The effect of PPP capital on economic growth

(1)

The Effect of PPP Capital on Economic Growth

By Tomás Malafaia

Student ID 11084685 July 2016

In Partial Fulfillment of the Requirements for the Degree of Master of Science in Economics Master of Science in Economics – Specialization: Monetary Policy, Banking and Regulation

Supervisor: Dr. W.E. Romp Co-reader: Dr. C.A. Stoltenberg

Faculty of Economics and Business Department of Macro & International Economics

Abstract

This thesis investigates the macroeconomic effects of the stock of Capital resulting from Public-Private Partnerships on the economic growth of European countries. Using a dataset consisting of 26 countries and 22 years (1990-2011), the Production Function approach (classic Cobb-Douglas augmented with PPP Capital) to the investigation of the effects of Public Capital on economic growth is implemented. Estimations performed by both OLS and Blundell and Bond’s GMM estimator point to a negative but statistically insignificant effect of PPP Capital on output, thus making the case that the recent (and predicted to continue in the future) expansion of this kind of investment will not hamper or enhance economic growth.

(2)

2

Statement of Originality

This document is written by Student [fill out your Given name and your Surname] who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Introduction

Public-Private Partnerships (PPP’s) date back to the 1980’s in the United Kingdom, originally under the Private-Finance Initiative (PFI) structure, which included the private partner almost solely on a financing capacity that provided the government with a way to perform infrastructure investment while evading expenditure controls (Spackman, 2002). Since then, however, PPP’s have evolved to an “undertaking which involves a sizable initial investment in a certain facility (a road, a bridge, an airport, a prison), and then the delivery of the services from this facility”, as defined by Sadka (2006). A key characteristic of these facilities, and the reason why the option of its complete privatization is usually dismissed, is that they are public goods - which implies market failures that the continuing involvement of the public partner is meant to solve.

Along with the mentioned evolution of its financial and operational structures, PPP’s started spreading throughout Europe in the 1990’s and well into the 2000’s. Chart 1 below shows the growth of European PPP projects in both Value (left axis, in million EUR) and Number (right axis). It clearly demonstrates the increasing popularity of this kind of investment, mainly fueled by big projects in Portugal, Spain and France. Nevertheless, its usage has decreased in the years after the financial crisis and its well-known consequences on Public Finances.

Chart 1 (Data source: EPEC)

Constraints such as those imposed by the Fiscal Pact create a situation where the choice between PPP investment and traditional public sector procurement is actually closer to that between PPP and no investment whatsoever (Jasiukevicius and Vasiliauskaite,

0 5.000 10.000 15.000 20.000 25.000 30.000 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 0 50 100 150

Number and Value of PPP Projects in Europe

(5)

5

2013). This, given the current state of most European countries’ Debt to GDP ratios and public deficits, is a strong reason to believe that this relatively new way of financing the creation of Public Capital may continue to be a growingly common occurrence. Therefore it becomes relevant to analyze the existing data and understand whether or not there is reason to consider that PPP’s have a positive, negative or inexistent effect on economic growth, which the present work intends to do.

The next section surveys the existent literature on PPP’s, particularly focusing on the work relating them to economic growth and the ways in which the effects of Public Capital on the latter can be measured. Section 3 lays out the methodology selected for this particular analysis. Section 4 focuses on the sources of data for each variable and mentions the number of countries and years in the sample. Section 5 describes the Empirical Approach utilized and how it is meant to tackle some of the more important econometrical issues identified. In Section 6, the estimation results are presented and critically assessed. Lastly, Section 7 summarizes and concludes.

Literature Review

Dating back to the 1990’s a considerable amount of literature devoted to explaining the political and economic motivations for Public-Private Partnerships has been produced. This extends to various aspects associated with PPP’s such as risk allocation between parties (Grimsey and Lewis, 2002), optimal cost sharing (Takashima et al., 2010), the transfer of value to the private sector through government guarantees (Alonso-Conde et al., 2007), efficient capital structures (Moszoro, 2010) and more.

