Changes in Economic Effect of Infrastructure and Financing Methods *

Transcription

1 Policy Research Institute, Ministry of Finance, Japan, Public Policy Review, Vol.12, No.1, March Changes in Economic Effect of Infrastructure and Financing Methods * Masaki Nakahigashi Associate Professor, Faculty of Economics, Niigata University / Senior Research Fellow, Policy Research Institute, Ministry of Finance, Japan Naoyuki Yoshino Dean, Asian Development Bank Institute / Professor Emeritus, Keio University Abstract This paper analyzes the recent changes in the economic effect of infrastructure focusing on the productivity effect of public investment among a declining trend in public investment, and discusses how to promote the spending of private-sector money for infrastructure investment through the creation of incentive mechanisms. In secondary and tertiary industries, the marginal productivity of public capital has declined in many regions of Japan, but even in recent years, there have still been positive productivity effects from public capital. Since the beginning of the 2000s until today, although the regional allocation of public investment has changed, the structure of regional disparity in the productivity effect of public capital has remained unchanged. The important factors for further promoting the use of private sector resources in infrastructure development include: (1) governmental involvement in high-risk infrastructure projects, (2) an earnings structure that appropriately reflects the economic effect of infrastructure (e.g. return of the part of spillover tax revenues to infrastructure investors) and (3) systems designed to increase incentives for infrastructure-operating entities. Keywords: infrastructure, productivity effect of public capital, public-private partnership, infrastructure finance JEL Classification: G23, G28, H54 I. Introduction The cumulative public debt of Japan is higher than that of the Greek government, due to the expansion of expenditure on social security for population aging (Yoshino and Taghizadeh-Hessary (2015, 2016)). It has become difficult to expand government spending, including public investment. In addition, after the ceiling collapse accident of Sasago Tunnel in December 2012, measures for containing the aging and maintaining the quality of infrastructure have become a challenge. Infrastructure is one of the essential services for our daily life, and therefore we require measures to enhance infrastructure development with ap- * We appreciate the valuable comments from participants in the review meetings of the Financial Review in Ministry of Finance. All errors remain our responsibility. Opinions shown in this paper are personal and do not indicate views and policies of the Asian Development Bank and the Asian Development Bank Institute.

2 48 M Nakahigashi, N Yoshino / Public Policy Review propriate maintenance. This paper analyzes two aspects. Firstly, we reveal through an analysis using the translog production function whether the productivity effect of public investment has changed in recent years. Secondly, we propose measures to promote infrastructure investment, made necessary by infrastructure aging, by the private sector rather than by the government, and describe measures to maintain infrastructure while reducing expenditure from public funds and to promote the economic effect of infrastructure. II. Changes to and evaluation of productivity effect of public capital in recent years In section II, we clarify whether the productivity effect of public capital has remained in recent years. Specifically, we assess contributions to productivity through infrastructure development. In Japan, much research on the productivity effect of public capital was carried out, especially around the year More recently, Annala, Batina and Feehan (2008), Miyara and Fukushige (2008) and Okubo (2008) base their analyses on data before the late 1990s. Hayashi (2009) and Miyagawa, Kawasaki and Edamura (2013) carry out their analyses using recent prefectural macro data, as compared with previous research, and indicated that output elasticity of public capital has increased in recent years. 1 We estimate the productivity effect of public capital by industry and region based on the production function, as is adopted in Yoshino, Nakajima and Nakahigashi (1999), using the 36 years of data from FY1975 to FY2010. We also clarify whether the output elasticity of public capital has increased in recent years. II-1. Estimation model of the economic effect of public capital In our analyses, output is the total of real private value-added, and factors of production consist of: (1) private capital, (2) labor and (3) public capital. We use a trans-log production function, which represents flexibility the production technology structure. lny t -lny=α K (lnk Pt -lnk P )+α L (lnl t -lnl)+α G (lnk G,t -lnk G ) +β KL (lnk Pt -lnk P )(lnl t -lnl)+β KG (lnk P,t -lnk P )(lnk G,t -lnk G ) +β LG (lnl t -lnl)(lnk G,t -lnk G )+ 1 β KK (lnk P,t -lnk P ) β LL (lnl t -lnl) β GG (lnk G,t -lnk G ) 2 Where the trans-log production function is a log-linear approximation around the average of each variable and the symmetry of second derivatives is assumed. Furthermore, it is 1 See Yoshino, Nakajima and Nakahigashi (1999), Nakahigashi (2003), and Yoshino and Nakahigashi (2004) for research on the productivity effect of public capital around See Nakahigashi (2015) for research on the productivity effect of public capital using Japanese macroeconomic data.

