Fiscal Equalization,Tiebout Competition,and Incentives for Efficiency in a Federalist Country

Abstract

The purpose of this article is twofold. First, it measures the efficiency in the provision of public goods by local jurisdictions applying data envelopment analysis (DEA). Second, it relates efficiency scores to a fiscal equalization scheme designed to mitigate the negative consequences of Tiebout competition. The data come from the twenty-six cantons of Switzerland (2000–2004), a country characterized by marked federalism. Results show the equalization scheme to indeed have a negative influence on performance, resulting in an efficiency–equity trade-off. However, substitution of earmarked payments by lump-sum payments as part of the 2008 reform is likely to enhance cantonal performance.

Keywords

DEA efficiency measurement federalism fiscal equalization public finance Switzerland Tiebout competition

During the past decade, growing tax burdens have combined with ecological and equity concerns to increase citizens’ interest in the efficient provision of public goods. Economists have been responding to this interest by trying to provide information about government performance that may contribute to an efficient use of tax revenues. Examples of efficiency measurement of public services include Drake and Simper (2003), who examined police departments in England and Wales; Worthington and Dollery (2001), who estimated the efficiency of waste management in South Wales; and Worthington (2001), who focused on US and English public education. Grossman, Mavros, and Wassmer (1999) conclude that competition between US cities serves to increase their efficiency, in line with the Tiebout hypothesis. As to continental Europe, Afonso and Fernandes (2006), Afonso and Scaglioni (2005), and De Borger and Kerstens (1996), as well as Vanden Eeckaut, Tulkens, and Jamar (1993), examined the efficiency of Lisbon, Italian, and Belgian local governments, respectively. Specifically, De Borger and Kerstens (1996) find that the tax rate and income per capita have an insignificant effect on the performance of Belgian local governments, while federal grants have a negative influence. At the country level, Afonso, Schuknecht, and Tanzi (2006), comparing new European Union (EU) member and emerging market states, conclude that trade openness and transparency in government have a positive but insignificant effect on efficiency, while public trust in politicians fosters inefficiency.

These studies have not taken into account one feature of federalist countries that may affect efficiency at the local level, namely, fiscal equalization schemes. Fiscal equalization is designed to reduce horizontal and vertical fiscal imbalances that often exist between lower-level jurisdictions to provide public goods. This reduction is achieved by payments from jurisdictions with above-average fiscal capacity to jurisdictions with below-average fiscal capacity. In this way, below-average jurisdictions are to be enabled to produce public goods at average tax rates (Thöny 2005). Equalization schemes exist in most countries, among them the United States, the EU, Germany, Austria, and Switzerland—sometimes even at the community level. However, little attention has been given to the influence of such programs on the performance of both contributing and receiving member states. Indeed, disparities in the provision of public goods could even increase because jurisdictions on the receiving end may lack incentives for efficiency. The efficiency of contributing states may be undermined, too, giving rise to the well-documented equity–efficiency trade-off (Stiglitz 1988).

The contribution of this article therefore is twofold. First, it measures the efficiency of all twenty-six Swiss cantons between 2000 and 2004. Aggregate output performance indicators including six major public services are constructed to calculate cantonal efficiency scores based on robust data envelopment analysis (DEA). Second, calculated efficiency scores are related to the fiscal equalization scheme operated by the Swiss federal state both in its present and in its new (allegedly improved) form, controlling for socioeconomic factors that also have an influence on cantonal performance.

To the best of our knowledge, this is the first contribution undertaking a macroeconomic efficiency measurement of public good provision in a federalist country that takes the incentive effects of a fiscal equalization scheme into account.

This article is organized as follows. The second section provides some background information about Swiss federalism. The third section contains a review of efficiency measurement methods to argue that DEA is the method of choice in the present context. The data used are described in the fourth section. The fifth section is devoted to the presentation of results of the DEA and of a Tobit model estimating the effect of the fiscal equalization scheme on DEA efficiency scores. The final section concludes with an outlook and suggestions for future research.

Swiss Federalism

Cantons as the Producers of Public Goods

Switzerland, a federal state with its constitution dating from 1,848, distinguishes between three levels of government: federal, twenty-six cantons,¹ and approximately 2,600 communities. Public services are financed and provided at all three levels but with differing authorities. While the communities act under cantonal oversight, the cantons still constitute the backbone of the state. By article 3 of the Swiss constitution, they are responsible for all public services that are delegated neither to the federal state nor to their affiliated local authorities. Cantons are sovereign governmental entities with their own constitution and separation of power (legislative, executive, and judiciary), resulting in an extremely decentralized provision of public services.

Table 1 shows public expenditure on the twelve major service categories according to the three levels of authority. To the extent that Olson (1969) equivalence principle applies, expenditure by an authority also means provision. According to that principle, more than 60 percent of public good provision is estimated to be controlled by the twenty-six cantons and their affiliated communities. However, this share varies between categories; it is particularly low for military defense and foreign relations, which are delegated to the federal state. It is highest in education and health, which also constitute two of the most important overall expenditure items.

Table 1.

Functional Structure of Public Good Provision, 2004

Expenditure	Federal State	Cantons	Communities	Total
In million CHF
Administration	1,918	3,299	3,637	8,855
Public safety	728	5,287	1,955	7,970
[Military defense]	4,637	157	185	4,979
[Foreign relations]	2,427	—	—	2,427
Education	5,231	14,399	8,055	27,684
Culture and sport	447	1,380	2,422	4,249
Health	200	12,203	6,922	19,326
[Social welfare]	13,805	8,026	5,911	27,742
Transportation	8,547	2,873	2,991	14,411
Environment and spatial (planning)	728	1,019	3,159	4,907
Public economy	4,546	1,287	512	6,344
[Finance and tax]	9,411	−984	1,059	9,486
Total expenditure	52,624	48,947	36,808	138,379

Note: Swiss Federal Statistical Office, 1 CHF = 0.8 USD (2004 exchange rates).

Tiebout’s (1956) hypothesis predicts a positive relationship between fiscal federalism and performance of government. Similar to a free market economy, where consumers buy from the producer offering the best performance–price ratio, citizens choose the jurisdiction where they get the best ratio between public services provided and tax paid. In the case of Switzerland, cantonal autonomy in combination with direct democratic control through popular initiatives and referenda has resulted in considerable heterogeneity in the mode of provision. Since citizens can migrate and shift capital freely between cantons, they indeed expose them to Tiebout competition.

However, this hypothesis assumes that there are no externalities and that differences in performance are entirely due to the efficiency of administration. Externalities exist if citizens from one canton cannot be prevented from using services provided by another canton without paying. They typically arise in health care, education, and culture, although cantons with specialized hospitals do charge higher fees to patients from elsewhere, those with a university levy higher tuitions, and those with an opera house often make other cantons contribute to their operating expense. As to the efficiency of administration, there are disparities that are due to topographic, demographic, and socioeconomic conditions, constituting a handicap that cannot be overcome by the affected canton. Both confounding influences will be controlled for (see Service Categories Retained and Determinants of DEA Efficiency sections, respectively) when assessing the influence of Tiebout competition on cantonal performance.