Valila (2005) compares PPP’s to traditional public investment, concentrating on the microeconomic causes for higher benefits and costs and trying to identify “the conditions under which public-private partnerships may be the optimal form of public sector intervention”. Under this analysis, the public involvement is considered to be beneficial to the extent to which it solves market failures (such as natural monopolies, externalities and public goods) and provides allocative efficiency, while the inclusion of the private partner is said to bolster the productive efficiency which has been found to

(6)

6

be lacking in an important amount of public investment projects1_{. This gain in efficiency,} however, does not mean ex-ante that economic growth will be higher, as the savings freed in this way can be applied to unproductive investment or spending.

Another subject mentioned in the same article, but also in several others such as Sadka (2006) and Spackman (2002), is the fact that PPP’s can often be equivalent to off-Balance Sheet borrowing for governments. This can be seen through two different scopes: on the one hand, budget restrictions caused by either limited access to financial markets or artificial rules (p.ex. the Stability and Growth Pact), create situations where PPP’s are the only option for governments to provide economies with structural investment and certain services which would have to be postponed or provided in a smaller scale; on the other, a more cynical view would focus on the fact that it can be a tool for governments to hide deficits and incur in consumption or other spending that leads to political gain. Maskin and Tirole (2008) develop a theoretical model in which they drop the assumption that the public partner maximizes social welfare2_{and show through} which mechanisms these hidden or implicit deficits can be ran by governments and how the public official’s incentives (the benefit of certain interest groups over others) may lead him or her away from the less costly option.

Not much has been done in the way of econometric analysis of this kind of investment, particularly relating it to economic growth, partly because it is still a relatively recent phenomenon and data spanning a considerable amount of years and countries has only recently been made available. One case of such analysis was performed by Jasiukevicius and Vasiliauskaite (2013) who measure correlation between GDP Growth and PPP Market Development (defined as both capital costs and number of projects) for EU countries. They look at the impact of the first variable on the second by using different time lags PPP market data and 5 shortening stages of different duration, and then performing a comparison of the correlations’ means in order to compare the strength

1_{Flyvbjerg et al. (2002) conducted a study of 258 projects of transport infrastructure procured}

using traditional public investment strategies, spanning 20 countries, and found an underestimation of the costs occurred in 90% of the cases.

2_{This option is supported by a considerable number of studies, one of which was done by}

Cadot et al. (2002) who found that the main motivation for the building of roads and railways in the re-election of politicians.

(7)

7

of the estimated correlations. Also, they hypothesize that projective GDP data will have a stronger relation with PPP market data, but no statistically significant results were found to support this. Finally their research points to quite different relations between groups of countries, suggesting that other factors (possibly country-specific) may be at play.

The present work intends to build on previous literature and add to it by finding whether or not PPP Capital seems to have a positive role in causing economic growth. There has been, to the extent the author was able to find in surveying the existent literature, no work dedicated to testing this hypothesis. Such analysis is to be performed using the Production-Function approach to the investigation of the effects of Public Capital on economic growth. In articles that follow this method, the underlying theoretical framework is usually an aggregate Cobb-Douglas production function augmented with public capital stock and assuming constant returns to scale across all inputs3_.

Romp and de Haan (2007) provide an extensive summary of the ways in which various authors have tried to deal with the most important problem related with the estimation of a production function – the potential for reverse causation if, as largely accepted, capital investments depend on income through a savings function4_{- the implication of} which is that it becomes “difficult to identify the results of regressions (…) as a production function relationship”. The proposed solutions identified by these authors include: designing a test in such a way that it is clear how the causality runs, estimating panel models, estimating simultaneous equation models and using instrumental variables. I chose the latter option, for which a more detailed explanation is provided in the “Empirical Approach” section.