3 Policy Research Institute, Ministry of Finance, Japan, Public Policy Review, Vol.12, No.1, March assumed that β KK < 0, β LL < 0 and β GG < 0 based on the law of diminishing marginal productivity. However, in a trans-log production function, there are many cross terms of factor variables and it is likely that multicollinearity will occur. Therefore, it is assumed that the production function is constant returns to scale with respect to labor and private capital, i.e. the homogeneous of degree one in the function of labor and private capital, and the factor payment for each factor input follows the marginal productivity proposition of the allocation of income. When meeting these assumptions, income is fully distributed to labor and private capital, and the income share of private capital is equal to the output elasticity of private capital and the income share of labor is equal to the output elasticity of labor. Therefore, in order to satisfy these assumptions, we impose the following parameter constraints: { α K +α L =1 β KK +β KL =0 (1) β KL +β LL =0 β KG +β LG =0 where these restrictions are a sufficient condition for this assumption. Our estimated equation system is as follows: (lny t -lny)-(lnl t -lnl) =α K[(lnK P,t -lnk P )-(lnl t -lnl)] +α G (lnk G,t -lnk G )+β KL[(lnK P,t -lnk P )(lnl t -lnl) - (lnk P,t -lnk P ) 2 - (lnl t -lnl) ] +β KG (lnk G,t -lnk G )[(lnk P,t -lnk P )-(lnl t -lnl)] β GG (lnk G,t -lnk G ) 2 +ε P,t (2) S L,t =1-α K +β KL [(lnk P,t -lnk P )-(lnl t -lnl)]+β KG (lnk G,t -lnk G )+ε S,t (3) where S L,t represents the labor share in period t, and ε P,t and ε S,t are the error terms of the labor share function and production function, respectively. In addition, under producers profit maximization, the following sign conditions are necessary: α K 0, α G 0, β KL > 0 and β GG < 0. II-2 Methods for measuring the productivity effect of public capital Productivity effect of public capital, in general, is represented by the output elasticity or marginal productivity by marginal increase in factor input. In our paper, the productivity effect of public capital is the increase in output by marginal increase in factor input, including the effect of variation in factor productivity, which is same as Yoshino and Nakano (1994),