Existing Fiscal Equalization Scheme

To overcome these disadvantages of fiscal federalism, Switzerland initiated a fiscal equalization program in 1959 to equalize cantonal disparities in the provision of public goods. According to an amendment of the federal constitution (article 135), cantonal disparities are to be mitigated with reference to Tiebout competition, the objective is to create a level playing field. By 2004, fiscal equalization has grown to some 1,000 CHF million of payments from the confederation to the cantons and another 1,500 CHF million between them. In relative terms, it totals almost 3 percent of cantonal and communal expenditure. The program is geared to the “financial potential” indicator, which has four components.

Financial potential is defined to increase with

Income: cantonal income per capita;

Tax power: taxable income, weighted by the tax burden per capita;

Tax burden: inverse of the cantonal and communal taxation as a share of income; and

Favorable topographic situation: share of a canton’s nonmountainous cropland in its total area, weighted by the number of inhabitants per unit of productive land.

A higher total index value results in less financial assistance. Figure 1 shows total payments per capita as of 2004. The canton of Zug (ZG) contributed the maximum of some CHF 1,250 (1 CHF = 0.8 USD in 2004) per capita to the program, followed by Basel-City (BS), Geneva (GE), and Zurich (ZH). At the other extreme, the 33,000 inhabitants of the canton of Obwalden (OW) in central Switzerland received some CHF 1,800 on average. In comparison, the extreme values of the German equalization scheme are a maximum of some CHF 600 paid by the land of Hessen and a maximum of some CHF 1,200 CHF per capita received by Berlin. These figures illustrate the importance of the Swiss fiscal equalization scheme.

One also needs to distinguish between earmarked (almost 70 percent of total) and general payments. While general payments can be used by the canton in ways it believes to generate the highest benefit for its citizens, earmarked subsidies may result in gold plating of projects and hence inefficiency (De Borger and Kerstens 1996).

Wrong incentives of the fiscal equalization program could have a sizable influence on cantonal performance and national welfare. Indeed, the existing program has been suspected of inducing the disparities it is designed to alleviate. Especially components 2 and 3 of the index formula are seen to create incentives for subsidized cantons to keep their tax burden high, for example, using their tax revenue for projects that contribute little to economic growth but enhance politicians’ popularity (Fischer, Beljean, and Fivaz 2003). In addition, cantons that are obliged to pay into the scheme have incentives to waste their money. They rather spend it on idle projects than give it to other cantons. These concerns have resulted in a reform proposal that passed a popular referendum in 2006. Starting in 2008, the share of earmarked payments was to be reduced to a minimum. Equalization payments are to be governed by a new formula, which distinguishes between resource and financial disparities.

Against this backdrop, this article seeks to answer two questions. First, does a fiscal equalization program as sizable as the Swiss contain incentives to provide public goods less efficiently, creating a trade-off between equity and efficiency? Second, does it matter whether transfer payments are earmarked or not?

Measuring Technical Efficiency with DEA

The characterization of Swiss cantons in the preceding section justifies viewing them as largely independent producers of a subset of public goods. For productivity measurement, they constitute decision-making units (DMU) that transform inputs into outputs, with productivity reflecting the quality of their administration. Following Koopmans (1951), technical efficiency in the provision of public goods thus can be measured with reference to a technology set $Γ$ ,

Γ = {(X, Y | Y \leq f (X))},

that describes the feasible set of input and output combinations

(X, Y)

of a production process. A DMU is called technically efficient if it lies on the boundary of

Γ

. On that boundary, it is not possible to produce more outputs Y for a given amount of inputs

\bar{X}

; or conversely, no smaller quantity of inputs X can produce a given output

\bar{Y}

There are various assumptions regarding the boundary of $Γ$ . For simplicity, we adopt those of Shephard (1970),

I s o X (y) = {x | x \in X (y), θ x \notin X, \forall 0 < θ < 1},

I s o Y (x) = {y | y \in Y (x), θ^{- 1} y \notin Y, \forall 0 < θ < 1} .

Here, the input and output isoquants

I s o (\cdot)

define sections with strong and weak technical efficiency, depending on the slope of the frontier, with θ denoting a scalar by which all inputs can be reduced without leaving the feasibility set. Accordingly, θ⁻¹ symbolizes the scaling-up factor for the outputs. However, the relevant technology set is almost never known in applied economic research, forcing the analyst to use observed rather than efficient input and output quantities. The pertinent methodology was developed by Farrell (1957); it has evolved into a distinction between parametric (econometric) and nonparametric (mathematical) methods (see Coelli et al. [2005] for respective overviews).

In public good provision analysis, DEA is the most common alternative. DEA dominates its main competitor, Stochastic frontier analysis (SFA) because DEA is more flexible (no specific functional form of the transformation process needs to be specified) and because DEA does not have to rely on price data for inputs and outputs, which often is lacking in the public sector.

DEA is the preferred technique for the present investigation, in particular because of lacking information about factor prices. Public sector accounts are notorious for neglecting capital user cost, and Switzerland is no exception. The DEA version employed here is an input-orientated one. The objective is to determine an efficient frontier $\hat{I s o X (y)}$ that is defined by the most productive DMUs. DEA amounts to solving a linear optimization problem for a particular ${D M U}_{C}$ or canton $c = 1, \dots, 26$ , with a 1 × K output vector $y_{c}$ and a 1 × M input vector $x_{c}$ .

\begin{matrix} {M a x}_{υ, ν} \\ s . t . \end{matrix} \begin{matrix} υ^{'} y_{c} \\ ν^{'} x_{c} = 1 \\ υ^{'} Y - ν^{'} X \leq 0 \\ υ, ν \geq 0 \end{matrix} .

Here, the 26 × K output matrix Y and 26 × M input matrix X represent the data for all twenty-six cantons. Thus, let a canton optimize its outputs

y_{c}

and inputs

x_{c}

by maximizing the distance between them valued using weights

υ

and

ν

. Note that these weights relate to the universe of all cantons and can be interpreted as shadow prices. Moreover, inputs are normalized to sum up (after weighting) to 1. The inequality

υ^{'} Y - ν^{'} X \leq 0

prevents outputs from increasing without bounds for a given bundle of inputs.

\hat{I s o X (y)}

is defined by those units for which

υ^{'} Y - ν^{'} X = 0

. Their efficiency score

\hat{{E F F}_{c}}

is 100 percent, while that of the other DMUs is given by their radial distance from the frontier.

However, the location of the efficient frontier strongly depends on the extreme DMUs (which lack comparators). One way to obtain robust DEA efficiency scores $\hat{E F F}$ is to iteratively exclude one DMU lying on the efficiency frontier. The new frontier then assigns this DMU a so-called super-efficiency score in excess of 100 percent without truncating the scores of the remaining DMUs (Andersen and Petersen 1993). The larger the super-efficiency of a DMU, the farther away it is from the remaining units in the technology set. Here, if this score is more than 1.5 times the distance between the twenty-fifth and the seventy-fifth percentile of all super-efficiency values, the pertinent DMU is excluded as an outlier (Thanassoulis 1999).