3_𝑌

𝑡 = 𝐴𝑡𝐿𝛼𝑡𝐾𝑡 𝛽

𝐺_𝑡𝜑 (where G represents the public capital stock). In per capita terms and taking the natural logarithm, this becomes: 𝑙𝑛𝑌𝑡

𝐿_𝑡= 𝑙𝑛𝐴𝑡+ 𝛽𝑙𝑛 𝐾_𝑡 𝐿_𝑡+ 𝜑𝑙𝑛 𝐺_𝑡 𝐿_𝑡. 4_∆𝐾 𝑡 = 𝑠𝑌𝑡− 𝑑𝐾𝑡.

(8)

8

Methodology

The selected approach closely follows that of Calderón and Servén (2002), which in turn is related to the one used in Canning (1999). It consists of the estimation of an aggregate production function augmented with PPP capital, rather than infrastructure capital as in the mentioned work.

Formally, the production function can be specified as:

𝑌 = 𝐴𝐾𝛼𝐻𝛽𝑍𝜑𝐿1−𝛼−𝛽−𝜑 (1)

Where Y represents total national income (GDP), A is technology, K is the physical capital stock (deduced of PPP Capital), H is human capital, Z is the PPP Capital stock and L is Labor. By expressing all variables in logs and assuming constant returns to scale, it can be rewritten as:

𝑦 = 𝑎 + 𝛼𝑘 + 𝛽ℎ + 𝜑𝑧 + (1 − 𝛼 − 𝛽 − 𝜑)𝑙 (2) Our focus will be on the 𝜑 parameter, which captures the elasticity of the dependent variable with respect to PPP capital. Such can only be asserted given that we have the means to remove Z from K, making sure that PPP capital does not appear twice in the equation. Using the data for PPP Investments by year, I then proceed to compute the PPP Capital Stock from each year and then subtract it from the total Capital Stock. Given that PPP capital is mostly constituted of infrastructure capital and, therefore, its depreciation occurs at a much slower than average rate and that the time span of the analysis performed is of only 22 years, no depreciation is considered for this kind of capital.

Data collection

In what concerns PPP projects, data was requested and received from the European PPP Expertise Center (EPEC). This institution collects it from a variety of sources and cross-checks it against the European Investment Bank’s own files, projects with a capital value inferior to EUR 5 million are excluded (EIB, 2010). Also regarding the EPEC’s methods, it is important to mention that not all of the projects that make up the dataset have been

(9)

9

confirmed and validated by EPEC members and, as a consequence, all usage of this information should be treated with caution. Data concerning all other variables (GDP, Capital Stock, Human Capital5_{and Labor}6_{) was collected from the Penn World Table,} version 8.1. Data span includes 26 European countries (see full list in Appendix B) and 22 years (1990-2011) for a total of 572 observations.

Chart 2 shows how the stock of PPP Capital has evolved, relative to GDP, in the 5 countries where the average stock is the highest. Portugal is the clear leader with a stock of PPP Capital above 7% of GDP in 2010 and 2011, while the UK is second with over 5%. The main jumps in the data occur in 2006 in Cyprus, due to a major project regarding the development and operation of the country’s two international airports (World Commerce Review, 2015), and in Portugal between 2008 and 2010, caused by 19 projects undertaken mainly in the transportation infrastructure sector.

Chart 2 (Data sources: EPEC and PWT 8.1)

5_{Index of human capital per person, based on years of schooling (Barro/Lee, 2012) and returns}

to education (Psacharopoulos, 1994)

6_{Number of persons engaged (in millions)}

0% 1% 2% 3% 4% 5% 6% 7% 8% 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

PPP Capital to GDP ratio in Top 5 countries

Cyprus Greece Hungary Portugal United Kingdom

(10)

10

Empirical Approach

Firstly, equation (2) is rewritten in terms of ratios to the labor force, for the purposes of estimation. Doing so yields:

𝑦_𝑖𝑡− 𝑙_𝑖𝑡 = 𝑎_𝑖 + 𝑏_𝑡+ 𝛼(𝑘_𝑖𝑡− 𝑙_𝑖𝑡) + 𝛽(ℎ_𝑖𝑡− 𝑙_𝑖𝑡) + 𝜑(𝑧_𝑖𝑡− 𝑙_𝑖𝑡) + 𝜀_𝑖𝑡 (3) Where the i and t subscripts refer to countries and years, respectively, 𝑎𝑖 is country-fixed productivity effect and 𝑏𝑡 is a time-fixed productivity effect. Finally, 𝜀𝑖𝑡 is an error term that is assumed to be random and uncorrelated between all observations.