4 50 M Nakahigashi, N Yoshino / Public Policy Review Yoshino, Nakajima and Nakahigashi (1999) and Nakahigashi (2003). Our marginal productivity of public capital from equation (2) is as follows: 2 dy dk =η Y G K +η η KG η +β K P KG Y η Y KG K P G η (1-η )+β +η η K G L-β KG L (4) K K P KP KL G η(1-η L L )+β KL K G where η, η K P L, η K, represent the output elasticity of private capital, labor and public capital respectively. Based on equation (2), the output elasticity of labor is the same as equation G (3) and the output elasticity of private capital and public capital is expressed as follows: η =α K P K+β KL [( lnl-lnl)-(lnk P -lnk P )]+β KG (lnk G -lnk G ) (5) - - η =α - K G G+β KG [( lnk P -lnk P )-(lnl-lnl)] +β GG (lnk G -lnk G ) (6) The first term on the right-hand side of equation (4) is the marginal productivity of public capital (direct effect of public capital), the second term on the right-hand side of equation (4) indicates the indirect effect of private capital, which shows the level of output increase by varying the productivity of private capital through marginal increase of public capital, and the third term on the right-hand side of equation (4) is the indirect effect of labor, which represents the level of increase in output by varying labor productivity through a marginal increase in public capital. II-3 Industrial and regional classification Our industrial classification is reconstructed based on economic activity in the System of National Accounts of Japan and we classify industries into primary industry, secondary industry and tertiary industry. 3 Our Electricity, Gas and Water Supply industry includes Electricity, Gas and Water Supply of Producers of Government Services, and our Service industry includes Service Activities of Producers of Government Services and Producers of Private Non-profit Services to Households. In our production function, it is assumed that parameters related to the public capital are able of taking a different value depending on region. This assumption considers the influence of the regional allocation policy for public investment, which reflects the unequal relationship between potential supply level of the infrastructure services and actual utilization level of infrastructure services among regions. It is assumed that parameters not related to public capital can take different values between regions because there is the possibility that technical differences among regions will occur in our rough industrial classification, and therefore all parameters in our production function can have different values depending on region. 2 See Yoshino, Nakajima and Nakahigashi (1999) for deriving equation (4). 3 In Prefectural Accounts of Japan, industrial classification changes after FY2005. Especially, in secondary and tertiary industries, industries included in each industry before FY2004 and that after FY2005 are different. Therefore, discontinuities in the time series data occur in secondary and tertiary industries.

5 Policy Research Institute, Ministry of Finance, Japan, Public Policy Review, Vol.12, No.1, March In addition, in order to internalize regional spillover effects of the productivity effect of public capital, we re-aggregated 46 prefectures, excluding Okinawa, into 11 regions. Table 1 shows the regional segmentation in our analysis, and the way in which data is created for use in our analysis is shown in the appendix. II-4 Specification method of the estimation model In addition to the above description, we describe four points for our estimation. Firstly, variables representing the effect of the private sector s technological progress are not included explicitly in equation (2) and equation (3). That is because we acknowledge that the technical progress of the private sector has been embedded in factor inputs such as capital and labor input. Secondly, it is assumed that the error term of the production function (u P,it ) consists of a usual error term (ε P,it ) and the region-specific constant value (μ i ), which represents the effect on output by elements other than factor inputs: u P, it =μ i +ε P, it for all i, t (7) Thirdly, the error vector ε t = (ε P,t u S,t )ʹ meets E (ε t ) = 0, E (ε t εʹt) = Σ in all t. Since equation (3) is derived from the production function of equation (2), the estimation model is a simultaneous equation system and there is the possibility that ε P,t and u S,t, which is the error term of labor share function, are correlated. Finally, Seemingly Unrelated Regression (SUR) is used in estimating our simultaneous Table 1 Regional segmentation in our analysis

6 52 M Nakahigashi, N Yoshino / Public Policy Review model. We can see that explanatory variables of the labor share function appear in explanatory variables in the production function. As indicated by Revankar (1974), SUR estimates of the labor share function in our system are same as least square estimates of the labor share function, and are the best linear unbiased estimator of the labor share function. Therefore, in specifying the estimation model we give the following two steps: first, the labor share function, equation (3), is specified by ordinary least squares estimation while considering the sign condition of parameters. Secondly, we determine the parameters of production function to be estimated based on the specification of labor share function in the first step and then specify the whole estimation model. III. Results of the productivity effect of public capital Based on the specifications of the estimation model in Section II, we show the estimated results using the panel data of 11 regions from FY1975 to FY2010 and estimate the productivity effect of public capital. Note that the estimation model, consisting of equation (2) and equation (3), is represented by a log-linear approximation around the average, which is the average from FY1975 to FY1994. III-1 Result for the production function Table 1, Table 2 and Table 3 show the estimated results of primary, secondary and tertiary industries, respectively. Parameters showing firstly in the respective table are common parameters in all regions, and following coefficient dummies underneath these parameters show the difference from estimates of common parameters. When there is an estimate of common parameter, the estimate of the coefficient dummy indicates the magnitude of the divergence from the common parameter estimate. When an estimate of the common parameter does not exist, the estimate of the coefficient dummy shows a region-specific parameter estimate. The following is described on parameters affected by public capital. Firstly, except for some regions of primary industry, it can be indicated that α G is significantly positive. At least in secondary and tertiary industries, positive α G appears and shows the increase in productivity by increasing in public capital. Secondly, when comparing parameter α G among regions, it is found that the effect of public capital in Southern Kanto and Tokai regions is higher than that in other regions. Furthermore, parameters β KG, which indicate the direction of the impact of the increase in public capital on the factor share, are positive in almost all regions and industries. Thus, it can be said that public capital is a capital-using/labor-saving factor in the sense that capital share is raised by increasing the public capital.

