. We test this argument’s implications in the Philippines. Using a close elections regression discontinuity design on a sample of mayors, we show that (i) horizontally dynastic mayors have higher levels of government spending, (ii) direct institutional constraints are the mechanism that drives this core result, and (iii) horizontally dynastic mayors do not lead to higher economic growth economic growth or lower poverty.
Keywords
Introduction
Political dynasties exist all over the world, from developing countries in Asia and Africa to industrialized Western democracies like Sweden and the United States. As a result, a growing literature has examined the ways that dynasties perpetuate themselves through the incumbency advantage (Chhibber, 2013; Feinstein, 2011; Fiva & Smith, 2018; Querubin, 2016; Van Coppenolle, 2017) and the implications of dynastic politics on political, economic, and social outcomes (George, 2019; Asako et al., 2015; Braganca et al., 2015; Folke et al., 2017a). The majority of this literature examines family relationships across time. Yet intertemporal linkages between family members are clearly not the only type of family linkage. Horizontal dynasties—family members holding multiple political offices at the same time—and the outcomes that accrue to horizontal dynastic status are still relatively unexplored.
It is important to study horizontal dynasties because the concurrent structure of these dynasties allows us to examine novel strategies that dynasties use to achieve their political and policy goals. In addition to traditional mechanisms such as the incumbency advantage, horizontally dynastic politicians have access to other strategies that non-dynastic politicians do not. In particular, if politicians from the same family hold different offices at the same point in time they can coordinate, and circumvent potential roadblocks, to implement joint policy goals. Horizontal dynasties thus operate via distinct mechanisms and may lead to different outcomes. This article answers the following question: How do horizontal dynasties affect economic outcomes and what is the mechanism that allows them to do so?
We argue that the horizontal dynastic structure allows politicians to more easily enact policy by replacing potential political rivals in other offices with a member of the same family, thus rendering conventional constraints on policymaking—institutional impediments such as veto power or coalitional politics between two potential rival politicians—ineffective. Replacing a potential political rival with a family member greases the wheels of policymaking; the various constraints of political infighting are replaced by joint effort towards mutual goals. This result is being driven by positions with the most direct constraints on the dynastic politician. In other words, policy change happens when potential rivals who share a similar jurisdiction over the reference politician (and hence are directly engaged in the policymaking process), and have the most access to institutional/legal constraints, are replaced by a member of the politician’s family. The argument further implies that members of the same family have similar preferences, and hence share political goals. However, policy change does not necessarily imply improvements in economic development. In contexts of low electoral accountability and where clientelism and other non-programmatic policies determine electoral success, the policy changes that accrue to horizontal dynastic status do not lead to higher levels of economic development, implying that policy change may have simply facilitated rent-seeking.
The predictions of this argument guide our empirical analysis. We study the Philippines over the past 20 years, an ideal setting for several reasons. First, horizontally dynastic politicians are present at multiple levels of local government: 15% of all municipal mayors and 45% of all provincial governors are horizontally dynastic. Second, because several existing papers on dynastic incumbency advantage and the socioeconomic effects of dynastic status are set in the Philippines (Querubin, 2016; Labonne et al., 2017; Mendoza et al., 2019), our research directly extends and provides channels beyond the incumbency advantage by which dynasties affect Philippine politics.
This article focuses on Philippine municipal mayors. We identify a mayor as horizontally dynastic if he/she has a relative who is concurrently serving in a local executive or legislative position within the province the mayor is a part of: the governor or vice governor of the province, or congresspersons, mayors and vice mayors in the same province. We choose to focus on mayors because they are the primary decision-makers in setting and executing policy in the municipality. On the other hand, these other local positions influence the mayor’s capacity to set policy. Most changes along local policy dimensions can be attributed to the preferences and incentives of the local mayor. We also choose mayors for statistical reasons. There are over 1600 mayors serving concurrently in any given year. The large number of mayors gives us enough sample size for identification.
Testing our predictions could be undermined by the endogeneity of horizontal dynastic status. The presence of horizontal dynasties may be both a cause or a consequence of the policy decisions of the mayor. A third factor, such as local economic and political conditions, may also bring about dynasties and policy change. To overcome such difficulties, our identification strategy employs a regression discontinuity design that leverages close elections of mayors’ relatives to account for the presence of a horizontally dynastic mayor. Using the incumbent mayor as the reference politician, we consider a relative who runs for another political office (the vice mayor, mayor, vice governor, governor, or congressman within the same province). If his/her relative barely wins, the mayor is considered horizontally dynastic because two of them are in office concurrently. We exploit close elections as a quasi-experiment that randomly assigns whether the mayor belongs to a dynasty (when a relative barely wins an election) or non-dynasty (when a relative barely loses an election). Following the dynastic incumbency literature, we use close elections as our source of exogenous variation.
Our empirical results are consistent with the predictions of our argument. We use government spending as our measure of policy change, and find that municipalities with horizontally dynastic mayors spend 4–5% more than municipalities with non-dynastic mayors. This result is robust to different functional forms of the running variable, to various optimal bandwidth selection methods and to the inclusion the inclusion of a variety of control variables. Furthermore, we show that this increase in spending is being driven by mayors who are horizontally dynastic with their vice mayors, suggesting that positions with the most direct constraints over the mayor—in terms of the formal laws that enable vetoes as well as proximity to the policymaking process—are driving the spending result. We also show some suggestive evidence that this particular type of dynastic linkage leads to increases in specific components of total spending: economic services and health. Despite this, we see no relationship between horizontal dynastic status and night light luminosity (a proxy for growth) and poverty incidence. The results imply that in settings like the Philippines, where clientelism and rent-seeking are prevalent and where government institutions are prone to mismanagement, the policy change that accrues to horizontal dynastic status does not necessarily lead to development. Finally, we show that these results are not driven by alternative explanations, such as dynasties serving as a proxy for political parties, or the networks of the dynasty.
This article extends and deepens scholarship on political dynasties. The recent boom in research on political dynasties has thus far focused on dynastic selection, most particularly the dynastic incumbency advantage—holding political office has a causal effect on the probability of having future relatives in office (Querubin, 2016; Fiva & Smith, 2018; Van Coppenolle, 2017)—or on the social and economic outcomes that accrue from dynastic status (Folke et al., 2017b; Geys & Smith, 2017). This article complements the existing literature by focusing on horizontal dynasties and linking this particular dynastic structure with economic outcomes such as government spending and economic growth, and arguing for the removal of direct institutional constraints as the mechanism that mediates dynastic structure with policy.