In order to tackle the issue mentioned in the “Literature Review” segment, namely the potential for reverse causation, lagged values of the explanatory variables are employed. This can be seen as the most appropriate way to follow an Instrumental Variable approach, given that suitable exogenous instruments are most likely unavailable. As mentioned in Calderón and Servén (2002), if the error term 𝜀_𝑖𝑡 is serially uncorrelated and the regressors are weakly exogenous, then the second and higher order lags of the regressors represent valid instruments. Also, Sargan tests of orthogonality between the instruments and the error term will be performed to confirm the validity of the former.

However, other issues besides the possibility of two-way causality remain to be considered. Specifically, cross-country heterogeneity and omitted common factors: in order to address the first, I allow for country-specific effects aimed to account for both varying production functions as well as eventual different (perhaps political) criteria used in the selection of PPP projects that may affect their productivity; as for the second, with the intention of avoiding residual correlation across countries, time-specific effects are allowed for (these may include the business and financial cycles, which are very likely to be coincident in most of the European economies analyzed).

A final concern regarding the estimation by OLS of equation (3) is the non-stationarity of the data, which is likely to skew our results in the way of overestimating the effects of the capital stock on GDP due to a spurious correlation (Romp and de Haan, 2007) (the results of such estimation are presented below in Table 2). As such, an alternative specification would be to take first differences of equation (3):

(11)

11

∆(𝑦_𝑖𝑡− 𝑙_𝑖𝑡) = (𝑏_𝑡− 𝑏_𝑡−1) + 𝛼∆(𝑘_𝑖𝑡 − 𝑙_𝑖𝑡) + 𝛽∆(ℎ_𝑖𝑡− 𝑙_𝑖𝑡) + 𝜑∆(𝑧_𝑖𝑡− 𝑙_𝑖𝑡) + ∆𝜀_𝑖𝑡 (4) I employ this tool and proceed to perform both OLS and Blundell and Bond’s difference-GMM regressions, for which the results are presented in Tables 3 and 4, respectively. The choice of the second method is meant to tackle the econometrical issue resulting from the correlation between the unobserved panel-level effects and the lagged dependent variable, which renders estimators such as OLS inconsistent (Blundell and Bond, 2008). It is an expansion of the Arellano-Bond difference-GMM estimator, and builds upon it by making the assumption that the fixed effects are not correlated with the first differences of instrument variables, and thus permitting the introduction of more instruments that can improve efficiency (Roodman, 2009). Additionally, this is a technique tailored for data sets of the “small T, large N” kind, like the one used for this work – with a large T, the number of instruments can become too great and the dynamic panel bias less significant, creating a situation where simpler estimators become more applicable and advisable.

Estimation Results

Before proceeding to the various estimations performed, I start by presenting the sample correlations between all variables, dependent (Real GDP) and independent (Capital Stock, Human Capital and PPP Stock), in Table 1.

Table 1

Sample Correlations in levels (below diagonal) and differences (above diagonal)

Real GDP Capital Stock Human Capital PPP Stock

Real GDP - 0.2201* 0.0501 0.0268

Capital Stock 0.9107* - 0.0365 -0.0148

Human Capital 0.1090* 0.0630 - 0.0463

PPP Stock 0.4599* 0.4473* 0.1833* -

(12)

12

In accordance with most of the literature on the subject, output displays a statistically significant (at the 99% level) positive correlation with all three of the explanatory variables, particularly so with the Capital Stock. Also as expected, the magnitude of the correlations is found to be considerably lower when the variables are expressed in differences – in this case, only the correlation between the Real GDP and the Capital Stock retains statistical significance. An additional point of interest is the fact that the Capital and PPP stocks turn out to be negatively correlated (even if insignificantly so), in first differences, which is aligned with argument presented earlier that PPP projects are sometimes used by governments as an alternative to other forms of investment, when budget restrictions so imply.

