Do government subsidies for impure public goods crowd out private donations? The case of zoos in Germany

Do government subsidies for impure public goods crowd out private donations? The case of zoos in Germany University of Heidelberg, Germany June 11, 2014 Abstract This paper examines the relationship between public support and private donations for impure public goods by analyzing the crowding eect for zoos in Germany. In contrast to purely public goods, the crowding effect on private donations is not only determined by government grants, it is also determined by the amount of revenues, by the number of potential contributors and by changes in governmental support. This paper adds to the literature of econometric research on public goods by analyzing the crowding eect with panel data of an impure public good rather than a simple public good. We nd signicant results of a crowding-out eect for absolute values under the 2SLS specication, however a crowding-in eect for per capita public support and donation levels. Moreover, we nd that the legal form, which has not been analyzed before, has a signicant impact on private giving when the zoo is a stock corporation. Keywords: Crowd-out; Government grants; Donations; Impure public goods JEL classication: H41; H42 1 Introduction The question whether government grants need to support public goods remains widely discussed. However, the question Do we need to support semi-public or impure public goods? opens up discussions which determine many parts of our daily lives. Whether we think of banks, which are semi-public and expect us to bail-in for potential losses, or a simple recreational example of zoos, which receive private donations and government grants from taxpayers' money. Both examples show that the gift of giving is often a burdon doubled, either through direct giving or indirectly, through taxpayers' income. Public goods are goods for which the market most often cannot supply the optimal amount by itself as the incentives to free ride are greater than the incentives to contribute. For this reason, the level of voluntary provision of a 1

public good fails to achieve the ecient level (see Bergstrom, Blume and Varian; 1986 and Bernheim; 1986). As markets of public goods often do not work, government intervention and governmental support are crucial for public goods to further exist. To overcome this problem, dierent options are available to the users, either by implementing a tax or granting a subsidy. However, public goods are often in fact semi-public, and by linking the use and benets of a public good to private consumption, one can increase the incentives to invest in a public good. Examples of semi-public goods, or impure public goods, are multifold, and there are dierent examples of where bundling private consumption to public investment have increased welfare throughout. In environmental economics, one example is given by agroforestry, where farmers have an incentive to invest in land and maintaining the environment and habitat for species. In public economics, one example is given by the tax relief people often receive when donating to a charitable institution. In energy and resource economics, one example is given by the use of renewables in Germany, where people receive a monetary compensation from the government when substracting their energy from renewable energy sources. In this paper, we will consider the case of zoos in Germany as these public-private institutions often receive government grants as well as private donations. There is a wide range of economic literature which discusses whether publicprivate relationships are useful and desireable. One direction proposes that governmental support for public goods crowd-out private donations. Warr (1982) and Roberts (1984) denote that private donations are crowded-out by government grants one-for-one, meaning that one dollar given by a government grant crowds out private donations by one dollar, leading to no change of the level of public good provision. Abrams and Schmitz (1978) conclude that crowdingout occurs, however less than one-for one. Andreoni (1989) also constitutes a partial crowding-out eect, as individuals not only receive utility from the total level of public good, but from the individual contribution itself, also known as the feeling of warm glow. Kotchen and Moore (2007) analyze dierent participation mechanisms on the private level for impure public goods, compared to pure public goods. Moreover, in Kotchen (2006), we can conclude the eect of government provision on the provision of a public good with respect to an impure public good. Empirical work has tested for this theory, often following the conclusion that the crowding eect is less than one-for-one. Even though most researchers nd this result, the question whether there is a crowding-in or crowding-out eect remains ambiguous. Kingma (1989) concludes that crowding-out occurs when testing for donations for public radio stations. He nds that the impure altruist model is more appropriate when scrutinizing charitable contributions, meaning that donations do not necessarily follow a purely altruistic motive. Moreover, empirical research constitutes a dierent side, that government grants crowd-in donations. One reason for this ambiguous result is that government grants work as a signal of investment worthiness to increase the level of private donations. For instance, Khanna and Sandler (2000) nd that government grants cause signicant crowding-in of voluntary donations to UK char- 2