7 Policy Research Institute, Ministry of Finance, Japan, Public Policy Review, Vol.12, No.1, March Table 2 Estimation result for production function in primary industry Note 1: ** and * denote statistical significance at 1% and 5% level, respectively. Note 2: In addition to variables shown in this table, the estimation model includes the regional fixed effect, but the estimated result of this effect has been omitted.

8 54 M Nakahigashi, N Yoshino / Public Policy Review Table 3 Estimation result for production function in secondary industry Note 1: ** and * denote statistical significance at 1% and 5%, respectively. Note 2: In addition to variables shown in this table, the estimation model includes the regional fixed effect and the discontinuity dummy, which is a constant dummy of each region after 2005 in order to take into account the discontinuity of data due to a change of industrial classification after These estimated results have been omitted.

9 Policy Research Institute, Ministry of Finance, Japan, Public Policy Review, Vol.12, No.1, March Table 4 Estimation result for production function in tertiary industry Note 1: ** denotes statistical significance at the 1% level. Note 2: In addition to variables shown in this table, the estimation model includes the regional fixed effect and the discontinuity dummy, which is a constant dummy of each region after 2005 in order to take into account the discontinuity of data due to a change of industrial classification after However, in this table, estimated results of these have been omitted.

10 56 M Nakahigashi, N Yoshino / Public Policy Review III-2 Derivation of point estimates of productivity effect of public capital We derive point estimates of productivity effect of public capital using the estimated results of the production function in Table 2, Table 3 and Table 4, and compare point estimates at FY2010 which is the final year of our analysis to that at FY1990, in which the Japanese economy was in a favorable situation. Marginal productivity is used in deriving the point estimates of the productivity effect of public capital. Table 5 denotes the productivity effect of public capital by industry and region at FY2010. When comparing the productivity effect of public capital between industries, tertiary industry has the largest effect and primary industry has the lowest. Especially in primary industries, both the direct effect and the indirect effect are near zero. When looking at the indirect effect in detail, in the secondary industry the indirect effect of labor is greater than that of private capital, in tertiary industry, on the other hand, the indirect effect of private capital is greater than that of labor. It can be seen that in secondary industry, under fixed factor prices, an increase in public capital will result in an increase in output via, mainly, an increase in labor productivity. In the tertiary industry, it can be seen that under fixed factor prices, an increase in public capital will result in an increase in output via an increase in la- Table 5 Point estimates of productivity effect of public capital at FY2010

11 Policy Research Institute, Ministry of Finance, Japan, Public Policy Review, Vol.12, No.1, March bor productivity. Next we compare the productivity effect of public capital at FY2010 to that at FY1990. Table 5 show the point estimate of the productivity effect of public capital at FY2010. Figure 1 and Figure 2 show the point estimate of the productivity effect of public capital in secondary industry and tertiary industry, respectively. Both figures show that its point estimate of FY2010 compared to FY1990 has been lowered. It can be seen that the effect of increasing productivity by marginal increase of public capital is smaller. Furthermore, when comparing its decrease, its decrease of secondary industry is larger than that of tertiary industry. III-3 Interval estimation of productivity effect of public capital If the estimates of parameters to be used in deriving the productivity effect of public capital are the realization of the random variables, it is possible to derive an interval estimation of the productivity effect of public capital. Then we compare the level of the productiv- Figure 1 Changes in productivity effect of public capital in secondary industry Figure 2 Changes in productivity effect of public capital in tertiary industry