Regarding the estimation results, Column 1 of each Table presents the empirical estimates obtained using non-lagged values of the explanatory variables (shown for all different estimations executed despite the econometrical issue of possible reverse causation, for the purpose of providing a more comprehensive “feel” for the data), Column 2 results from lagging the independent variables by up to and including two periods, while in Column 3 these were lagged by up to and including three periods – individual coefficients are then added in order to obtain long-term coefficients (this requires the assumption of no feedback effects, which is made here).

I start by presenting, in Table 2 below, the results from the estimation by OLS of the proposed Production Function augmented with PPP Stock, in levels. In the case of all 3 Columns of this particular Table, the utilized model includes country fixed-effects and a full set of year dummies that turned out to be largely significant.

(13)

13

Table 2

PPP Stock–Augmented Production Function: OLS Levels Dependent Variable: Real GDP

per worker (in logs)

[1] No Lags [2] Two Lags [3] Three Lags Capital Stock 0.473*** 0.497*** 0.486*** Human Capital -0.632*** -0.516** -0.554** PPP Stock -0.000015 -0.0086** -0.012*** Intercept 5.625*** 5.193*** 5.341*** R**2 0.7982 0.8114 0.8005 Number of Observations 572 572 572 Number of Countries 26 26 26

Notes: All variables are measured per worker and expressed in logs. ** p< 0.05, *** p<0.001.

Using either two or three lags for the independent variables, the coefficients show a positive statistically significant effect for the Capital Stock and negative statistically significant effects for both Human Capital and the PPP stock (although in a much smaller magnitude for the latter). However, these estimates result from specifications that are vulnerable to the issue of spurious correlation due to the non-stationarity of the data, as previously alluded to.

I now turn to the estimation, again by OLS, of equation (4) – the augmented Production Function in differences. This alternative specification provides stationarity of the data and, as such, is not subject to the vulnerability the previous estimates were. Again, long-term coefficients of twice and three times lagged independent variables were used, and a complete set of year dummies was included in the model. Table 3 displays the ensuing results.

(14)

14

Table 3

PPP Stock–Augmented Production Function: OLS First-Differences Dependent Variable: Real GDP

[1] No Lags [2] Two Lags [3] Three Lags Capital Stock 0.353*** 0.198*** 0.191*** Human Capital 0.222 0.435 0.00564 PPP Stock -0.000623 -0.000456 -0.00482 R**2 0.2583 0.2603 0.3198 Number of Observations 572 572 572 Number of Countries 26 26 26

Note: All variables are measured per worker and expressed in logs. *** p<0.001.

The obtained coefficients identify positive effects for the Capital Stock and Human Capital, with the only instance of statistical significance being that of the Capital Stock, and negative effects for the PPP Stock. Results are similar in the cases of explanatory variables being lagged twice and three times. In any case, these estimates were encountered by the usage of a specification that continues to be vulnerable to an important issue – the inconsistency of the estimators caused by the correlation of the dependent variables (at all lags) with the unobserved panel-level effects.

In order to solve this problem, I apply the estimator constructed by Blundell and Bond (1998) which is, in essence, an extension of Arellano and Bond’s difference-GMM estimator. This provides us with consistent estimates for the model that is being used, and allows for the testing of the validity of the instruments by performing Sargan tests of orthogonality between them and the error term. Table 4 shows the outcome of this specification.