ities. Payne (1998) obtains similar results when analyzing academic research institutions. To grasp an environmental eect which may aect contribution behavior in a dierent way, Heutel (2007) concludes that social charities and environmental charities behave in a totally dierent way. Whereas government grants to social charities crowd-in donations, this result no longer holds with signicance for environmental charities. Moreover, crowding-out can manifest in both directions, i.e. in response to the level of private donations, the government may alter its funding (Heutel 2009). An example of involuntary contributions to public goods is discussed by Andreoni (1993), for which he founds that voluntary contributions are incompletely crowded-out by involuntary contributions by means of a per-capita lump sum tax. Moreover, subjects who are forced to contribute are signicantly more cooperative. Borgonovi (2006) analyzes a panel of theatres in the U.S. and nds out that the crowding eect induced by the level of public support takes an inverted U-shape. At relatively low levels government or local subsidies crowd-in private donations while at higher levels it displaces them. In this paper, we will reanalyze the theory of the crowding eect, however with a stronger focus on impure public goods. Zoos in Germany oer a unique example to provide insight into the crowding eect of semi-pure public goods, as they exhibit both purely private characteristics by being a spot for recreational activities, and public characteristics, as animal viewing cannot exclude other visitors from this activity and in a broader sense, zoos are supposed to contribute to wildlife conservation. Zoos are facing increasing popularity in Germany. Having more than 300 zoos and nature parks, the country experienced 34.3 million zoo visitors in 2011 for only those zoos belonging to the German Association of Zoo Directors (VDZ, 2013). As the country faces less than 82 million inhabitants, the number of zoo visitors is relatively high compared to overall population and zoos remain one of Germany's favorite recreation spots. Even though most zoos charge entrance fees, they still depend on great amounts of governmental grants each year. The structure of the paper is the following. We will rst introduce a basic model of governmental subsidies for impure public goods to determine the eect of government grants for an impure public good on the provision of that good. Then we will conduct econometric analysis to determine the role of governmental subsidies on private donations for impure public goods in Germany with respect to zoos. The data set contains 21 zoos in Germany over a time period of eight years, between 2005 until 2012. The data is unbalanced, leading to a total of 149 observations for the dependent variable donations. Data is obtained from the annual income statements of zoos in Germany, which are published at Bundesanzeiger, the ocial site of the German government for annual income statements and balance sheets, and which are publicly available. If the income statements were not available via Bundesanzeiger, the data was obtained from annual income statements of local governments in Germany, so-called Beteiligungsberichte, which often publish individual income statements of zoos when the zoos received government grants from the local government. For our inde- 3

pendent variables, we do not only consider government grants. Moreover, we want to include generating revenues from other activities than donations and government grants. These revenues can include entrance ticket revenues, revenues from local shops and revenues from zoo membership fees. This stands in contrast to the existing literature on crowding-in and crowding-out, as previous literature analyzes non-prot organisations or charities, with no real usage eect. To extent the literature on the crowding-eect, we also consider the legal form of a zoo, as these dier across zoos. One would presume that non-prot organisations receive higher grants than stock corporations. We will provide evidence that this assumption does not hold. We test for the OLS model and xed-eect and random-eects model. We nd clear results that the crowding eect for the 2SLS-model is highly signicant. However, the growth rates of government grants have no signicant impact on private donations. The legal form or status of a zoo plays an important role, leading to signicant positive results as well. We will test for average individual contribution levels as well (along the lines of Andreoni), to test whether this makes a dierence. We nd signicant results of crowding-out for aggregate contribution levels by using a 2SLS approach, however signicant crowding-in for individual contribution levels under the normal OLS estimation. These dierent estimation procedures result from the fact that aggregate revenues are highly endogenous and correlated with other independent variables, whereas revenue per visitor is exogenous. In chapter 2, we will discuss the crowding eect from a theoretical point of view. In chapter 3, we will describe our data set we use to determine whether there crowding-in or crowding-out occurs for German zoos. In chapter 4, we will conduct econometric analyses to determine this eect. In chapter 5, we will provide robustness of our results, by including the fact that revenues can be highly endogenous, determined by the total number of users of an impure public good. In chapter 6, a conclusion will follow by discussing our results and ndings. 2 The theoretical model Visitors and non-visitors of zoos have the opportunity to make a voluntary contribution to the provision of a semi-public good or impure public good. Impure public goods are considered impure, when users can pursue an activity which generates both a private and public good. In our case, users provide for a private good X, which can be interpreted as recreation, amusement or individual time spending, and a public good Y. In the context of zoos, we can interpret the public component as animal viewing or biodiversity conservation. X is determined by the entrance ticket and other revenue-generating activities (except for donations and government grants) people pay to consume the quasi-public good, i.e. X = αg i. The total provision of the public good, is determined by 4