12 58 M Nakahigashi, N Yoshino / Public Policy Review ity effect of public capital from a statistical point of view through interval estimation at FY2010 and FY1990. Productivity effect of public capital can be calculated from equation (4). If all factor inputs in our model are defined as non-random variables, the only estimated parameters are random variables. Then we can readily derive the direct effect of public capital because the direct effect can be represented by a linear combination of random variables. The indirect effect, on the other hand, cannot be derived easily because the indirect effect is the ratio of a random variable to a random variable. Therefore, in our paper, interval estimations are adopted only in the direct effect, i.e. the marginal productivity of public capital. Our interval estimation is the confidence interval at the 95% level under the assumption that error terms are normally-distributed random variables. Table 6 and Table 7 show the interval estimation of productivity effect of public capital at FY2010 and FY1990, respectively. It should be noted that the interval estimates of the productivity effect of public capital are calculated using all the estimated parameters regardless of the statistical significance and that we calculate the output elasticity as well as marginal productivity. The first point of view is on the industrial difference in the productivity effect of public capital. The productivity effect of primary industry is very low compared to that of other industries. In primary industry, it is found that public capital does not contribute very much to productivity. In the secondary and tertiary industries, on the other hand, the productivity effect of public capital has been significantly positive at FY2010. It can be seen that even in Table 6 95% confidence interval of productivity effect of public capital (FY2010)

13 Policy Research Institute, Ministry of Finance, Japan, Public Policy Review, Vol.12, No.1, March Table 7 95% confidence interval of productivity effect of public capital (FY1990) recent years the productivity effects of public capital have existed. The second point of view is on the regional differences in the productivity effect of public capital. We compare the productivity effect of FY1990 to that of FY2010. In the secondary industry, although output elasticity of all regions except Chugoku and Southern Kyushu regions are unchanged, marginal productivity of the productivity effect decrease in all regions except Tohoku and Southern Kyushu. In the tertiary industry, although the output elasticity of public capital increases significantly in all regions, the marginal productivity of public capital decreases or is unchanged. Some research, e.g. Hayashi (2009) and Miyagawa, Kawasaki and Edamura (2013), shows that there is a possibility that the output elasticity of public capital has been growing in recent years. Our results find that especially in tertiary industry output elasticity of public capital increase in recent years, marginal productivity of public capital, on the other hand, has decreased in almost all regions. This should be noted in order to compare it with the efficiency of public capital. Finally, we try to classify the regional marginal productivity of public capital based on interval estimations of its productivity effect. In secondary industry, the structure of its regional differences in marginal productivity of public capital was almost the same at FY2010 and FY1990. The lowest region in terms of marginal productivity is Hokkaido, and the second lowest regions are Tohoku, Hokuriku and Shikoku. On the other hand, the highest regions of marginal productivity are the metropolitan regions of Southern Kanto, Tokai, and Kinki. Furthermore, in the tertiary industry, structures of regional differences in marginal