(15)

15

Table 4

PPP Stock–Augmented Production Function: GMM Dependent Variable: Real GDP

[1] No Lags [2] Two Lags [3] Three Lags Capital Stock 0.126*** -0.0184 -0.121 Human Capital 0.955* 1.142* 0.611** PPP Stock 0.000199 -0.00152 -0.000377

Sargan Test (p-value) 0.9906 0.4232 0.1384

Number of Observations 572 572 572

Number of Countries 26 26 26

Note: All variables are measured per worker and expressed in logs. * p<0.05, ** p<0.01, ***

p<0.001.

The coefficients resulting from the implementation of the chosen GMM estimator, both with two and three times lagged values of the independent variables, display a positive and statistically significant effect of Human Capital on output and insignificant negative effects of the Capital and PPP stocks on the same. The Sargan Test p-values appear to be supportive of the model employed, and the similarity of the results exhibited in Columns 2 and 3 both in signal and magnitude invite a positive appreciation regarding their robustness.

Concluding Remarks

During the last two and a half decades, PPP projects have been implemented throughout Europe, with nearly every country utilizing this financial structure to some extent, spanning critical sectors such as health, transportation and waste management, among others. In the cases of Portugal and the UK, the ratio of PPP Capital stock to GDP has risen to values above 5% due to projects of great magnitude like the Vasco da Gama bridge in Lisbon or the redevelopment of the Royal Liverpool University Hospital. Moreover, current budget restrictions in light of high budget deficits and Debt to GDP

(16)

16

ratios may lead to a situation where this kind of investment becomes even more relevant.

The task of asserting whether or not PPP Capital has the capability to hamper or enhance economic growth is one that presents a number of econometrical difficulties, a fact which implies a careful consideration of the methodology to be put in place. In the present work, I started with a classic Cobb-Douglas production function augmented with PPP Capital and proceeded to rewrite it for estimation purposes by taking logs of the variables and expressing them in ratios to the labor force. Finally, estimations were performed by OLS (in levels and in first differences) and by Blundell and Bond’s GMM estimator.

Given the econometrical issues explained in the previous sections, the focus of my assessment of the results will be on the final model. The empirical estimates display, for every specification used, a very low level of statistical significance, meaning that the evidence seems to support the claim made by Valila (2005) that “there is no macroeconomic case for – or against – public-private partnerships”. This could symptomatic of offsetting effects, of a lack of data or simply the possibility that PPP Capital does not, in fact, have a noteworthy impact on output. The lack of significance of the coefficients is line with the results obtained in previous work on Public and Infrastructure capital stock that used similar theoretical frameworks, as exemplified by Cadot et al. (1999) or Bonaglia et al. (2001).

Given that the study of PPP’s from a macroeconomic standpoint is still fairly undeveloped, there is vast room for future research to be performed on the subject. One of the ways in which that could be directed would be to investigate the effects on economic growth of PPP’s by sector, given that it is quite plausible that they could be different from one another. However, the necessary data for such an undertaking is yet unavailable. Additionally, as referenced by Romp and de Haan (2007), the Production Function method may not be capable of providing the necessary understanding on the complexity of the channels through which economic growth is influenced by infrastructure, given that such link is quite surely characterized by non-linearities and heterogeneity. Consequently, the evaluation of PPP projects and their effects on productivity may to be performed on a case by case basis.

(17)

17

References

Alonso-Conde, Ana Belen, Brown, Christine and Rojo-Suarez, Javier (2007), "Public private partnerships: Incentives, risk transfer and real options", Review of Financial Economics, Vol. 16(4), pp. 335-349.

Blundell, Richard, and Bond, Stephen (1998), "Initial conditions and moment restrictions in dynamic panel data models", Journal of Econometrics, 87, pp. 115–143.

Bonaglia, Federico, La Ferrara, Eliana and Marcellino, Massimiliano (2001), "Public Capital and Economic Performance: Evidence from Italy", IGIER Working Paper No. 163. Cadot, Olivier, Roller, Lars-Hendrik and Stephan, Andreas (1999), "A Political Economy Model of Infrastructure Allocation: An Empirical Assessment", CEPR Discussion Paper No. 2336.