the aggregate level of contributions and by the level of government funding, such that Y = n i=1 Y i + Y GOV, where Y i is household i's contribution and Y i = βg i + d i with d i denoting donations. In this case, g i denotes the revenuegenerating activity, and Y GOV denotes the amount of government funding. Thus, the impure public good is provided via three ways of nancing: impure public good expenditures, i.e. revenues, donations from the private or corporate sector, and public funding by means of government grants. Each household takes the contribution of all other households, denoted Y i = Y Y GOV Y i, as exogenously given (the Nash assumption) and solves the following utility maximization problem: max {U(y i, Y i + Y i + Y GOV ; θ) p g g i + p d d i = m i } (1) g i,d i This setup resembles the standard model for private provision of an impure public good without the conventional purely private good (Kotchen 2006). We can add G i to both sides of the budget constraint in (1) and rewrite the household's problem with a choice over the aggregate level of Y rather than g i : { max U(y i, Y i + Y i + Y GOV ; θ) p β } X i,y α X i + Y = m i + [Y i + Y GOV ] ; Y i Y (2) where the additional constraint Y i Y follows from the nonnegativity of Y i. The price of recreation p X is given by p β α. The solution to the problem yields a continuous demand function for the public good Y that can be written as Y = max {f(m i + Y i + Y GOV ; θ), Y i + Y GOV } (3) wheref( ) is demand for Y ignoring the inequality constraint. We assume that Y is a normal good, which implies that f( ) is strictly increasing. Now substracting Y i + Y GOV from both sides of (3), we have each household's best response function for a contribution: Y i = max {f(m i + Y i + Y GOV ; θ) Y i Y GOV, 0} (4) Letting Y denote an equilibrium level of contributions to nance the public good, we can solve for each household's equilibrium contribution. Assume Y i Y in (2), invert f( ), and add Yi to both sides. Solving for the household's contribution yields Y i = m i p X [f 1 (Y ; θ i )+Y ]. Now we will dene a critical level of income m (θ) = f 1 (Y ; θ) Y. We can then write a household's equilibrium contribution to the impure public good as { Yi 0 ifm i m (θ) = (5) m i p X m (θ) ifm i m (θ) 5

From (5) it follows that an increase in governmental subsidies, Y GOV, increases demand for the public good Y, because f/ Y GOV 0, however individual contributions decrease, as long as Y is a normal good, i.e. f/ Y GOV 1. This implies that if governmental grants for zoos increase, individual contributions should decrease, because Y i / Y GOV = f/ Y GOV 1 0 and if revenues from visitations remains unchanged, donations should be crowded-out either completely or partially, less than one-for-one. However, demand for the impure public good is determined by two activities, donations and revenues. The question arises, whether an increase in government funding crowds-out donations or revenue-contribution or both. 3 The data set For our data set we use a panel of 21 zoos in Germany over a time period of 8 years, ranging between 2005 and 2012. The data is unbalanced, leading to a total number of observations of 149 observations for the dependent variable. The data was collected from the income statements and balance sheets of dierent zoos in Germany. The data was either taken from the German Bundesanzeiger, a webpage which lists all balance sheets and income statements of dierent companies, or from dierent municipal balance sheets and income statements, when cities provide for funding for dierent zoos. Some data is publicly available, mostly for stock corporations. The panel contains zoos with dierent legal forms, from stock operations (German: AG) to private operations, and from non-prot (ggmbh, e.v.) to prot-maximizing private operations (German: GmbH). As to our knowledge, this topic has not been analyzed in the literature of crowding-out before, as charity organizations are considered as non-prot organizations. In table 1, we can nd the total number of zoos listed according to their legal form and distinguished between zoos located in the East of Germany and in the rest of Germany. Around 50 % of zoos located in East Germany were non-prot Ltds. The number of stock corporations in other regions were relatively higher (20%) than in East Germany (16.67%). Zoos which are Ltds make up over 50 % in the rest of Germany, compared to only one third in the East region. 6