Our results also contribute to a growing literature on how social structures affect electoral success in young democracies, and in particular the Philippines. Our work builds on Querubin (2016), who provides evidence for the dynastic incumbency advantage—individuals who barely win political office are much more likely to have future relatives in office. We also provide supporting evidence for empirical work that has linked dynasties to negative development outcomes in the Philippines (Tusalem & Pe-Aguirre, 2013; Mendoza et al., 2016). Finally, we build on a series of papers (Cruz et al., 2017, 2020; Cruz, 2019) that show that a variety of political networks (networks among voters and politicians’ extended families for example) facilitate strategies that lead to electoral success. Like these papers, we agree that political linkages matter. We provide our own unique contribution in two ways. First, we break down the definition of dynasties to show how a particular type of dynasty (in this case a horizontal dynasty) uses the specifics of this structure to craft electoral strategies. Second, while the extant literature focuses on clientelism as politicians’ core electoral strategy, we emphasize policymaking and public spending (and by proxy public goods provision) as an alternative or augmenting strategy by which politicians may achieve political goals.
Horizontal Dynasties: Policy, Direct Institutional Constraints, and Development
In this section, we propose a simple argument that links a politician’s horizontal dynastic status with policy and development outcomes. The argument is set-up as follows: Consider a reference politician, which we call the mayor. 1 Mayors set and execute policy in their municipality. Mayors attempt to win elections and extract rents. In principle, they may use a variety of tools to achieve these goals. Here, we focus on policymaking. For example, a policy like building a road may benefit a mayor’s constituents and win their votes, or otherwise be a conduit for the mayor’s own enrichment. Regardless of the mayor’s underlying goals, they face constraints in the form of other rival politicians. These rival politicians wield institutional tools, such as the ability to veto a mayor’s policy, that will prevent the mayor from enacting his/her preferred policies. Mayors therefore need to find a way to circumvent these constraints in order to achieve their political objectives.
Horizontal dynastic linkages allow mayors to circumvent these constraints from rival politicians by replacing a would-be rival with a member of their family. Consider the following simple scenario: A mayor wants to build a road but is vetoed by a rival politician. The road does not get built. The alternative scenario is one where the rival is replaced by a member of the mayor’s family. Consider again the same road-building project. The mayor proposes the project. But now, with the family member in place of the rival, the road is now built. Extending this logic to policy more broadly, we can say that horizontally dynastic mayors will implement differential policy versus non-dynastic mayors.
There are a variety of mechanisms that may mediate the relationship between horizontally dynastic mayors and spending. We argue that direct institutional constraints and their removal by replacing a potential rival politician with a member of the family are a relevant mechanism mediating this result. By direct constraints, we mean the legal and institutional tools a rival politician has to constrain the mayor through, for example, the vetoing of ordinances and the jurisdiction-based constitutional mandate to serve as a check on the mayor’s local autonomy. Some horizontal linkages have more of these institutional constraints than others, and if direct institutional constraints are indeed driving spending, then the spending result should be disproportionately driven by these positions. Thus, the increased spending for horizontally dynastic mayors will be driven by positions with the most direct institutional constraints.
While horizontally dynastic mayors differentially affect policy, it is unclear whether these policy differences imply improvements in welfare and development. On the one hand, studies have argued that policy may be used as a way to generate kickbacks for corrupt politicians (Boas et al., 2014; Rose-Ackerman & Palifka, 2016). If this is the case, then horizontally dynastic politicians further facilitate rent-seeking because they eliminate potential checks on corrupt politicians. On the other hand, policy may be welfare enhancing and politicians hoping to enact such policies do so in the hopes of maintaining political office (Ashworth, 2012; De Mesquita et al., 2005; Lake & Baum, 2001). In this sense, horizontal dynasties facilitate development because they are able to circumvent the inefficiencies inherent to checks and balances in order to provide citizens with policies they desire.
In the context of the Philippines, we expect the former case to accrue. Horizontally dynastic mayors do not spur development. First, the Philippines lacks electoral accountability (Hutchcroft, 2000; Quimpo, 2007). In the absence of accountability channels, winning elections does not hinge on public service provision and instead becomes a matter of facilitating clientelistic political exchange (Cruz, 2019; Cruz et al., 2017). Furthermore, the Philippines has a weak bureaucracy and low state capacity. Even if spending were ostensibly to be used for welfare-enhancing purposes, the quality of the output may be low and spending may be inefficient. Our argument is consistent with an empirical literature that claims that in places with poor governance, government spending is not correlated with improved development outcomes (Devarajan et al., 1996; Rajkumar & Swaroop, 2008).
The argument yields three testable hypotheses:
At the outset, we acknowledge limiting conditions for our argument. The argument likely does not apply to horizontally dynastic politicians who are powerful enough to completely override the political system. These horizontally dynastic politicians do not need to work within (and hence attempt to circumvent) checks and balances. Perhaps a more appropriate model for these “strong” dynasties is warlord governance (Mukhopadhyay, 2014) or the Mafia (Gambetta, 1996). Furthermore, the argument likely does not hold in countries with stable party systems. In these countries, parties already facilitate the role of coordination and preference alignment between members of the party organization (Aldrich, 1995). This implies that horizontal dynasties may be an institutional solution to the problem of weak, unstable party alignments that plague countries like the Philippines (Hicken, 2009; Montinola, 1999).
The Philippines: Background and Institutional Context
This article focuses on mayors in the context of the Philippines, spanning the early 1990s to the present. We focus on mayors for specific reasons: first, mayors are considered the central actor in local politics, close enough to the people to be accountable to their needs while having substantive relationships with other local officials within the municipality, across municipalities, and with other branches of local government. Second, mayors also face institutional and extralegal constraints (or opportunities for coordination) from other local politicians.
Local Political Structure
Mayors are the head of the executive branch of a Philippine city or municipality, the backbone of local politics in the Philippines. 2 Mayors preside over a variety of executive functions at the local level, and the extent of their control ranges from the exercise of general supervision over all projects and programs at the city level, initiating and maximizing the generation of revenue, and ensuring the delivery of basic services (Local Government Code (henceforth LGC) Section 455). As a branch of the executive, mayors are tasked with enforcing local laws or ordinances that are created through a city legislature, composed of the municipal board.