Calderón, César and Servén, Luis (2002), "The Output Cost of Latin America’s Infrastructure Gap", Central Bank of Chile Working Paper No. 186.

Canning, David (1999), “Infrastructure’s contribution to aggregate output”, World Bank Policy Research Discussion Paper 2246.

Feenstra, Robert C., Inklaar, Robert and Timmer, Marcel P. (2015), "The Next Generation of the Penn World Table" forthcoming in American Economic Review, available for download at www.ggdc.net/pwt.

Flyvbjerg, Bent, Holm, Mette Skamris and Buhl, Soren (2002), "Underestimating Costs in Public Works Projects: Error or Lie?”, Journal of the American Planning Association, 68.3, pp. 279-95.

Grimsey, Darrin and Lewis, Mervyn K. (2002), "Evaluating the risks of public private partnerships for infrastructure projects", International Journal of Project Management, Vol. 20(2), pp. 107-118.

Jasiukevicius, Linas & Vasiliauskaite, Asta (2013), "The relation between economic growth and Public-Private Partnership market development in the countries of the European Union", Kaunas University of Technology, Economics and Management, 18(2), pp. 226-236.

(18)

18

Kappeler, Andreas and Nemoz, Mathieu (2010), "Public-Private Partnerships in Europe – Before and During the Recent Financial Crisis", European Investment Bank, Economic and Financial Report 10/04.

Maskin, Eric and Tirole, Jean (2008), "Public–private partnerships and government spending limits", International Journal of Industrial Organization, Vol. 26(2); pp. 412-420.

Moszoro, Marian (2010), "A theory of Efficient Public-Private Capital Structures", IESE Business School, 1-23.

Romp, Ward and De Haan, Jakob (2007), “Public Capital and Economic Growth: A Critical Survey”, Perspektiven der Wirtschaftspolitik 8 (s1), pp. 6–52.

Roodman, David (2009), "How to do xtabond2: An introduction to difference and system GMM in Stata", The Stata Journal, Number 1, pp. 86-136

Sadka, Efraim (2006), "Public-Private partnerships: A political economics perspective", International Monetary Fund, WP/06/77.

Spackman, Michael (2002), “PPPs: Lessons from the British Approach,” Economic Systems, Vol. 26, No. 3, pp. 283–301.

Takashima, Ryuta, Yagi, Kyoko and Takamori, Hiroshi (2010), "Government guarantees and risk sharing in public–private partnerships", Hiroshi Review of Financial Economics, Vol. 19(2), pp. 78-83.

Valila, Timo (2005), "How expensive are cost savings? On the economics of public-private partnerships", EIB papers, 10(1), pp. 95-115.

World Commerce Review (2015), "Cyprus' Airports PPP - Best Transport Project in Europe", World Commerce Review, June 2015.

(19)

19

Appendix A: Descriptive Statistics

A.1) Data as collected from sources

Variable Obs. Mean Std. Dev. Min Max

Real GDP 572 440159.6 631307.4 9374.706 2905727 Capital Stock 572 1460098 2189180 29804.75 10404254 Human Capital 572 2.91371 0.233 2.231248 3.535638 PPP Stock 572 3834.11 14647.73 1 135601 Labor 572 8.181565 9.934602 0.18724 41.3802

A.2) Data as used for estimation purposes – in logs and per worker

Variable Obs. Mean Std. Dev. Min Max

Real GDP 572 10.68576 0.4775765 9.095164 11.63418

Capital Stock 572 12 0.5692399 10.25894 12.84867

Human

Capital 572 1.066186 0 0.8025612 1.262894

(20)

20

Appendix B: Countries included in Dataset

Austria Italy

Belgium Latvia

Bulgaria Lithuania

Cyprus Luxembourg

Czech Republic Netherlands

Denmark Poland Estonia Portugal Finland Romania France Slovakia Germany Slovenia Greece Spain Hungary Sweden

Ireland United Kingdom

The effect of PPP capital on economic growth