The data also contains the level of government funding, resulting directly from local (city-based) governments. Funding on a country-level is left out in our analysis, as almost all zoos are managed on a local level. We also collected data for the number of employees each zoo had on average, as this works as an indicator for the size of the zoo. The numbers of visitors for each year is also included into our dataset, as this reects size and attraction as well. Moreover, in contrast to previous papers on the crowding eect, we have included revenues for each zoo for each year as well, in order to capture the eect of revenues on donations. As each visitor contributes to zoos by paying entrance fees, we want to capture the eect of this impure public good on donations. Donations can be linked to the level of revenues, as visitors or donors might feel that they are contributing to the impure public good by sole usage. In Graph 1, we can see the evolvement of average donations, government grants and revenues for German zoos between 2005 and 2012. A spike in government grants should lead to a decrease in either donations or revenues, or both. Government grants appear relatively constant over the years, except for a spike in 2011. We cannot notice a downturn in donations or revenues for that year, presuming that a crowding-eect is not apparent. 7

All numbers are CPI-inated with 2010 as the base year and we have included some dummy variables such as region to incorporate regional dierences. We have also made a distinction between the East Germany and West Germany, as there are great disparities between income across Germany. 4 Econometric analysis We want to determine the eect of government grants on voluntary contributions, donations i,t, i.e. donations i,t = f(r i,t ; GG i,t ; visitors i,t ; unemployment jt ; θ) (6) donations i,t denotes the donations per zoo i over a period of time t 1, R i,t denotes the revenues of zoo i between 2005 and 2012, GG i,t denotes government grants received by a zoo i between 2005 and 2012. The number of visitors is given by visitors i,t. The independent variable employee i,t captures the average number of people working in the zoo for each year. 1 the data on donations was sometimes not explicitly given in each income statement, depending on the size of each zoo. If the zoo did not publish these results, the number for other income from the income statement was taken. As other income also contains e.g. rentals if zoos have real estate, this number is slightly biased. However, changes in other income can be accounted to changes in donations, as other income sources are relatively stable. 8

The parameter θ incorporates individual characteristics such as the geographical location within Germany, and the legal form of the zoo. The legal form of the zoo is important as there are a fair amount of non-prot zoos as well as prot-maximizing zoos which receive government grants for nancial support. Moreover, non-prot zoos receive greater government grants to capture the losses these zoos make. We can specify donations as: D it = α i,t +β 1i,t GG it +β 2i,t R it ++β 3i,t visitors it +β 4i,t unemployment jt +β 5i,t θ i +ɛ 1it (7) where α i,t denotes a constant, β 1i,t,..., β 4i,t are coecients to be determined and ɛ 1it denotes the error term. We intend to determine the eect of local government grants to zoos on donations, i.e. d(donations)/d(gg) to determine whether there is a crowdingout or crowding-in eect. If government grants crowd-in donations more than one-for-one, it should hold that dd/dgg 1, if it crowds-in less than one-for-one, 0 dd/dgg 1. And if government grants crowd out private donations, dd/dgg 0. We have also included the growth rate of government grants as a potential independent variable (along the lines of Borgonovi 2006), and revenue per person to incorporate individual contributions. In total we have N x t observations with some observations for 2012 missing due some income sheets not yet published, leaving to 149 observations of the dependent variable. The next table shows the description of variables and summary statistics for our panel. The amounts are reported in constant dollars, with 2010 as the base year. From our table, can see that the maximum value of government grants was over 27 million Euros. The uctuation in the growth rate of government grants was relatively low. The average private donation is less than and around half of the average value in government grants. Revenues include revenues for zoos achieved from their operating activity, excluding governmental support in funding. The variable visitors contains the number of visitors to a specic zoo and amount as a maximum value to almost 3,2 million for the Berlin Zoogarten zoo in 2007, which is a stock corporation. 2 The number of employees range from a minimum of 20, to a total of 529 employees. The number of employees and visitors for zoos which are stock corporations tend to be higher. The revenue per person is on average 5.73 Euros, with a maximum value of 19.79 Euros. Revenues face high volatility, with a standard deviation of over 5,700 thousand euro. 2 Visitors with an annual card allowing for unlimited entry to a zoo are counted dierently, depending on each counting method of a zoo. Some zoos count an owner of an annual card as a 20-times yearly visit. Other zoos equal annual cards to a ten-times visit. 9