Research has identified mayors, their incentives and their behaviors as among the major drivers of local Philippine politics. In his study of local politics in Cavite, Sidel (1999) argues that the position of mayor is a major prize, the benefits associated with it being “the awarding of building permits, the passage of municipal zoning ordinances, the use of government-owned land, the allocation of public works, the approval of reclamation projects… [and] the awarding of petty monopoly franchises and concessions” (p.33).
Mayors operate within a larger infrastructure of local politics (see Figure 1). Among the local political linkages, the relationship between the mayor and the vice mayor is most direct. By law (delineated in the LGC), vice mayors assist the mayors in implementing policy at the municipality level. Assistance also means serving as a check on the mayor, thus granting the vice mayor with direct institutional constraints on the mayor. Most importantly, vice mayors serve as the presiding officer of the municipal board, which has veto power over the mayor’s ordinances—the primary legal means by which mayors set laws and enact policies. Major policies are enacted through ordinances. Particularly relevant to this article is the fact that the budget, and thus the line items for spending in the fiscal year, can only be enacted through an ordinance that is approved by the vice mayor-led municipal board, via LGC Section 319 and the LGC Implementing Rules and Regulations Sections 414–415.3 The specifics of this budget are written up in the Revenue Code, as outlined by LGC Section 13. Political structure in the Philippines.
Moreover, the vice mayor’s mandate, and the fact that both he/she and the mayor have jurisdiction over the municipality, implies that they are required to engage themselves with all of the mayor’s policies. Thus, vice mayors may exert formal opposition to the mayor very directly and oppose the mayor’s most relevant policies. This jurisdictional autonomy granting checks and balances responsibility to the vice mayor vis-à-vis the mayor is enshrined in Article 10 of the 1987 Philippine Constitution, and further delineated in LGC Section 16, a catch-all provision for local autonomy.
Cities and municipalities make up one of 81 provinces, the most aggregated level of local government in the Philippines. The executive branch of a province is headed by a governor and vice governor, who attain office through plurality elections. Although the political responsibilities of governors and vice governors are similar in each jurisdiction, the provinces themselves vary widely in terms of measures such as levels of economic development. Governors also have oversight responsibilities over mayors. LGC Section 39 mandates governors to ensure that mayors’ projects fall within prescribed duties and responsibilities. Since they have oversight responsibilities over the mayor, they may constrain them. However, they do not have formal vetoes over mayors’ ordinances and do not formally preside over the municipal board like the vice mayor. They do have veto power over executive orders—minor local policies that do not include the majority of budget items. This implies that they may serve as an institutional barrier, but since they have no veto power over the more important ordinances and the autonomy of municipalities is legally protected by the Constitution, they have less direct constraints on mayors than vice mayors do. Note that governors may of course impose indirect constraints on mayors through means like delaying approval of projects, but these are less effective than straight-up legal constraints.
Multiple cities and municipalities also comprise a legislative jurisdiction called a district. Districts are headed by congresspersons, who are in charge of representing their district in the House of Representatives, which is in turn responsible for drafting national legislation. 4 Congresspersons affect mayors’ spending through less direct means. The congressperson or a congressional representative sits on the local development council, and this may influence mayors’ policies with regard to public spending. While mayors do have substantial discretion over final policy, congresspersons may exert some influence in these councils via indirect coordination, that is, by tying discretionary pork barrel spending to particular mayoral policies.
The remaining linkages—mayors in other municipalities and vice governors—influence the mayor through indirect means. Mayors may collaborate with other mayors in large scalable projects, and vice governors may influence municipal outcomes through their influence on the governor. Neither has direct (or formal) constraints on a given municipal mayor.
For this article, mayors will be considered horizontally dynastic if they are related to any of the local positions expounded on above—vice mayor, mayor, congressperson, vice governor or, governor.
These positions are all simultaneously elected at-large every 3 years. Each position faces a term limit of three consecutive terms. In general, elections for all local positions have an average number of 2.9–3.5 candidates running for any position and a mean vote share concentration of 0.50–0.54. 5 At first glance, mayoral and vice mayoral elections seem to be more competitive, having fewer candidates compared to gubernatorial and congress elections (2.9–3.0 vs. 3.5 candidates). However, looking at the HHI shows that mayoral, gubernatorial, and congress elections feature vote shares that are equally concentrated (see Supplementary Appendix B for a complete breakdown by position). Positions like the vice mayor and mayor also serve as stepping stones to higher office like that of the mayor and governor, respectively. 6 For example consider Dennis Pineda, from the Pineda dynasty of Pampanga province and its incumbent governor. He served as mayor of Lubao, Pampanga for three consecutive terms from 2001–2010. He was vice governor from 2013 to 2019 before his current position. Also note that he replaced his mother, Lilia Pineda, as governor. We also see this pattern in the data: 22% of governors have held the mayoral office in the past.
Horizontal dynasties are also more prevalent at higher levels of governance. For example, 45% of governors are horizontally dynastic, while only 15% of mayors are horizontally dynastic. Mayor–governor links are thus potentially indicative of more advanced dynasties, as not all dynasties successfully reach higher levels of government. This has implications for the later analysis breaking down the particular dynastic linkage with the mayor, since dynastic pairs with governors potentially differ in resources or ability (which affect spending) from dynastic pairs with vice mayors. We will discuss these concerns when we test H2. Finally, voters in the Philippines vote for individuals and not parties. It is therefore possible that a voter votes for a mayor and another local politician (such as the vice mayor or the governor) from an opposing party.
Political Dynasties in the Philippines
Political dynasties—a collection of family members occupying political office either concurrently or through time—are the core unit of local political organization, existing alongside and embedded within formal institutions such as political parties (Fegan, 1993). 7 Prominent political analysts have noted that dynasties are an organizational expression of the century-long dominance of local governments by local elites. Political clans “are an enduring feature of Philippine politics,” and have “displayed an uncanny ability to adapt to the changing landscape of Philippine politics” (Coronel, 2004). That is, while Philippine politics features the formal rules and institutions that usually accompany democratic governments, these rules and institutions are underpinned by informal structures, in this case the political family (McCoy, 2009).