In a second step, the obtained robust efficiency scores $\hat{{E F F}_{c t}}$ (of cantons $c = 1, \dots, 26$ in year $t = 2000, \dots, 2004$ ) are related to a set of variables characterizing the Swiss fiscal equalization program and the disparities between cantons in order to address the two research questions stated at the end of Swiss Federalism section,

\hat{{E F F}_{c t}} = γ_{0} + γ_{1} Z_{1, c t} + γ_{2} Z_{2, c t} + φ_{c t} .

Commonly, Tobit estimation is applied to account for the fact that scores cannot exceed 1.00 (the lower limit of 0 is less relevant because it is never binding). Specification details are given in Estimation of the Determinants of DEA Efficiency section. Alternatively, one could also use the super-efficiency scores of the outlier detection process as a dependent variable and estimate ordinary least squares (OLS). However, to be consistent with the theoretical background of Farell (1957) and because the two variants lead to similar conclusions, we proceed with the more common robust efficiency scores as the dependent variable. See De Borger and Kerstens (1996) and Drake and Simper (2003) for applications of Tobit estimates.

Data

Service Categories Retained

The data come from the Federal Statistical Office, covering the years 2000–2004. As shown in Table 1, not all categories of services listed are predominantly subject to cantonal control. Moreover, the quality of data is insufficient for some categories. Therefore, only six of the twelve are retained for this investigation, namely, (1) administration, (2) public safety, (5) education, (7) health, (9) transportation and (11) public economy (they will be renumbered 1–6 in the following). Further, in order to exclude spillovers as far as possible, only primary and secondary education (without tertiary and vocational components), private road transportation (without regional public transportation), and farming and forestry are included in the analysis. More refined adjustments for spillovers (known to exist especially in health care) were not possible. They are controlled for in the second step Tobit estimation.

Constructing an Aggregate Output Performance Index

Measuring technical efficiency with DEA makes clear that the choice of output variables has an important influence on the results of a DEA. However, in public good provision, choice is no simple task because of two reasons. First, most of the outputs are not directly quantifiable. Second, public services are too many for individually entering them in a DEA. In this article, we try to overcome these difficulties by running a cost driver analysis and by constructing an aggregate output performance index.

Selecting the Output Variables

Since outputs of the public sector are difficult to measure, activity-based indicators serve as a substitute, in line with previous studies (Afonso, Schuknecht, and Tanzi 2006). In this article, two to six indicators for each of the six retained categories—twenty-two in total—are selected to proxy the output of public goods provided by a canton (see Table 2 ).

Table 2.

Output Indicators for the Six Governmental Activities Investigated

Public service	Output		Description, remarks
(1) Administration
Legislative, Executive	B1	Population	Population served and number of firms served as proxies for administration services provided
General administration	B2	Number of firms
(2) Public safety
Judicature	B1	Number of delinquencies	The assumption is that all citizens and dwelling units have the same preferences for public safety and a similar probability of suffering from crime and fire risk
Police	B2	Population
Fire department	B3	Number of dwelling units
(3) Education
Kindergarten	B1	Number of students	The numbers of kindergarten, primary, secondary, and high school students serve as indicators of output values. Quality of education is neglected because of lacking annual indicator
Primary education	B1	Number of students
Secondary education	B3	Number of students
High school education	B4	Number of students
(4) Health
Hospitals (specialized)	B1	Number of patient cases	Case-mix adjusted number of cases served as a severity-adjusted output for specialized and primary hospitals. The output of rehabilitation and psychiatric clinics, nursing homes, and retirement homes is measured by the number of patient days. No quality adjustment was possible because of lacking data
Hospitals (primary)	B1	Number of patient cases
Hospitals (psychiatric)	B3	Number of patient days
Rehab clinic	B4	Number of patient days
Nursing homes	B5	Number of patient days
Retirement homes	B6	Number of patient days
(5) Transportation
Cantonal roads	B1	Road length (km)	Road length serves as a proxy for maintenance. Number of cars is used as a utilization indicator for the roads
Communal roads	B2	Road length (km)
Road utilization	B3	Number of cars
(6) Public economy
Farming	B1	Farming area (km²)	The assumption is that farming and forest areas serve recreation purposes. The share of mountain area and organic farming area adjusts for differences in quality
	B1	Mountain area (km²)
	B3	Organic area (km²)
Forestry	B4	Forest area (km²)

Our selection was based on two concerns: choosing the most relevant variables and making sure that they cover the years 2000–04 for each canton. The relevance of the selected variables is checked with an analysis of cost drivers in the six service categories. Therefore, the dependent variables are category-specific real expenditure $C_{c t}^{j}, (j = 1, \dots, 6)$ of canton $c, (c = 1, \dots, 26)$ in year $t, (t = 2000, \dots, 2004)$ . They are related to the output indicators $(Y_{k, c t}^{j}, k = 1, \dots, K_{j} f o r 2 \leq K_{j} \leq 6)$ and a time trend $({t r e n d}_{t}^{j}, t = 1, \dots, 5)$ ,

\begin{matrix} C_{c t}^{1} & = & β_{0}^{1} + β_{1}^{1} Y_{1, c t}^{1} + β_{2}^{1} Y_{2, c t}^{1} + \dots + β_{K_{1}, c t}^{1} + α^{1} {t r e n d}_{t}^{1} + ε_{c t}^{1} \\ ⋮ \\ C_{c t}^{6} & = & β_{0}^{6} + β_{1}^{6} Y_{1, c t}^{6} + β_{2}^{6} Y_{2, c t}^{6} + \dots + β_{K_{6}, c t}^{6} + α^{6} {t r e n d}_{t}^{6} + ε_{c t}^{6} \end{matrix} .

Since cantons are exposed to similar shocks, error terms

ε_{i t}^{j}

are likely to be correlated, calling for seemingly unrelated regression estimation (SURE).

Depending on the service category, SURE confirms the relevance of the selected twenty-two output indicators and the correlation between the service categories. Pertinent econometric results are shown in Table 3 together with the correlation matrix of the residuals. Two criteria were applied to judge the relevance of the output indicators. First, they need to be positively related to cost as a summary measure of input qualities, in keeping with the production theory laid out in Measuring Technical Efficiency with DEA section. Second, they should importantly contribute to the explanatory power of the cost driver analysis, indicated by the significance level of their coefficients.

Table 3.

Seemingly Unrelated Regression Results for Six Public Service Categories^a

Variables	(1)	(2)	(3)	(4)	(5)	(6)
B1	9.74E+04^*** 2.04E+04	−7.18E+04^** 2.22E+04	1.25E+08^*** 1.91E+07	1.61E+05^*** 2.42E+04	3.17E+04^*** 4.83E+03	2.50E+00^*** 9.33E+02
B2	1.55E+05^*** 2.14E+04	1.76E+05^*** 5.19E+04	2.19E+08^*** 1.35E+07	3.07E+05^*** 2.26E+04	2.17E+04^*** 5.85E+03	7.80E+03^*** 1.82E+03
B3	—	1.69E+05^** 5.49E+04	5.79E+07^*** 1.28E+07	4.78E+04^*** 9.75E+03	6.17E−01^*** 3.10E+03	1.29E+04^*** 2.56E+03
B4	—	—	7.06E+07^*** 1.16E+07	3.98E+03 1.08E+04	—	1.67E+04^*** 1.88E+03
B5	—	—	—	1.08E+05^** 3.93E+04	—
B6	—	—	—	2.38E+04 1.72E+04	—	—
Adjusted R ²	.94	.91	.98	.96	.93	.99
Observed	130	130	130	130	130	130
Correlation matrix of the residuals
1	1.00	0.20	0.03	0.37	−0.17	0.03
2		1.00	0.81	0.46	0.48	0.09
3			1.00	0.41	0.53	0.05
4				1.00	0.09	−0.15
5					1.00	0.17
6						1.00

Note: (1) = administration, (2) = public safety, (3) = education, (4) = health, (5) = transportation, and (6) = public economy.