Table 3 further distinguishes between each variable and the variation of each variable between the dierent zoos and within each zoo. In contrast to government grants, there are strong dierences between the variation of donations between all zoos and within each zoo over the time period of 8 years. For government grants, GG it, the between variation does not vary to a great extent from the within variation. This is due to the fact that in one year, a zoo can receive a great amount of local funding, whereas in the following year, it can receive zero funding, also depending on the losses each zoo makes. The numbers for the growth rate in government grants shows a greater variation within each zoo over the years than between zoos. The mean of revenues amounts to almost 4,738 thousand, with a strong between variation of almost 5,474 thousand. For all variables, the between-variation is stronger than the within-variation, leading to greater disparities between dierent zoos. As the legal form for each zoo does not change over time, this variable is used for a time-invariant or xed eect for each zoo. The variation between zoos regarding revenue per capita is more than twice as great as the within variation for each zoo. Table 4 shows the correlation matrix for the independent variables. Government grants are hardly correlated with donations (0.0351). Along the lines of Borgonovi (2006), we use both the absolute value as well as the growth rate of government grants, supposing that present values are in- uenced by its predecessors. However, we do not use the change in government 10

grants, in order to prevent multicollinearity between the independent variables, but growth rates between each period instead. Thus, donations denote the absolute value of donations, and GrowthGG it denotes the growth rate of government grants. The growth rate of government funding is negatively correlated with donations and hardly shows any correlation (-0.01138). 11

Revenues however, are strongly correlated with donations (0.6391), and highly correlated with the number of visitors (0.9495). One more visitor increases revenues by 1, leading to a one-on-one relationship. This is no surprising result, as the number of visitors should increase total revenues. Moreover, 12

revenues are strongly correlated with the legal form with 0.7125, i.e. stock corporations and Ltds should be able to generate higher revenues than non-prot Ltds and registered associations. The number of visitors is positively correlated with the level of donations (0.6889). The fact that there is a positive correlation between the number of visitors and the level of donations is also explainable, as the more people visit a zoo, the rate of potential donors should increase. The unemployment rate for each location of the zoo faces a correlation of 0.4027 with donations. Moreover, the number of employees working in a zoo are highly correlated with total revenues (0.7326), the number of visitors (0.6966) and donations (0.5832). Next, we run the regression to see if there is a crowding eect for the impure public good, and if government grants crowd-in or crowd-out private donations to German zoos. Moreover, we want to determine whether revenues are substitutes or complements for donations, i.e. if revenues crowd-in or crowd-out. 13

In the rst regression, we run a simple OLS estimation with donations as the dependent variable and only government grants GG it as the independent variable. We also included a time variable T ime as well as some dummies regarding the legal status and the area in Germany (East region). Surprisingly, a stock corporation crowds-in donations and is highly signicant (2,840,801), as well as a zoo situated in the East of Germany. One suitable explanation would be that stock corporations tend to be bigger that normal Ltds, leading to more visitors and more potential donors. In the second regression, we included revenues into our analysis and control for the city's unemployment rate. In this specication, government grants crowd-out donations signicantly at the 5 % level, however less than one-for- 14