Data and Descriptive Statistics
Sample Construction
The article focuses on mayors’ relationships with vice mayors, mayors, congresspersons, vice governors, and governors within the same province. A horizontally dynastic mayor is identified when that mayor and a politician from any one of these positions is from the same family concurrently. We obtain elections data for these positions from the Commission on Elections from 1992 until 2013. During this period, there were 74–80 provinces, 200 to 234 districts, and 1610 to 1634 cities and municipalities in the Philippines. All local elections are held simultaneously every 3 years, and all local positions have a three-term limit. The initial sample includes 625 governors and vice governors, 1691 congresspersons, and 12,891 mayors and vice mayors, totaling 28,723 local officials across eight election years.
From the initial sample of 12,891 mayors, we focus on those mayors with exactly one relative running for office at the same election year, leaving us with 1962 observations in the final sample. 8
Summary Statistics, Initial versus Final Sample.
Note: The initial sample consists of all mayors while the final sample consists of all mayors with exactly one relative in office at the same time.
∗∗∗, ∗∗ and ∗ denote statistical significance at 1%, 5%, and 10% level, respectively.
Independent Variable: Horizontal Dynasty
The key independent variable for this study is whether the mayor is horizontally dynastic or not. Using the final sample, we define a mayor in city i at election year t to be horizontally dynastic if he/she has exactly one family member occupying a local political office j within the larger province where the municipality is located in, also at election year t. Hence, the requirements for satisfying the horizontal dynasty definition are: (1) the mayor has exactly one relative in another office, (2) both the mayor and his/her relative win the election, (3) both are in office at the same election year, (4) both come from the same province, and (5) the relative occupies a local position as defined above. For example, a winning mayor in the province of Cebu is horizontally dynastic in 2013 if he/she has a relative who is also elected as the governor, vice governor, congressperson, mayor, or vice mayor in the same province in 2013. It is important to note that this definition of dynasties is more specific than the definition taken by other authors (Querubin, 2016). 9 Because our argument focuses on the concurrency of dynastic status, limiting dynastic status to multiple family members within a given year constitutes the most valid variable to test our argument.
We identify family members by utilizing naming traditions and matching surnames (Braganca et al., 2015; Querubin, 2016). We follow the existing literature by identifying family networks via a matching procedure where individuals with identical surnames residing in the same province are considered related. Supplementary Appendix C expounds on the matching process in greater detail and addresses potential concerns such as false matches (positive and negative) in the data.
Dependent Variables
Hypothesis 1 states that horizontal dynastic status (politicians from the same family holding political positions simultaneously) leads to differential policy than non-dynastic status. Our measure of policy outcome is government spending. Furthermore, we claim that in the Philippine context, horizontally dynastic mayors will spend more than non-dynastic mayors. The data on government spending is at the municipal level obtained from the Bureau of Local Government Finance. We consider the 3 year average of a municipality’s spending after every election. For example, for the election year 2007, we consider the dependent variable to be the municipal government spending for 2008–2010 since mayors only have control over the budget after they assume office and budget planning starts the year before its enactment. Finally, we use log per capita measure to ensure comparability across differently sized municipalities, to decrease the impact of outliers, and to make the coefficients easier to interpret. Consistent with our argument, this measure includes all the programs that benefit the mayor’s constituency—including public and social services such as schools, hospitals, and roads, public goods such as security, and exclusive goods such as titles, permits, and other forms of property rights.
We choose government spending as our measure of policy outcomes for the following reasons: First, the level of government spending represents an observable measure that politicians attempt to manipulate, and hence is often used as a policy outcome in the literature (Alt & Lowry, 1994; Alesina & Rosenthal, 1995; Fiorina, 1996). Second, the level of government spending is consistent with both winning votes and extracting rents (Birdsall, 1996; Dixit & Londregan, 1996). For example, if a politician wants to win elected office, he or she may increase the level of spending on schools and hospitals to increase welfare and hence win votes. On the other hand, politicians who hope to engage in rent-seeking may also increase local government spending, creating a bigger pie for them to pilfer from. An increase in government spending thus captures a variety of political goals—from winning votes to stealing resources. Finally, Philippine spending data is reliable, standardized across years, and available for the entirety of our sample.
While we argue that in the Philippine context spending is higher for horizontally dynastic mayors, in principle, horizontally dynastic mayors may instead reduce local government spending. An alternative argument is that horizontal dynasties solve the common pool resource problem (Weingast et al., 1981). By this logic, two politicians both spend beyond what is socially optimal because each seeks to provide for their constituency. Horizontal dynasties align preferences and therefore allow both politicians (particularly the mayor) to reduce spending. The specifics of the Philippine context do not fit this argument. Mayors have control over budget allocation and executing policy, and therefore control over the municipality’s resource pool (Esguerra, 2001). The implication of this arrangement is that political rivals do not have access to the mayor’s resources. They thus constrain the mayor by blocking policy, for example via a vice mayor vetoing the mayor’s road construction project.
Finally, the argument presumes that credit for higher government spending goes to the mayor (and not to other politicians like the vice mayor or the governor). In the Philippines, mayors use their significant control over both the budget and the execution of government projects to claim credit (Cruz & Schneider, 2017). Figure 2 shows mayors putting their names on completed projects in order to claim credit for them. Mayors and government projects.
Hypothesis 2 states that direct institutional constraints are the mechanisms that mediate the positive link between horizontally dynastic mayors and spending. This analysis examines the effect of each horizontal link to the mayor, but the dependent variable is still municipal spending.
This Pasig City public high school is named after its longest-serving mayor. The Eusebio dynasty has ruled Pasig City since 1992, starting from the patriarch (Vicente Eusebio), his wife (Soledad Cruz-Eusebio), the son and incumbent (Robert Eusebio), and the incumbent’s wife (Maribel Andaya-Eusebio). Currently, brothers Robert and Richard Eusebio are the incumbent mayor and congressman, respectively.