^aTime dummies for the years 2001–04 are not shown.

*Significant at the 90 percent confidence level.

** Significant at the 95 percent confidence level.

***Significant at the 99 percent confidence level.

With one exception, the output indicators are positively related to cost. The negative sign of $β_{1}^{2}$ in Table 3 (number of delinquencies) could be the result of systematic measurement error. Some cantons report the many petty of cases (which cause little expense), while others limit their reporting to the relatively few major offenses (which cost a lot). These differences may induce a negative partial correlation between expenditure on public safety and the number of delinquencies. Furthermore, except for the two coefficients $β_{4}^{4}$ (rehab clinics) and $β_{6}^{4}$ (retirement homes), all indicators are significant at the 95 percent confidence level. Both indicators are nevertheless retained for reasons of completeness. Finally, the adjusted $R^{2}$ reaches at least 90 percent, confirming the combined relevance of the selected indicators.

The Output Performance Index

Including all twenty-two output indicators is still not possible in an annual DEA with twenty-six cantons. This is because DEA necessarily identifies at least one canton as efficient with respect to the twenty-two output–cost combinations. Thus, at least twenty-two of the twenty-six cantons would be identified as efficient, reducing the power of the analysis. One possibility is to aggregate the retained twenty-two output indicators $Y_{k, c t}^{j}$ to a performance index $Ψ_{c t}^{j}$ for each service category ( $j = 1, \dots, 6$ ), resulting in

Ψ_{c t}^{j} = \sum_{k = 1}^{K_{j}} Y_{k, c t}^{j} \cdot p_{k}^{j} .

In this calculation, the problematic output indicator delinquencies is subjected to a linear monotone transformation such that

\hat{Y_{i}} = - Y_{i} + r \geq 0

(Seiford and Zhu 2002). As to the weighting parameters

p_{k}^{j}

pertaining to the two to six output indicators per service category, there are two alternatives. One is to use the estimated

β_{k}^{j}

from equation (5). Here, all cantons get the same weighting parameters

p_{k}^{j}

. A more flexible variant is to use the canton-specific shadow prices

υ_{c}

from a DEA applied to each of the six service categories. These shadow prices reflect the marginal cost of expanding a particular service by one unit produced by an efficient DMU. For the inefficient DMUs, a radial projection onto the efficiency frontier permits to determine the pertinent shadow prices. In this variant, no canton is discriminated because of an inappropriate weighting parameter,

p_{k}^{j}

, as would be the case in the first alternative. For each canton, the aggregation is based on those cantonal specific weighting parameters

υ_{c}

, which maximize their output performance index. For reasons of consistency, this alternative is retained for our analysis. The results of this calculation (with

Ψ_{c}^{j}

values for 2004) are displayed in Table 4 .

Table 4.

Output Performance Indicators Ψij, Twenty-Six Cantons (2004)

Cantons	(1)	(2)	(3)	(4)	(5)	(6)
ZH	804,351	609,182	143,830,9	241,433,8	293,337	182,691
BE	608,731	467,307	101,858,9	227,987,2	270,257	535,600
LU	226,021	160,632	534,219	490,410	84,259	187,310
UR	22,354	16,503	49,531	39,184	12,713	26,513
SZ	87,616	61,522	175,165	128,367	370,09	73,474
OW	21,130	17,194	46,818	33,907	11,161	34,362
NW	28,856	17,994	48,612	35,418	10,613	20,764
GL	25,754	19,353	51,962	61,451	10,999	29,547
ZG	128,873	47,613	134,137	132,750	27,843	31,162
FR	160,165	113,271	363,600	343,419	75,380	180,878
SO	157,621	112,625	312,784	269,547	65,085	85,709
BS	123,028	104,393	232,726	634,082	31,674	1,302
BL	169,043	120,690	383,400	317,776	62,031	58,940
SH	47,015	34,727	96,062	113,555	31,302	36,222
AR	33,668	24,723	74,755	99,397	17,765	43,370
AI	10,082	7,635	29,755	15,113	4,366	24,714
SG	292,344	208,803	620,278	668,789	108,448	205,445
GR	120,802	130,689	221,249	300,494	97,956	306,163
AG	360,075	255,663	858,094	691,775	145,986	146,288
TG	148,445	105,537	332,003	360,721	71,424	117,408
TI	220,732	185,993	420,807	654,548	100,196	84,566
VD	413,159	327,087	986,906	141,136,0	188,657	274,138
VS	183,488	174242	399,383	459,095	134,423	138,802
NE	106,986	80,628	270,524	284,766	44,079	91,707
GE	288,564	204,038	635,494	150,813,8	106,019	24,811
JU	44,022	31,404	95,142	121,188	29,632	106,580

Note: (1) = administration, (2) = public safety, (3) = education, (4) = health, (5) = transportation, and (6) = public economy. AG = Argovia, AI = Appenzell Inner-Rhodes, AR = Appenzell Outer-Rhodes, BE = Bern, BL = Basel-Country, BS = Basel-City, FR = Fribourg, GE = Geneva, GL = Glarus, GR = Grisons, JU = Jura, LU = Lucerne, NE = Neuchatel, NW = Nidwalden, OW = Obwalden, SG = St. Gall, SH = Schafhausen, SO = Solothurn, SZ = Schwyz, TG = Thurgovia, TI = Ticino, UR = Uri, VD = Vaud, VS = Valais, ZG = Zug, and ZH = Zurich.

While the numbers are difficult to interpret in general, the entries for administration (column 1) reflect size of the cantonal population served because the two output indicators are population and number of firms.

Input Variables

The inputs are measured as real expenditure (CHF of 2000) on the six service categories. This is a widespread practice (Afonso, Schuknecht, and Tanzi 2006; De Borger and Kerstens 1996). For the categories transportation and health, only operating expenses are included (total expenditure minus investments in new infrastructure) because annual investments contain a strong transitory component.

Determinants of DEA Efficiency

Recall the two research questions:

1. Does a fiscal equalization program as sizable as the Switzerland contain incentives to provide public goods less efficiently, creating a trade-off between equity and efficiency?

2. Does it matter whether transfer payments are earmarked or not?

The first question is investigated using three models. Model A relates DEA efficiency scores to the financial potential, which determines the amount of fiscal equalization between cantons. Model B checks whether this influence depends only on the size of the financial flows, regardless of their direction. In model C, fiscal equalization paid and received is allowed to have an asymmetric impact on efficiency. The explanatory variables are defined as follows (endogeneity issues are addressed in estimation of the determinants of DEA efficiency section).