one. This means that an increase in government funding by one, decreases donations by 0.219. Moreover, revenues crowd-in donations (0.311) and our result is signicant at the 1 % level, meaning that revenues work as a complement for donations. This would imply that people do not consider their sole usage as a substitute for contributing. Moreover, our constant has a negative sign and is highly signicant, concluding that donations drop signicantly if there were no revenue income, or government grants. The independent variables explain 62.6 % of the variation in the dependent variable. In the third regression, we included the number of visitors for our specication. The number of visitors crowd-in donations, one additional visitor increases donations by more than 2.316 euros. Although government grants crowd-in donations in (3), our results are insignicant. The location of a zoo in the East of Germany has a positive impact on donations and is signicant at the 10 % level. Revenues slightly crowd-out donations, however insigncantly. In the fourth regression, we have included both the absolute value of government grants and the growth rate in state funding into our analysis. In this case, government grants still crowd-out donations by -0.154%, however the result is insignicant, and the same holds for the growth rate of government funding. As Andreoni (1993) introduced the concept of warm glow, visitors could retreive utility out of individual contribution rates rather than absolute contribution levels. For this reason, we dierentiate between total values and per capita values, as the number of visitors has a great impact on the crowding eect. Moreover, we watnt to dierentiate between random and xed eects estimation. In our example, such a variable could be the legal status of a zoo: a level-one variable, since it varies over all zoos, the level-one units. The xed eect of this variable is the average eect in the entire population of organisations, expressed by the regression coecient. Since most often it is not assumed that the average eect of an interesting explanatory variable is exactly zero, almost always the model will include the xed eect of all explanatory variables under consideration. For this reason it will be necessary to specify also a random eect of this variable, meaning that it is assumed that the eect varies randomly within the population of zoos, and the researcher is interested to test and estimate the variance of these random eects across this population. Such an eect is also called a random slope (Everitt and Howell ; 2005). In Table 6, we run a simple regression with per capita values to determine dierent in crowding-in or crowding-out. For our OLS estimation, we nd highly signicant results for government grants per visitor, i.e. the government grant associated to each visitor crowds-in an individual donation by 0.147. This means that a one-euro increase in per capita government funding increases per capita donations by 0.147 euros and this result is signicant at the 1 % level. However, the number of visitors does not impact each individual's donation. Surprisingly, under an OLS estimation, revenue per person crowds-out each individual's donation (-0.039), however crowds-in under xed and random eects estimation, with both results being insignicant. Under the random effects model, the stock corporation crowds-in donations by 2.187 and remains 15

signicant at the 10 % level. The variation in the dependent variable can be explained by the independent variables to a low extent, implying that there might be other variables missing and which are not specied in the model. 5 Robustness of Results There are several ways to estimate the measurement of crowding-in or crowdingout. The biggest concern is whether there are omitted variables in the empirical specication that are correlated with the revenues or government expenditures. 16

5.1 Omitted Variables Omitted variables that are correlated with private donations and government grants will bias the measure of crowd-out in the OLS regression. Adding additional variables to include into our specication could try to overcome this problem and may be used to reduce any bias in the measure of crowding-out. One possible way to go about this would be to include the distance and connection of a zoo to the central transportation network of that area. This would imply using data from GPS software to determine the local position of a zoo. Another option would be to include the number of dierent species in each zoo to account for diversity as a measure of attractiveness. This variable is however, dicult to capture and not always published for each zoo for each year, as these numbers vary to a great extent throughout the year. 5.2 2SLS From our analysis, we can assume that the OLS estimation is not probably not appropriate, as multicollinearity among the independent variables might be present. Moreover, total revenues are driven by the total number of visitors. In contrast to previous studies on the crowding eect, we do not consider government grants to be highly endogenous, as the correlations appear to low. One reason for this is that government funding is often decided years in advance, depending on the local city's budget. This implies that there is no matching eect, which is also noticable from graph 1, which hardly shows uctuations for government grants. We therefore run a two-stage least-squares estimation to consider for potential endogeneity of revenues in order to capture potential crowding-in or crowding-out eects. In order to account for potential endogeneity, we need to nd explanatory variables which are not correlated with the error term, however correlated with our endogenous regressor. Visitors and the number of employees work as a good proxi for revenues. The system in this case is overidentied. We have tested for endogeneity of revenues before running the regression by conducting the Durbin and Wu-Hausman test. Under all specications, revenues appear to be endogenous with a signicant result ranging from a 1-% until a 10 %-level. For this reason, the 2SLS estimation conducts more reliable results for absolute donation levels. When taking into account the problem of multicollinearity, the system of equations is dened as: and D it = α i,t + β 1i,t GG it + β 2i,t R it + β 3i,t θ i + ɛ it (8) R it = β 1i,t visitors it + β 2i,t employee it + +ɛ it (9) 17