Hypothesis 3 states that horizontally dynastic mayors do not spur economic development. We operationalize the development implications of horizontal dynasties by examining their effect on economic growth and poverty reduction. We use the log of night light luminosity as a proxy for economic growth. Moreover, night time lights have been shown to be very highly correlated with traditional measures such as GDP, and especially advantageous for measuring output at the lowest levels of development (Henderson et al., 2012; Pierskalla et al., 2017). This article employs night lights data from the Defense Meteorological Satellite Program’s Operational Linescan System. Luminosity intensity scores are scaled from 0–63 for each square kilometer. The unweighted average of each 1 km × 1 km grid cell within the boundaries of the municipality is calculated to obtain a luminosity score for each municipality. We take the night lights data 2 years after every election as the outcome variable. We also use municipal poverty rates as a second measure of economic development. We obtained the small-area poverty estimates from the Philippine Statistics Authority, which is based on the triennial Family Income and Expenditures Survey from 2000 to 2012. For consistency with the night lights measure, we assign the poverty rate 2 years after every election as the corresponding outcome variable.
It is worth noting that both night light luminosity and poverty rates are not perfect proxies for development. First, there may be a lag between policy, in terms of spending, and development, in terms of luminosity and poverty reduction. The developmental benefits that accrue from spending may bear fruit beyond the first few years of a mayor’s term. We choose not to test longer term economic development because this leads to an identification problem—it is not clear to which mayor do we credit long-term development. Second, economic development is a multifaceted concept, and our measures may not fully capture the potential ways spending may affect it. Still, we believe that the evidence provided above is suggestive of our argument that horizontally dynastic mayors do not lead to short-run development.
Identification Strategy and Empirical Results
Testing H1: Horizontal Dynasties and Spending
We first test whether horizontally dynastic mayors spend more than non-dynastic mayors. However, this presents an identification challenge since a naive comparison of horizontally dynastic and non-dynastic mayors may confound the effect of dynastic status with other variables that lead to higher spending. In practice, dynastic and non-dynastic areas are different along several dimensions: population, poverty incidence, political competition, urbanity, and many others. For example, wealthier and more developed areas collect higher tax revenues, which can lead to higher spending and attract the formation of dynasties due to the potential for rent-seeking. Another confounder is politician ability, where more talented politicians are better able to navigate existing rules to increase spending. In addition, highly skilled politicians may have greater chances of starting dynasties. Hence, the statistical relationship we see may therefore be an artifact of political ability or local development, and not of intrafamily cooperation.
In order to estimate the causal effect of dynasties on spending, the ideal experiment would be to randomly assign horizontal dynastic status to two otherwise similar mayors (one is horizontally dynastic, while the other is non-dynastic) and observe consequent differences in their respective municipalities’ spending. In this imagined setting, the treatment is assigning a mayor to belong to a dynasty: that is, the mayor and at least one relative are simultaneously elected and occupy political office, whether as a vice mayor, mayor, congressman, vice governor, or governor in the same province.
While this is impossible to experimentally manipulate, using regression discontinuity designs closely mimics this experimental scenario (Lee, 2008). By definition, a mayor achieves horizontally dynastic status when he/she has a relative occupying another political office within the province during the same election term. Whether a mayor’s relative barely wins or loses is therefore synonymous to quasi-random assignment of horizontal dynastic status to that particular mayor. Consistent with previous literature, close elections serve as a “near-experimental” setting where a candidate (i.e., the mayor’s relative) is deemed to win or lose at random. Hence, comparing mayors whose relative barely wins (and therefore achieve horizontally dynastic status) with those whose relative barely loses (and therefore do not achieve horizontally dynastic status) allows us to causally identify the effect of horizontal dynasties on local government spending. 10
Thus, the treatment group consists of mayors who have relatives simultaneously elected for any of the following positions: the provincial governor or vice governor, a congressperson, mayor or vice mayor, and the control group then includes mayors who have relatives who failed to get elected (in the same year) for any of the above local positions. 11
Note that we choose to define the mayor’s dynastic status broadly, identifying a mayor as horizontally dynastic if he/she has concurrent family members in office across a variety of positions. We do this because, conceptually, coordination with mayors can exist across several positions, and we show the net effect to ensure that we are not cherry-picking horizontal linkages. This does leave room to further delineate a mechanism. Thus, in the next section, we examine the effect of specific dynastic linkages on municipal spending.
Our identification strategy differs somewhat from the standard RD. While typical RD designs consider the same politician’s close election, forcing variable, treatment, and his/her eventual outcome, ours looks at a relative’s close election and forcing variable to determine the mayor’s treatment and eventual outcome. The difference stems from the definition of our independent variable and our sampling choice.
First, the label of horizontal dynasty implies that a mayor is dynastic if at least one of his/her family members also holds local political office. Hence, a mayor’s treatment is defined by the electoral outcome of his/her relative. While this is not standard, the underlying ideas of RD designs remain consistent. Echoing the close elections RD literature, the close election of a politician (e.g., the relative) can be considered an as-if random assignment. In our case, the relative’s electoral outcome assigns mayors into treatment (those with winning relatives) and control (those with losing relatives) groups. Using this design, there is a one-to-one mapping between a relative’s electoral outcomes and the mayor’s dynastic status. Hence, the quasi-random election of the relative dictates the quasi-random dynastic or non-dynastic status of the mayor.
Second, our choice of sample defines the treatment and control categories, and is crucial in ensuring that we consider the correct counterfactuals. 12 We do not take all close elections available—only the close elections of winning mayors’ relatives. The former would measure the impact of winning which is applicable only to the candidate, while the latter would measure the impact of a relative winning and the mayor becoming horizontally dynastic which has an effect on the mayor and the dynasty. The latter is the parameter we seek to causally estimate. Although subtle, this sampling choice is central to understanding that the treatment group becomes not just the winning politician, but the horizontal dynasty.
Marshall (2019) provides an extensive discussion on these non-standard RDDs and notes that these designs may identify not only the specific characteristic of interest (in our case, dynastic status), but also candidate- or context-level characteristics that ensure winning candidates who possess the specified characteristic remain in close races. To address this issue, we perform a battery of robustness checks by including a variety of politician-level and district-level variables to control for the potential effect of these compensating differentials.
Following the standard RDD specification, we exploit close elections to test our first hypothesis
Figure 3 presents suggestive evidence of the effects of horizontal dynastic status on spending. We see a discontinuous jump in spending around the cutoff, which implies that winning and hence gaining horizontal dynastic status leads to greater spending.
Horizontal Dynasties and Spending.