Index of financial potential (F.POT)

The Swiss fiscal equalization program is based on this indicator, with higher value implying less federal financial assistance. It is used in model A.

Index of financial equalization (F.EQ)

F.EQ is a modification of F.POT. It measures the absolute value of the deviation from the value α at which no aid is contributed or received; formally, $F . E Q = a b s (F . P O T - α)$ . Note that α differs from the mean value of $F . P O T$ . The higher this index value, the larger is the amount of fiscal equalization. It is used in model B.

Dummies for paying and receiving cantons (F.GIV = 1, F.REC = 1)

F.GIV equals 1 for cantons who are payers, while F.REC equals 1 for those who are recipients. Cantons that are neither recipients nor payers constitute the benchmark group in both cases. These variables appear in model C.

The second research question calls for the introduction of the following variables.

Subsidies per capita (SUBS)

This variable measures earmarked payments, which are suspected to induce a particularly high degree of inefficiency (see Swiss Federalism section again).

In addition, the following variables serve to control for other influences on cantonal efficiency scores that cannot be controlled for in the DEA but could influence efficiency scores.

Direct democracy (DIR.DEM)

The degree of direct democratic control (popular initiatives and mandatory referenda on expensive public projects) was already found to be relevant by Pommerehne and Zweifel (1991) in the context of tax evasion. More recently, Fischer (2004) and Feld and Matsusaka (2003) found the amount of public services provided to be negatively related to an index of democratic control developed by Stutzer (1999). This index is used here as well, with the expectation of a positive relationship with efficiency.

Decentralization (DEC)

Decentralized provision of public services within a canton has an ambiguous effect on efficiency. On one hand, it might cause a lack of human and technical resources in small cantons, resulting in higher cost of administration (see, e.g., Smith 1985). On the other hand, Tiebout (1956) argues that decentralization facilitates competition, which fosters efficiency. In this work, DEC is the share of cantonal expenditure that is transferred to the communities.

Income per capita (INCOME)

This is a component of F.POT that according to De Borger and Kerstens (1996) has additional information content. They predict that efficiency of local government decreases with increasing income per capita because citizens in high-wage jurisdictions face high opportunity costs when trying to monitor the efficiency of public good provision.

Tax burden (TAX)

This component of F.POT has additional information content as well. In line with Tiebout (1956), a canton’s efficiency awareness is predicted to increase with a stronger participation in tax competition. Since a low value of TAX indicates a strong engagement in tax competition, it is hypothesized to go along with a high degree of efficiency, ceteris paribus.

Disparities (TOPOGR, I.STRUCT, and POP.STRUCT)

These variables reflect exogenously given disparities, which are expected to cause higher cost and hence lower efficiency in the provision of public services. They enter the new fiscal equalization formula. TOPOGR adjusts for geographic differences, while I.STRUCT controls for difference of community size, the employment rate, and population density. POP.STRUCT denotes the shares of immigrants and citizens older than 80 years, with equal weights.

Cost of housing (P.HOUS)

The cost of housing differs substantially between cantons. It is an important component of the cost of living, which is adjusted for in the wages of public employees and hence influences the cost of providing public services.

Culture (CULT.F = 1)

The French- and German-speaking parts of Switzerland differ in many ways, possibly also in terms of efficiency Fischer (2004). Thus, CULT.F = 1 if the canton is predominantly French-speaking.

Year of observation

$(Y \under 2001 = 1, Y \under 2002 = 1, Y \under 2003 = 1, a n d Y \under 2004 = 1) .$ This set of dummy variables indicates the year of observation (base year is 2000).

Empirical Results

This section first discusses the robust DEA efficiency scores. The assumption (to be relaxed in the following) is that the twenty-six cantons belong to the same universe, meaning that all cantons face the same circumstances in their provision of public goods. In a second step, efficiency scores are related to fiscal equalization and other socioeconomic factors of interest.

DEA

With the six output indicators derived from equation (6) and expenditures changing from year to year, an annual DEA for the years 2000–04 can be performed. Table 5 shows the results for the year 2004. The robust efficiency scores are calculated under the assumption of constant returns to scale, indicating potential cost improvements achievable by a radial movement to a technically and scale-efficient reference point on the frontier. There are two super-efficient cases that are assigned a score of 1.00 (see Measuring Technical Efficiency with DEA section again).

Table 5.

DEA Efficiency Scores, Twenty-Six Swiss Cantons (2004)

Cantons	Rank	(1–6)^a	(1)	(2)	(3)	(4)	(5)	(6)	SD
ZH	25	0.74	0.71	0.42	0.67	0.76	0.53	0.84	.15
BE	11	0.88	0.87	0.70	0.76	0.81	0.71	0.88	.08
LU	20	0.82	0.64	0.74	0.91	0.63	0.67	0.79	.11
UR	22	0.82	0.71	0.65	0.82	0.76	0.69	0.73	.06
SZ	15	0.86	[1.00]	0.75	0.81	0.56	0.71	0.75	.14
OW	5	0.95	0.78	0.97	0.93	0.72	0.88	0.78	.10
NW	10	0.89	[1.00]	0.79	0.78	0.64	0.78	0.77	.12
GL	6	0.95	0.81	0.91	0.85	0.97	0.74	0.77	.09
ZG	16	0.85	0.95	0.51	0.69	0.81	0.69	0.92	.16
FR	14	0.86	0.79	0.74	0.93	0.69	0.67	0.77	.09
SO	12	0.88	0.78	0.73	0.81	0.61	0.81	0.96	.11
BS	26	0.64	0.84	0.42	0.78	0.72	0.43	0.21	.25
BL	18	0.85	0.68	0.70	0.84	0.69	0.62	[1.00]	.14
SH	13	0.88	0.71	0.59	0.84	0.73	0.99	0.82	.13
AR	8	0.93	0.64	0.91	0.86	0.93	0.61	0.98	.16
AI	2	0.97	0.92	0.86	0.98	0.83	0.61	0.97	.14
SG	19	0.84	0.89	0.65	0.80	0.74	0.49	0.89	.16
GR	17	0.85	0.61	0.90	0.88	0.63	0.52	[1.00]	.19
AG	3	0.95	0.82	0.82	[1.00]	0.59	0.95	0.91	.14
TG	[1]	[1.00]	0.84	0.66	0.85	[1.00]	[1.00]	0.99	.14
TI	7	0.95	0.62	0.79	[1.00]	0.85	0.78	[1.00]	.14
VD	21	0.82	0.67	0.60	[1.00]	0.55	0.71	0.84	.16
VS	4	0.95	0.82	[1.00]	0.96	0.76	0.77	0.78	.10
NE	23	0.77	0.64	0.55	0.98	0.39	0.53	[1.00]	.25
GE	24	0.75	0.42	0.38	0.91	[1.00]	0.65	0.64	.25
JU	9	0.89	0.70	0.81	0.88	0.45	0.93	[1.00]	.20
Number of efficiency		1	2	1	3	2	1	5
Outliers		0	0	0	0	TG	0	TI
Mean		0.87	0.76	0.71	0.87	0.72	0.71	0.85
Min		0.64	0.42	0.38	0.67	0.39	0.43	0.21
SD		.08	.14	.17	.09	.16	.15	.17

Note: (1) = administration, (2) = public safety, (3) = education, (4) = health, (5) = transportation, and (6) = public economy.