In (8) we estimate whether government grants GG it crowd-in or crowd-out donations D it. Moreover, donations are determined by the number of visitors. θ incorporates city-specic indicators as specied before, such as the local city's unemployment rate of regional-specic aspects such as the location in either the East of Germany or not, or the legal form. In Table 7, we can see the eects of crowding for donations to zoos on an absolute level. In all specications, we use donations as our dependent variable and government grants as one of our independent variables. In (1), we run a simple regression with revenues and government grants to determine the eect of government funding on donations. We nd a highly signicant eect of crowding-in for revenues and for the legal form of a stock corporation. This could reect the fact that stock corporations tend to be greater than smaller Ltds, attracting 18

more visitors. Moreover, visitors probably are not familiar with the legal form of a zoo. The legal status of a zoo for being a stock corporation increases donations signicantly under (1) and (2). Another reasonable explanation would be that stock corporations tend to be greater in size, leading to a higher number of visitors and with that, to more donations. Revenues are crowded-in by 0.186, meaning that an increase in revenues by one, increases donations by 0.186 and is signicant at the 1 % level. In specication (2), we add the unemployment rate in that particular city as an independent variable to the equation. This serves as a proxy for the wealth of that city in which a particular zoo is located. For all specications, revenues denote the endogenous variable, and is determined by the total number of visitors and employees, as employees reect a size of a zoo. For (2), we nd that government grants crowd-out donations (-0.265) and is signicant at the 5 % level. Revenues crowd-in donations signicantly, meaning that revenue contribution is considered a complement to donating rather than a substitute. More surprising is the unemployment rate, which increases donations rather than decreasing it. One reason could be that a higher unemployment rate in a city reects a social structure in that particular city. This could indicate whether people are more sociable towards giving. In specication (3) when including the growth rate of government grants as an instrumental variable and regional-specic dummies in order to determine the eect on donations, we still obtain signicant results of crowding-out of donations for government funding (-0.212) and a positive impact of an unemployment rate in the zoo's city it is located. A higher growth rate in government grants crowd-out donations, however our results are insignicant. In contrast to specication (1), a stock corporation as a legal status crowds-out donations signicantly (302,332). 5.3 Individual contributions In table 8, we use per capita values to determine whether a crowding eect exists for each individual's contribution level and government funding on the individual level. For this, we have conducted the Durbin and Wu-Hausman test to test for endogeneity for revenues per capita. In this case however, multicollinearity between variables did not exist. Due to this result, we have conducted a simple OLS regression as given in Table 6, and compared our results with the LIML and GMM estimator. In this case, government funding crowds-in donations under all three estimations, and is signicant at the 1 % or 5 % level respectively. Moreover, an increase in government grants per capita by one unit crowds-in donations per capita by 0.147 units. The number of visitors does not have an impact on donations per capita, and and revenues per capita crowd-out donations per capita. This implies that on the individual level, people regard usage payments for impure public goods as substitutes to donations. However, our results are not signicant. The legal 19

status of a zoo has no signicant impact on per capita donations, except for the GMM estimation, for which the stock corporation has a signicant impact on per capita donations at the 5-% level respectively. Revenues per capita crowd-out donations, however not signicantly. A legal status of a non-prot Ltds crowds-in donations, however not at a signicant level. A zoo located in the East of Germany crowds-in donations, however insignicantly. Overall, the independent variables hardly explain the variation in the dependent variable, with relatively low R 2 -values. 20

6 Conclusion This paper uses a data set coming from income statements for zoos in Germany in order to study the relationship between private donations, public grants and revenues. We nd that for absolute values, there is weak crowding-out of donations. Furthermore, zoos located in the East of Germany, which is considered a less stable economic environment, has a positive impact on donations. A higher unemployment rate in the city of zoo location also is an indicator for more pro-social behavior. More interestingly, stock corporations have a positive impact on giving. This could either result in the fact that users are unaware of the legal form, meaning that donors should have a higher incentive to donate to non-prot Ltds, or that stock corporations tend to be bigger, leading to a greater level of attractiveness for visitors. The legal form can work as a signal and inuences a user's incentive to provide for a public good. Normally, one would assume that a non-prot organisation would lead a private donor to contribute sooner than when the organisation works as a prot-maximizing institution. This is a novel result in the literature, as in previous papers, authors have studied non-prot charity organisations. When we consider individual average contributions, we nd signicant crowdingin. This means that government grants are necessary for zoos on an individual level, in order to incentivize people to donate to impure public goods. Furthermore, government grants should be related to the number of visitors a zoo faces, in order to capture the eect of crowding-in or crowding-out. Unfortunately some aws are present, as the data set consists of a small sample of zoos. This is due to a lack of data made available to the public. Stock corporations have more stringent publication policies regarding their balance sheets and income statements than Ltds. Due to this lenient behavior towards publication policies for Ltds, the data on donations faces an upwards bias, as the numbers in some cases contain income from rental activities of zoos when owning real estate. However, as other income activities can be considered relatively stable across time, donations are the number which drive strong dierences in numbers throughout the years. For policy implications, it would be wise to test whether zoo visitors take into consideration their individual contribution levels rather than aggregate contribution levels. This can be conducted through questionnaires, in order to capture preferences. Moreover, local governments should not dismiss the eects of individual contributions through revenues to zoos, in order to determine optimal prices for zoo businesses. Our results add to the literature in a novel way, as we include the legal form of institutions into our analysis in order to determine the eects of a legal status on contributional behavior. Besides that, we include the option of contributing to a public good by means of a private good, which resembles our revenue contribution each visitor makes. This approach is present in the theory of impure public goods, however not in empirical analysis. To adapt 21