Note: Robust standard errors clustered at the province level are in parentheses. Dependent variable is log spending per capita in a municipality or city. Horizontal dynasty is a dummy that takes the value 1 (0) if the mayor’s relative wins (loses) the election. ∗∗∗,∗∗, and ∗ denote statistical significance at 1%, 5% and 10% level, respectively.
In all specifications, we see that having another relative in government leads mayors to spend more on their respective municipalities. In the more conservative specifications with fixed effects, municipalities led by dynastic mayors spend 4–5% more than their non-dynastic counterparts. These estimates are statistically significant and economically meaningful. They are equivalent to the municipality’s average spending on education and housing. The positive and statistically significant relationship is robust to the inclusion of fixed effects and to various functional forms of the winning margin. However, coefficients from specifications with fixed effects differ from those without fixed effects. In general, what is driving the attenuation is the inclusion of the year fixed effects. This is due to the source of variation of our dependent variable, log spending per capita. While cross-municipality differences are minimized due to the per capita adjustment, spending largely varies across time, experiencing a five-fold increase from 1992 to 2013. Municipality fixed effects marginally decrease the coefficients in full samples, but lower them more in optimal bandwidth approaches where sample sizes are halved. By reducing the average number of observations per group or municipality to less than two, the smaller samples make it more challenging to uncover effects due to lower or no variation within municipalities. Combining both fixed effects, we obtain lower, but still statistically significant, estimates.
We then address the identifying assumptions and the robustness of results in the RD design. First, we control for the urban imbalance, past wealth and politician-level variables such as tenure, gender, and past dynastic status. Using a specification chart in Supplementary Figure D4, we show that controlling for these variables does not impact our coefficients, which remain generally statistically significant.
14
Second, we test for balance on a number of predetermined variables that may cause discontinuities in spending. We find that covariate means are balanced around the cutoff (see Supplementary Figures D2 and D1). To ensure the validity of the RD design, the running variable has to affect the potential outcomes only through the treatment assignment indicator. We explore the potential role of gender and party affiliation, both potentially affected by the running variable and can influence our spending measure.
15
However, we show that the intermediate variables do not seem to be discontinuous around the threshold (see Figure 4). Next, one concern would be that mayors can differentially exert influence when family members are running in a close race, making their relatives more likely to win close elections, and thus undermining a core RD assumption. If this were the case, then we would see a discontinuity in the density around the cutoff. To address this concern, we check for potential manipulation using the McCrary (2008) test. Figure 5 shows that the vote share distribution is continuous around the cutoff, implying that we find no evidence of manipulation. Finally, for robustness, we show in Figure 6 that coefficient estimates are stable for a wide range of bandwidths. Checks for discontinuity (H1). Note: The figure shows whether intermediate variables like party affiliation and gender are also discontinuous around the vote margin threshold. This provides evidence that the forcing variable does not seem to affect other important variables, apart from dynastic status. McCrary density test 


While the RD provides us with a causal estimate of the mayor’s horizontal dynastic status on spending, its generalizability can be limited in two ways. First, our design focuses on dynastic pairs: precisely two politicians from the same family holding office contemporaneously. Yet the theory can be straightforwardly extended to families with more than two members holding office at the same time. Second, RDs provide causal results for close elections, yet these results do not necessarily speak to municipal elections that do not fall within the bounds of RD analysis. This limits generalizability because campaign strategies and the underlying characteristics of both the mayors and municipalities may differ between close elections and non-close elections. The specifics of the RD sample have implications for the theory. If the sample is limited to places where institutional rules constrain policymaking, then in places where rules are not binding (warlord municipalities for example), the benefits to having a family member in office concurrently may not be as beneficial. Regardless of the underlying implications of the sample, we address these limitations and attempt to generalize our theory with the OLS analysis. In Supplementary Appendix E, we find that OLS results—using the full sample of mayors and various definitions of horizontal dynasties accounting for more than two concurrent family members in office—mirror those of the RD. These results support the generalizability of the theory to elections that are not necessarily close and to a wider definition of horizontal dynasties.
Testing H2: Direct Institutional Constraints, Preference Alignment, and Spending
The previous section has provided causal evidence for the relationship between a mayor’s horizontal dynastic status and municipal spending. Further tests also suggest that the theory is generalizable to horizontally dynastic mayors beyond those in the RD sample and valid for more than two members of the family holding office concurrently.
In this section, we further refine our mechanism. We argue that direct institutional constraints, and thus the removal of these constraints by replacing a potential political rival with a family member, are the primary mechanism driving our results. We define these direct institutional constraints as the constitutional or legal constraints that allow a rival politician to use his/her mandated powers to prevent the mayor from enacting policies and increasing spending. Thus, the particular linkages that should be driving the spending result should be from positions that have the most direct institutional constraints over the mayor.
Yet, other mechanisms may exist. Instead of direct institutional constraints, potential political rivals may enforce a variety of “informal” tactics to curb the mayor’s spending. Higher level officials may resort to intimidation and violence to curb spending. More “benign” stalling tactics, such as simply refusing to meet with the mayor or stalling to sign approval documents also serve to constrain a mayor’s ability to implement policy.
In the Philippine context, vice mayors impose the most direct institutional constraints on the mayor’s ability to spend. Therefore, municipalities where mayors are horizontally dynastic with their vice mayors are driving the overall increase in spending. Recall that the vice mayor’s role as a check on the mayor is constitutionally mandated (the local autonomy provision) in Article 10 of the 1987 Constitution and reinforced in Section 16 of the Local Government Code. More specifically, vice mayors lead the municipal board, which has veto power over the mayor’s budget and hence all the spending line items (LGC Sections 319 and 414–415).
As noted previously, other positions, such as the governor and congressperson, have some direct institutional capacity to constrain the mayor. But their institutional constraints (especially for congresspersons) are far more limited. Moreover, they do not have the jurisdictional mandate that vice mayors do that is codified in the Constitution. Furthermore, it is worth noting that other positions such as vice governor or mayors from other municipalities have no (or at best extremely indirect) constitutional constraints over the mayor. On the other hand, these positions also have the greatest capacity for “informally” constraining mayors.
Thus, evidence in support of direct institutional constraints would be a significant and positive relationship between spending and mayors horizontally dynastic with their vice mayors. Evidence in support of more informal constraints would be a positive and significant relationship between the mayor and the other horizontal linkages.