AG = Argovia, AI = Appenzell Inner-Rhodes, AR = Appenzell Outer-Rhodes, BE = Bern, BL = Basel-Country, BS = Basel-City, FR = Fribourg, GE = Geneva, GL = Glarus, GR = Grisons, JU = Jura, LU = Lucerne, NE = Neuchatel, NW = Nidwalden, OW = Obwalden, SG = St. Gall, SH = Schafhausen, SO = Solothurn, SZ = Schwyz, TG = Thurgovia, TI = Ticino, UR = Uri, VD = Vaud, VS = Valais, ZG = Zug, and ZH = Zurich.

^aMean of the six categories, normalized by the maximum value.

Starting with the overall scores, the rural canton of Thurgovia (TG) attains 100 percent technical efficiency (score of 1.00). Two more cantons (again rural) come close, namely, Appenzell Inner-Rhodes (AI; 0.97) and Argovia (AG; 0.97). Indeed, 30 percent of all cantons have a performance score higher than 0.90. At the other extreme, BS is identified as the most inefficient canton (0.64). Thus, its expenditure could have been lowered by 37 percent while still maintaining the same output level. Other urban cantons, namely, ZH, 0.74, and GE, 0.75, already perform much better. However, differences between rural and urban cantons are not surprising. The well-known disparities caused by higher population densities and more complex industry structures, which by the way are taken into account in the fiscal equalization program, cannot be incorporated in DEA. But the second step analysis adjusts for it with three variables from the new fiscal equalization program to enable unbiased estimates of the hypotheses.

The question arises of whether the year 2004 is representative of the observation period 2000–04. Figure 2 provides an answer, ranking cantons according to their five-year median values along with their estimated quartile ranges and 95 percent confidence bands.

The findings of Table 5 are confirmed in that TG remains leader while BS consistently is last. One reason for volatility over time could be investment in infrastructure. For example, ZG shows an improvement from rank 19 in 2000 to 15 in 2004 but drops to place 23 in 2003 because of spending heavily on investment without charging projects to the capital account. Yet, comparable Glarus (GL) with a similar degree of volatility in performance achieved a consistent improvement over the five years, from 0.84 (rank 18) to 0.94 (rank 6). In sum, variations over time are too limited and unsystematic to undermine the robustness of the overall ranking.

Another question of interest is whether the leader TG is the champion in all six categories of public service distinguished. If this were the case, Tiebout competition would unfold with full vigor. However, Table 5 shows that TG has a low efficiency score in public safety (0.66). Conversely, last-ranked BS does attain an average value in administration (0.84), permitting cantonal government to cater to voters especially interested in administrative services. Moreover, low overall scores do not necessarily go along with high standard errors across the six categories (see last column of Table 5). Bottom-ranked BS has a high standard error of 0.25, while Uri (UR) with rank 23 has one of only 0.06. Thus, small and homogenous UR can survive Tiebout competition since neighboring (more urban) Lucerne (LU) has twice as much variation (SD = 0.11), while its rank is almost the same. In sum, Tiebout competition is limited even in a country as markedly federalist as Switzerland.

In a federal state, another major issue is centralization versus decentralization. In the case of Switzerland, the debate has been focusing on education (see Barankay and Lockwood 2007). Lack of coordination between the cantons has been cited as a reason for the rather mediocre performance of the Swiss educational system in the Programme for International Student Assessment (PISA) study. However, these criticisms might be overstated. The average performance score for education (3) is 0.87 (SD = 0.09). This beats the score of 0.76 (SD = 0.14) for public administration (1), which is generally believed to perform well in international comparison.

Estimation of the Determinants of DEA Efficiency

Next, it is of interest to see whether fiscal equalization has an influence on the efficiency scores of the twenty-six cantons over the years 2000–04. In total, 130 observations (26 × 5) are available for estimating equation (4) of Measuring Technical Efficiency with DEA section. Disparities in the provision of public goods are reflected by the indicators discussed in Determinants of DEA Efficiency section.

Estimation results for the three models outlined in the Determinants of DEA Efficiency section are displayed in Table 6 , after performing tests for endogeneity, heteroscedasticity, and nonlinearity. Fiscal equalization could be endogenous to efficiency because highly efficient jurisdictions are made to contribute to the program. However, a Hausman test does not suggest rejection of the exogeneity assumption. This is not really surprising because the Swiss fiscal equalization is not adjusted every year, possibly making an observation period of five years too short for detecting reverse causality. Heteroscedasticity is not a problem either according to a Breusch–Pagan test. Finally, linearity need not be rejected with the exception of SUBS2, TAX2, DEC2, and POP.STRUCT2. Earmarked payments as well as tax burden, decentralization, and population structure have a nonlinear influence on the performance of the cantons. Several interaction terms proved significant, too; their inclusion does not markedly affect parameter estimates, however. Estimation results turn out to be robust for the three models. Most of the variables have expected signs and are significant at the 90 percent confidence level or better.

Table 6.

Tobit Estimates of Efficiency Scores

	Model A^a		Model B^a		Model C^a
Variables	Coef.	Elasticity	Coef.	Elasticity	Coef.	Elasticity
F.POT	−5.8E−03^***	−7.0E−01
F.EQ			−7.6E−04^***	−3.8E−02
F.GIV					−2.1E−01^*	−0.0836
F.REC					−2.8E−02	−0.0238
SUBS	−1.6E−04^***	−2.7E−01	−1.8^***	−3.0E−01	−1.3E−04^***	−0.2192
DIR.DEM	−3.0E−01^***	−1.7E+00	−7.1E−02^***	−3.9E−01	−1.4E−01^***	−0.7881
DEC	2.0E+00^**	1.1E+00	3.3E+00^***	1.7E+00	2.9E+00^**	1.5045
INCOME	−2.8E−04	−3.7E−02
TAX	−5.6E−03^***	−7.9E−01	−6.4E−03^***	−9.0E−01	−7.7E−03^***	−1.0818
TOPOGR	−1.4E−04^**	−2.4E−02	−6.2E−06	−1.1E−03	−5.7E−05	−0.0102
I.STRUCT	−3.3E−01^***	−7.2E−01	−2.3E−01^***	−5.0E−01	−2.3E−01^***	−0.4842
POP.STRUCT	−3.1E−01^***	−5.3E−01	−2.7E−01^***	−4.8E−01	−2.6E−01^***	−0.4603
P.HOUS	−6.3E−01^***	−8.5E−01	−6.2E−01^***	−8.4E−01	−6.5E−01^***	−0.8874
CULT.F	−6.0E−02^***	−1.8E−02	−4.1E−02^***	−1.2E−02	−4.0E−02^***	−0.0118
SUBS2	2.3E−08^***	8.8E−02	1.5E−08^***	5.6E−02	1.3E−08^***	0.0496
TAX2	3.1E−05^***	4.8E−01	3.2E−05^***	4.9E−01	3.9E−05^***	0.5957
POP.STRUCT2	7.9E−02^***	2.7E−01	6.3E−02^***	2.1E−01	5.8E−02^***	0.1987
DEC2	−5.0E+00^***	−1.1E+00	−5.0E+00^***	−1.1E+00	−5.4E+00^***	−1.1569
SUBS:F.POT	1.2E−07	1.4E−02
SUBS:F.EQ			1.1E−06^***	7.7E−02
SUBS:F.GIV					2.1E−05	0.0059
SUBS:F.REC					2.3E−05	0.0302
F.POT:DIR.DEM	1.3E−03^***	6.4E−01
F.GIV:DIR.DEM					4.0E−02^**	0.0690
F.REC:DIR.DEM					2.8E−05	0.0001
DIR.DEM:DEC	4.2E−01^***	8.9E−01	1.2E−01^***	2.6E−01	2.9E−01^***	0.6218
I.STRUCT:P.HOUS	2.9E−01^***	6.5E−01	2.0E−01^***	4.5E−01	2.0E−01^***	0.4546
Observation	114		114		114
Log likelihood	384.1		374.9		372.2