results of empirical studies on public goods to impure public goods is erroneous and would lead to inconsistent estimates. Future empirical research should incorporate these dierent options for contributing to impure public goods. 22

References [1] Burton A. Abrams and Mark D. Schmitz. The crowding-out eect of governmental transfers on private charitable contributions. Public Choice, 33:2939, 1978. [2] James Andreoni. Impure altruism and donations to public goods: A theory of warm-glow giving. The Economic Journal, pages 464477, 1990. [3] James Andreoni. An experimental test of the public-goods crowding-out hypothesis. The American Economic Review, pages 13171327, 1993. [4] James Andreoni and Abigail Payne. Do government grants to private charities crowd out giving or fund-raising? American Economic Review, pages 792812, 2003. [5] T. Bergstrom, L. Blume, and H. Varian. On the private provision of public goods. Journal of Public Economics, pages 2559, 1986. [6] Douglas Bernheim. On the voluntary and involuntary provision of public goods. American Economic Review, 76:789793, 1986. [7] Francesca Borgonovi. Do public grants to american theatres crowd-out private donations? Public Choice, 126:429451, 2006. [8] B.S. Everitt and D.C. (eds) Howell. Fixed and random eects. In Encyclopedia in Statistics in Behavioral Science, pages 664665. Wiley, Snijders, Tom, 2005. [9] William Greene. Econometric Analysis. Pearson, 2012. [10] J. Hausman. Specication tests in econometrics. Econometrica, pages 1251 1271, 1978. [11] Fumio Hayashi. Econometrics. Princeton University Press, 2011. [12] Garth Heutel. Environmental and social service charities: Private and public sources of funding. Proceedings of the National Tax Association 100th Annual Conference on Taxation, pages 250259, 2007. [13] Garth Heutel. Crowding out and crowding in of private donations and government grants. National Bureau of Economic Research, pages 059, 2009. [14] Jyoti Khanna, J. Posnett, and Todd Sandler. Charity donations in the uk: New evidence based on panel data. Journal of Public Economics, pages 257272, 1995. [15] Jyoti Khanna and Todd Sandler. Partners in giving: The crowding-in eects of uk government grants. European Economic Review, pages 1543 1556, 2000. 23

[16] Bruce R. Kingma. An accurate measurement of the crowd-out eect, income eect and price eect for charitable contributions. Journal of Political Economy, 97:11971207, 1989. [17] Matthew Kotchen. Green markets and the private provision of public goods. Journal of Political Economy, 4:816834, 2006. [18] Matthew J. Kotchen and Michael R. Moore. Private provision of environmental public goods: Household participation in green electricity program. Journal of Environmental Economics and Management, pages 116, 2007. [19] Abigail A. Payne. Do public transfers crowd-out private charitable giving? some econometric evidence for the federal republic of germany. Kiel working paper, 1982. [20] Abigail A. Payne. Does the government crowd-out private donations? new evidence from a sample of non-prot rms. Journal of Public Economics, 69:323345, 1998. [21] Russell D. Roberts. A positive model of private charity and public transfers. Journal of Political Economy, pages 136148, 1984. [22] R. Steinberg. Does government spending crowd-out donations? interpreting the evidence. Annals of Public and Cooperative Economics, 62:591617, 1991. [23] Peter G. Warr. Pareto optimal redistribution and private charity. Journal of Public Economics, pages 131138, 1982. 24

APPENDIX 25

26

27