To test this, we expand the original RD equation and include an indicator for each position. We also interact each position indicator with the forcing variable
Our results are consistent with direct institutional constraints; spending is being driven by mayors that are horizontally dynastic with vice mayors. For this section, we focus on the full sample given that the breakdown by position requires significantly more variation to be able to detect significant effects. In Figure 7, each bar represents the coefficients (in black diamonds) and confidence intervals (in shades of gray) of a particular dynastic linkage for a given specification combining the inclusion of fixed effects (none, year only, municipality only or both) and the function of the forcing variable (none, linear or quadratic). The first set of specifications (in dark gray) represents the estimates for links with the vice mayor, and the following sets (in alternating light and dark gray) for mayors, vice governors, governors, and congresspersons, respectively. The black (gray) circles at the bottom of the figure mean inclusion in (exclusion from) the regression. For example, the first bar is the vice mayor’s coefficient estimate for the specification with no fixed effects and no control function. Specification chart by position. Note: The figure splits the original horizontal dynasty variable into separate variables for dynastic linkages with vice mayors, mayors, vice governors, governors, and congresspersons (see position panel). Each bar represents a specification of a dynastic linkage, with or without fixed effects, using no, linear or quadratic functions of the running variable. The black diamond represents the point estimate, while the light (dark) gray bars are the 95% (90%) confidence intervals. The black (gray) circles mean inclusion in (exclusion from) the regression.
We find that the coefficients for vice mayor are not only generally statistically significant, but also are larger in magnitude compared to other positions. This provides suggestive evidence of the larger role that vice mayors play in influencing policy decisions made by the mayor. Alignment with vice mayors allows mayors to skirt around these direct institutional constraints, and provides them greater control over policy, which in this case, implies higher spending for their constituents.
A potential confounding explanation is that the spending differences across types of dynastic linkages are not driven by institutional constraints but reflect differences in the inherent quality of these dynastic pairs. For example, mayor–governor links are potentially indicative of more advanced dynasties, as not all dynasties successfully reach higher levels of government. This may suggest that certain dynastic linkages (mayor–governor or mayor–congressperson) represent higher quality politicians, who are better able to circumvent formal rules to implement their preferred policies. If this were the case, we would expect spending to be larger for these more advanced dynasties, but our results do not support such a model.
A potential caveat to these results is that we cannot test a plausible alternative mechanism: increased spending is being driven by transfers from higher levels of government. This would be particularly relevant to the position of congressperson, which for the entire duration of our sample could provide mayors with discretionary pork barrel spending. We cannot test this because pork barrel transfers are not available in our dataset. Nonetheless, other work has shown that political alliances between congresspersons and mayors lead to increased discretionary transfers (Atkinson et al., 2015; Hutchcroft, 2014; Ravanilla, 2017).
Next, we test an implicit assumption of our theory: that politicians from the same family have aligned preferences and work together to achieve shared policy goals. This preference alignment mirrors arguments of the role of parties and coalitions in aligning preferences between politicians and thus making policies easier to implement (Tsebelis, 1995; Treisman, 2000). In particular, we follow the logic of Cox and McCubbins (2001), who argue that “sameness of purpose”—in other words preference alignment—may allow politicians to circumvent institutional constraints without removing the constraint itself. Note that while preference alignment is difficult to test, the argument implies that aligned preferences will lead to reduced conflict across political offices. Politicians work together via access to multiple offices rather than fight over a particular position.
To test this, we examine whether mayor–vice mayor pairs from the same family are less likely to come in political conflict, that is, whether vice mayors are more or less likely to run against the incumbent mayor if they are related. While this methodology is not without caveats, we believe that testing conflict over political office is a sufficiently good measure of the alignment of preferences between politicians. This result bears out. We find that mayor–vice mayor pairs from the same family are less likely to compete with each other, compared to those from different families. To save space, we detail the specifics of our identification strategy, and address further concerns with the data, in Supplementary Appendix F.
Finally, we break down spending into its component parts to see which elements of spending are driving the results. Figure 8 displays these results. Similar to the previous specification chart, each bar represents the coefficient and confidence interval of a spending component for a given specification combining the inclusion of fixed effects (none, year only, municipality only, or both) and the function of the forcing variable (none, linear, or quadratic). Given that vice mayors have been shown to be driving the spending result, we only present their coefficients on the figure. The first set of specifications (in light gray) represents the coefficients for public services, and the following sets (in alternating dark and light gray) are for social services, economic services, education, health, housing, and debt services, respectively.
16
As before, the black (gray) circles at the bottom of the figure mean inclusion in (exclusion from) the regression. Specification chart by spending component. Note: The figure considers the dynastic link with the vice mayor as the independent variable and the different components of spending as the dependent variable. Each bar represents a specification of a spending component, with the full or bandwidth sample, using no, linear or quadratic functions of the running variable. The black diamond represents the point estimate, while the light (dark) gray bars are the 95% (90%) confidence intervals. The black (gray) circles mean inclusion in (exclusion from) the regression.
We see that among the components of municipal spending, the mayor–vice mayor dynastic linkage has a moderate impact on economic services and health. Given that the health system has been decentralized while the education sector has remained centralized, it is interesting to observe positive impacts on health but null effects on education. Unfortunately, the vagueness of the local government data does not lend itself to further theorizing. It is worth noting that the increase in economic services and health should ostensibly lead to higher economic growth and lower poverty. If increased spending, particularly in these specific components, does not lead to increased development, then this strengthens our claim that waste and corruption are driving the spending result.
Testing H3: Horizontal Dynasties and Economic Development
We now empirically test the relationship between horizontally dynastic mayors and economic development. The argument posits that in the context of the Philippines—rampant rent-seeking, vote-buying, clientelism, weak bureaucratic institutions, and chronic government mismanagement,—higher government spending associated with horizontally dynastic mayors does not lead to higher levels of development. Similar to Hypothesis 1, the treatment group (horizontally dynastic mayors) consists of mayors who have relatives simultaneously elected for any of the following: governor, vice governor, mayor, vice mayor, or congressperson. The control group then includes mayors who have relatives who failed to get elected (in the same year) for any of the above local positions.