^aTime dummies for the years 2001–04 are not shown.

*Significant at the 90 percent confidence level.

**Significant at the 95 percent confidence level.

***Significant at the 99 percent confidence level.

In model A, a negative sign is obtained for F.POT. Use of index of financial potential that determines fiscal equalization payments therefore seems to lower efficiency systematically (elasticity of −0.7) after controlling for exogenously given disparities and other variables affecting the cost of public good provision. Thus, cantons with high financial potential may have an incentive to underperform. In model B, the absolute value of payments enters with F.EQ. Not surprisingly, F.EQ has a significantly negative sign too, suggesting that fiscal equalization as such lowers technical efficiency in the provision of public goods. Finally, model C indicates that paying cantons (elasticity of −0.084) are more influenced than receiving cantons (elasticity of −0.024) with regard to efficiency.

In sum, the evidence of Table 6 provides an answer to question (1) of Swiss Federalism section by supporting the notion that fiscal equalization undermines cantonal efficiency in Switzerland for both receivers and payers, but even more for payers, who are the cantons with high financial potential. This difference is intuitive because payers have more reason to respond to fiscal equalization with inefficiency than receivers. Expecting no benefit from redistribution, they rather waste their money than to give it to financially disadvantaged cantons. Thus, any public good with a positive net benefit is provided, whereas only those with above-average net benefits contribute to the canton’s technical efficiency. Being financially constrained, receiving cantons want to ensure that their most productive projects are financed; they extend this list only in order to justify their need for redistribution. While estimated elasticities are below one throughout, fiscal equalization in the case of Switzerland does give rise to the equity–efficiency trade-off described by Stiglitz (1988).

Question 2 of Swiss Federalism section asks whether earmarked federal subsidies have an especially strong (negative) effect on cantonal efficiency. Whereas general payments can be used by the canton where it believes to generate the highest benefit for its citizens, earmarked subsidies may result in gold plating of projects and hence inefficiency. Indeed, Table 6 shows SUBS to have a negative sign in all three models with estimated elasticities between −0.2 and −0.3. Therefore, subsidies may encourage inefficiency, as claimed in the Swiss case by Frey et al. (1994). Therefore, the new equalization formula of 2008, which minimizes earmarked payments has the potential to reduce technical inefficiency in the provision of public goods compared to its predecessor.

Some of the other explanatory variables are of interest as well. Foremost, DIR.DEM and DEC, which capture two unique features of Switzerland, contradict theoretical expectations. The negative sign of DIR.DEM suggests that direct democratic control lowers rather than increases technical efficiency. This seems to contradict the findings of Fischer (2004) as well as Feld and Matsusaka (2003), who however studied the amount of public services provided rather than technical efficiency. Still, lower amounts can go along with lower efficiency if direct democracy should mainly delay (notably through referenda) planning that is “on target” in terms of efficiency. On the other hand, decentralization has the expected effect in that the coefficient of DEC is positive throughout, confirming Barankay and Lockwood (2006) who examined the impact of decentralization on productive efficiency in public education. The negative effects emphasized by Smith (1985) apparently are more than compensated by the positive ones due to Tiebout competition, which however are subject to diminishing marginal returns (see the negative coefficient of DEC2).

In addition, TAX shows the expected negative sign, suggesting that cantons with a low tax burden exhibit higher performance, a state of affairs conducive to strong Tiebout competition. However, the positive sign of the quadratic term points to a rapidly diminishing effect as soon as the tax burden starts to increase, with the critical value of 90.32 in model A and 100 in model B, respectively (the average tax burden of Switzerland is set to 100). The positive sign of TAX found by De Borger and Kerstens (1996) therefore also holds for Switzerland as soon as it exceeds the average. Thus, both extremely low and high tax burdens cause efficiency gains because of tax competition on one hand and because of increasing monitoring by citizens on the other.

Finally, it is of interest for policy to know whether the determinants entering the new fiscal equalization formula (TOPOGR, I.STRUCT, and POP.STRUCT) to adjust for resource disparities are relevant or not. The three variables are negatively related to DEA efficiency scores, regardless of model specification. Therefore, the 2008 reform is likely to achieve its objective because it introduces exogenous factors in the equalization formula that seem to have a significant influence on the heterogeneity of public good provision. Finally, the negative coefficient of P.HOUS shows that the cost of housing factors into the cost of public services and hence inefficiency. Since it is largely exogenous, it could also be included in the fiscal equalization formula.

Concluding Remarks

The purpose of this article was to measure efficiency in the provision of public services applying DEA, which maximizes the distance between an output bundle and an input bundle. The country analyzed is Switzerland, which is characterized by a high degree of federalism permitting Tiebout competition on one hand and a sizable fiscal equalization program on the other. DEA shadow prices serve to derive the weights for aggregating the six public service categories into an overall output indicator for the twenty-six cantons, while inputs are equated to their real expenditure over the years 2000–04. In a second step, DEA efficiency scores are related to the indicator “financial potential” that governs the Swiss fiscal equalization scheme designed to alleviate disparities between cantons.