Similar to our previous regressions, we estimate the following model but with the development outcomes as our dependent variable
Note: Robust standard errors clustered at the province level are in parentheses. Dependent variable is log night lights per capita in a municipality or city. Horizontal dynasty is a dummy that takes the value 1 (0) if the mayor’s relative wins (loses) the election. ∗∗∗,∗∗, and ∗ denote statistical significance at 1%, 5% and 10% level, respectively.
Note: Robust standard errors clustered at the province level are in parentheses. Dependent variable is poverty incidence in a municipality or city. Horizontal dynasty is a dummy that takes the value 1 (0) if the mayor’s relative wins (loses) the election. ∗∗∗,∗∗, and ∗ denote statistical significance at 1%, 5% and 10% level, respectively.
The takeaway is that horizontal dynastic status leads to higher levels of government spending but not to higher levels of development. In particular, the results suggest that increased spending is potentially driven by some combination of mismanagement, poor government performance, corruption, and rent-seeking. corruption and rent-seeking.
To test the RD’s identifying assumptions, we augment our previous test for covariate balance to include the night lights and poverty variable. We find that covariate means are smooth around the cutoff (see Supplementary Figure D2). In addition, we check for potential manipulation using the McCrary (2008) test and find no evidence of manipulation (see Supplementary Figure D5). 17
Identification Checks
Political Parties and Political Dynasties
Despite the significant effects documented in the previous sections, it is possible that what is driving our results is not horizontal dynastic status per se, but the fact that members of the same family are more likely to belong to the same party. Party alignment may also serve to align preferences, induce coordination, and increase government spending in the municipality (Aldrich, 1995). To address this concern, we include a control for whether the dynasty pair belongs to the same party or not. As the results in Table 3 Panel B show, being in the same party does not matter, or when it does, the effect of family similarity is not diminished and remains statistically significant. This result corroborates our first hypothesis: horizontally dynastic mayors spend more than non-dynastic mayors, even after controlling for belonging to the same party.
Apart from the statistical results, there are conceptual reasons as to why parties do not serve as vehicles for coordination in the context of the Philippines. Parties in the Philippines have low levels of party institutionalization, little differentiation between parties in terms of ideology, and fleeting alliances of convenience rather than stable unions of like-minded politicians (Hicken, 2009; Montinola, 1999). In particular, these fleeting alliances of convenience suggest that politicians from the same party cannot work together over policy because they cannot trust that their preferences will be aligned throughout the span of the policy. For example, two politicians from the same party may agree to build a road, but one politician may jump to a rival party in the middle of road construction. This would potentially prevent the road from being completed, and in the case of a rent-seeking politician, potentially prevent extraction from taking place.
Local Resources and Family Networks
Note: Robust standard errors clustered at the province level are in parentheses. Dependent variable is log spending per capita in a municipality or city. Horizontal dynasty is a dummy that takes the value 1 (0) if the mayor’s relative wins (loses) the election. Family networks is the number of members in a political dynasty who run for any local office within the province. ∗∗∗,∗∗, and ∗, denote statistical significance at 1%, 5% and 10% level, respectively.
This result synergizes well with existing results, also on the Philippines, that emphasize the importance of political networks for politicians’ electoral success (Cruz, 2019; Cruz et al., 2017). These papers argue that politician–voter networks facilitate political exchange, group monitoring, and clientelism. This result builds on these papers by noting a potential limitation of political networks. We show that intrafamily elite networks do not necessarily lead to preferred policy when one member of the family is not in power. Resources or innate ability may not be enough; what is needed is an institutional means to affect policy.
In Supplementary Appendix G, we discuss the potential role that other alternative mechanisms—politician experience and state capacity—could be playing in our results. Reassuringly, we find that neither of these mechanisms is driving our results. 19
Conclusion
This article has argued that horizontally dynastic Philippine mayors increase levels of government spending. This result is being driven by mayor–vice mayor horizontal linkages, which suggest that the mechanism behind increased spending is the removal of institutional constraints. But increases in spending do not lead to higher levels of development, indicating that spending may be a source of rent extraction or some other type of inefficiency. We tested this argument’s implications by using a regression discontinuity design leveraging close elections for local political office. Our results show that: (i) horizontally dynastic mayors have higher levels of spending than non-dynastic mayors, (ii) direct institutional constraints are a relevant mechanism mediating the above result, and (iii) horizontally dynastic mayors do not lead to higher levels of development in terms of higher economic growth economic growth or lower poverty.
Although empirically specific to the Philippines, the argument provides a general framework exploring how horizontal political dynasties navigate the constraints foisted upon politicians by political conflict. If political actors can organize themselves in social groups that (i) share similar preferences over political goals and (ii) concurrently hold positions of political power, then these groups may be able to strategically place themselves across various political positions in a way that undermines formal checks and balances. Potential scope conditions for our argument include: (i) different political offices with the potential for political conflict due to politicians’ potentially divergent preferences, and (ii) the weakness or inability of formal institutions, such as political parties, to coordinate behavior across offices.
This article builds on a growing body of work on the consequences of political dynasties. We contribute to this literature by examining how the link between a dynasty’s horizontal structure interacts with the formal rules of the political system to facilitate policymaking. We also build on literature specific to the Philippines, emphasizing the role of political networks in enabling political goals. We add to this literature by emphasizing the role of formal rules and how they interact with (dynastic) political networks.
Although we have explored an argument of how dynasties function, more work needs to be done on this topic. For example, it would be fruitful to consider how dynastic strength affects economic and political outcomes. Furthermore, existing research has not yet examined the causes or consequences of dynastic prevalence across forms of government. The potential differences between dynastic functioning in strong and weak democracies, or in autocracies, are still a relatively unexplored area, especially in the theoretical sense. Ultimately, understanding informal political structures like political dynasties provides us with a deeper comprehension of the ecosystem of organizations that politicians form to circumvent or aid formal institutions like political parties, providing us with a richer appreciation of the dynamic elements of politics.
Supplemental Material
sj-pdf-1-cps-10.1177_00104140211024292 – Supplemental Material for When Running for Office Runs in the Family: Horizontal Dynasties, Policy, and Development in the Philippines
Supplemental Material, sj-pdf-1-cps-10.1177_00104140211024292 for When Running for Office Runs in the Family: Horizontal Dynasties, Policy, and Development in the Philippines by Dean Dulay and Laurence Go in Comparative Political Studies
Footnotes
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.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