The main results are the following. First, efficiency scores indicate better performance of small rural cantons than of urban ones and are robust over the five years investigated. A comparison over the six service categories further shows that cantons with a high overall performance do not automatically outperform in all of them, preventing any one of them from becoming dominant in Tiebout competition. Second, financial equalization is negatively related to cantonal efficiency, with an especially marked effect on payers. Schemes designed to mitigate disparities that are deemed unacceptable not only by politicians but also by the citizenry (the pertinent constitutional amendment survived a popular referendum in the case of Switzerland) may thus have the undesirable side effect of undermining incentives for efficiency. Jurisdictions who are payers and receivers both seek to keep their “financial potential” low—the former because this serves to ease their burden, the latter because they expect to receive more transfer payments and subsidies notably by producing public services at higher than minimum cost. Therefore, the equity–efficiency trade-off noted by Stiglitz (1988) seems indeed to exist in the case of Switzerland. Third, earmarked federal subsidies (the main component of transfer payments prior to the 2008 reform) are negatively related to cantonal efficiency as well. Substitution of these earmarked payments by freely disposable lump-sum ones as part of the new equalization program implemented in 2008 is therefore likely to enhance cantonal performance. This analysis suffers from several limitations. Above all, DEA efficiency scores constitute a technocratic measure, being silent on the question of whether the services provided reflect the preferences of citizens. Also, some of the explanatory variables used to predict efficiency scores may not be fully exogenous in the long term. In particular, INCOME possibly not only influences efficiency as a taste variable but could be the consequence of cantonal efficiency as well. In spite of these limitations, the analysis not only identifies the equity–efficiency trade-off that federally organized countries (such as Switzerland) face when implementing a fiscal equalization scheme but also provides guidance on how to structure it in terms of earmarked and freely disposable payments.

Footnotes

Acknowledgments

The authors gratefully acknowledge the comments and suggestions from the editor and an anonymous referee, and research support provided by Johannes Schoder and Boris Krey.

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Afonso

Antonio

Fernandes

Sónia P.

(2006). “Measuring Local Government Spending Efficiency: Evidence for the Lisbon Region.” Regional Studies, 40, 39–53.

Afonso

Antonio

Scaglioni

Carla.

2005. “Public Services Efficiency Provision in Italian Regions: A Non-Parametric Analysis.” ISEG working papers 02, Technical University of Lisbon.

Afonso

Antonio

Schuknecht

Ludger

Tanzi

Vito.

2006. “Public Sector Efficiency: Evidence for new EU Member States and Emerging Markets.” ISEG working papers 01, Technical University of Lisbon.

Andersen

Per

Petersen

Niels C.

(1993). “A Procedure for Ranking Efficient Units in Data Envelopment Analysis.” Management Science, 39, 1261–4.

Barankay

Iwan

Lockwood

Ben.

(2007). “Decentralization and the Productive Efficiency of Government: Evidence from Swiss Cantons.” Journal of Public Economics, 91, 1197–218.

Coelli

Timothy J.

Rao

D. S. Prasada

O’Donnel

Christopher J.

Battese

George E.

(2005). An Introduction to Efficiency and Productivity Analysis. New York, NY: Springer.

De Borger

Bruno

Kerstens

Kristiaan.

(1996). “Cost Efficiency of Belgian Local Governments: A Comparative Analysis of FDH, DEA, and Econometric Approaches.” Regional Science and Urban Economics, 26, 145–70.

Drake

Leigh

Simper

Richard.

(2003). “The Measurement of English and Welsh Police Force Efficiency: A Comparison of Distance Function Models.” European Journal of Operational Research, 147, 165–86.

Farrell

(1957). “The Measurement of Productive Efficiency.” Journal of the Royal Statistical Society, 120, 253–81.

10.

Feld

Lars P.

Matsusaka

John G.

(2003). “Budget Referendums and Government Spending: Evidence from Swiss Cantons.” Journal of Public Economics, 87, 2703–24.

11.

Fischer

Roland.

2004. “Die Unterschiede in Der Steuerbelastung Der Kantone: Eine Analyse Auf Der Basis Eines Mikroökonomischen Haushaltsmodells [Differences in cantonal tax burden: An analysis based on a macroeconomic budgeting model].” Fiscal Authority Department of Switzerland Working Papers 06, Switzerland.

12.

Fischer

Roland

Tobias

Beljean

Fivaz

Jan.

(2003). “Mehr Chancengleichheit und Eigenverantwortung: Der neue Schweizer Finanzausgleich zwischen Bund und Kantonen [More Equity and Cantonal Authority: The New Swiss Financial Equalization Program].” Vierteljahrshefte zur Wirtschaftsforschung, 72, 407–22.

13.

Frey

R. L

Dafflon

Jeanrenaud

Meier

1994. “Der Finanzausgleich zwischen Bund und Kantonen [Financial equalization between the federal state and the cantons].” Fiscal Authority Department, Switzerland.

14.

Grossman

Philip J.

Mavros

Panayiotis

Wassmer

Robert W.

(1999). “Public Sector Technical Inefficiency in Large U.S. Cities.” Journal of Urban Economics, 46, 278–99.

15.

Koopmans

Tjalling C.

(1951). “An Analysis of Production as an Efficient Combination of Activities.” In Koopmans

Tjalling C.

(Ed.), Activity Analysis of Production and Allocation (pp. 33–97). New York: John Wiley.

16.

Olson

Mancur.

(1969). “The Principle of Fiscal Equivalence: The Division of Responsibilities Among Different Levels of Government.” The American Economic Review, 59, 479–87.

17.

Pommerehne

Werner

Zweifel

Peter.

(1991). “Success of a Tax Amnesty: At the Polls, for the FISC?” Public Choice, 72, 131–65.

18.

Seiford

Lawrence

Zhu

Joe.

(2002). “Modeling Undesirable Factors in Efficiency Evaluation.” European Journal of Operational Research, 142, 16–20.

19.

Shephard

Ronald W.

(1970). Theory of Cost and Production Functions. Princeton, NJ: Princeton University Press.

20.

Smith

Brian C.

(1985). Decentralization: The Territorial Dimension of State. London: Allen and Unwin.

21.

Stiglitz

Joseph E.

(1988). Economics of the Public Sector. New York, NY: Norton.

22.

Stutzer

Alois.

1999. “A Democracy Index for the Swiss Cantons.” Institute for Empirical Research in Economics Working Papers 23, University of Zurich.

23.

Thanassoulis

Emmanuel.

(1999). “Setting Achievement Targets for School Children.” Education Economics, 7, 101–19.

24.

Thöny

Bernhard.

2005. Der Finanzausgleich im Wettbewerbsföderalismus – Eine Theorie des Finanzausgleichs und eine Analyse der innerkantonalen Finanzausgleichssysteme in der Schweiz [Financial equalization in federalism—A theory of the equalization program and an analysis of the inner cantonal financial equalization systems of Switzerland]. Bamberg, Germany: Difo-Druck Gmbh.

25.

Tiebout

Charles M.

(1956). “A Pure Theory of Local Expenditures.” Journal of Political Economy, 64, 416–24.

26.

Vanden Eeckaut

Philippe

Tulkens

Henry

Jamar

Marie-Astrid.

(1993). “Cost Efficiency in Belgian Municipalities.” In Fried

Harold O.

Lovell

C. A. Knox

Schmidt

Shelton S.

(Eds.), The Measurement of Productive Efficiency: Techniques and Applications (pp. 300–34). New York: Oxford University Press.

27.

Worthington

Andrew.

(2001). “An Empirical Survey of Frontier Efficiency Measurement Techniques in Education.” Education Economics, 9, 245–68.

28.

Worthington

Andrew

Dollery

Brian E.

(2001). “Measuring Efficiency in Local Government: An Analysis of New South Wales Municipalities’ Domestic Waste Management Function.” Policy Studies Journal, 29, 232–50.