14 60 M Nakahigashi, N Yoshino / Public Policy Review productivity are almost the same at FY1990 and FY2010. The lowest region in terms of marginal productivity is Hokkaido, and the second lowest regions are Tohoku, Hokuriku, Shikoku, Northern Kyushu and Southern Kyushu. On the other hand, the highest region is Southern Kanto and the second highest is Tokai. From our analysis, since 2000, although economic efficiency has come to be a consideration in the regional allocation of public investment, it is found that the structure of regional differences in the productivity effect of public capital has not changed significantly between FY1990 and FY2010. IV. Infrastructure financing through private funds In recent years, public-private partnerships (PPP), including the use of private funds, are being emphasized. Advantages to infrastructure development utilizing private funds are: (1) coming under pressure to shorten period of construction of infrastructure as much as possible and to complete the infrastructure as early as possible, (2) to complete the infrastructure with minimal construction cost, and (3) to make every effort to be profitable while reducing costs after completion of the infrastructure. However, the Cabinet Office PFI Promotion Office (2015) shows that the accumulated number of PFI projects which is a major measure of the utilization of private resources, since FY1999 when PFI Law was enacted until FY2014 is 489 and the accumulated business scale is about 4.5 trillion yen, which is not large compared to the amount of all infrastructure projects. In this section, we will discuss the methods and conditions necessary in promoting expanded use of private funds in infrastructure development. IV-1 Public-private cooperation in high-risk projects Infrastructure projects have a variety of risks, for example due to: (1) regime change, for example causing stoppages due to change in local governor before completing the infrastructure project, (2) cost increase, for example creating an additional interest burden through extensions in construction period and delays in land acquisition, (3) revenue, for example facing unexpected decreases in revenue due to fee setting and decreased traffic, and (4) additional burdens, for example through compensation for noise occurring after the completion of an infrastructure project. 4 Private investors apply various ideas in order to avoid possible risks and earn benefits. Some investors, however, may force the transfer of risks onto the public sector. In Japan, infrastructure projects have been carried out by third-party sectors established by regional governments of various regions, and as a consequence there has been a deficit in the public sector. In these cases, it will be essential to decide risk sharing between public and private sectors in advance. Especially, Viability Gap Funding (VGF) would be one of the appropriate measures to 4 See OECD (2008) concerning the classification of risk under PPP projects.

15 Policy Research Institute, Ministry of Finance, Japan, Public Policy Review, Vol.12, No.1, March apply to infrastructure projects that are indispensable for the public, but which are high-risk and low-earning. For example, Figure 3 shows the case in which a highway project is supplied with 30% funding initially by the government and an increase in returns to private investors. Through the injection of funds from the public sector, the rate of return in the private sector would increase to 10/7 (about 1.428) times of the initial return. Even in projects in which private funds were not contributed due to expectation of low revenue, it is possible to introduce private funds. However, in this example, if the ratio of the injection by the public sector is too high, a moral hazard problem of private investors appears. On the other hand, when this ratio is too low, there is a possibility that the private sector would not put their money into the project at all. 5 IV-2 Increase in rate of return on infrastructure projects through the internalization of spillover effects Infrastructure projects have benefits other than revenue such as tolls. For example, a company that uses a highway could get more benefits than its usage cost from cost reduction and increase in sales through reducing the transit time of raw materials and products. Yoshino and Pontines (2015) estimate the increase in tax revenues of regions along the Figure 3 Injection of public funds to a highway project 5 Under this framework, a revenue bond is one of the measures for raising private funds. See Yoshino (2010) and Yoshino (2012) for revenue bonds.

16 62 M Nakahigashi, N Yoshino / Public Policy Review Philippine highway STAR (Southern Tagalog Arterial Road). In particular, Yoshino and Pontines (2015) evaluate the increase in change rates of business tax revenue and property tax revenue by the opening of the STAR highway. In order to compare the change rate of tax revenues in affected regions to that in unaffected regions they use a Difference-in-Differences, which is used in policy analysis. 6 Estimated equation of the increase in tax revenues due to the STAR highway compared with unaffected regions is as follows. Time t is represented on the basis of the completion point of the highway. In addition, the tax revenue (logarithmic value) can be expressed by following equations: The tax revenue at the time t of the affected region A of the highway (T At ): T At =α A +X At β+ δ t +ε At t The tax revenue at the time t of the unaffected region of the highway (T Nt ): (8) T Nt =α N +X Nt β+ε Nt (9) where α is a constant term and Xʹ is the vector of macroeconomic and microeconomic variables that affect the tax revenue, such as the number of employees and the amount of investment by private companies. δ t in equation (8) represents the effect of affected regions by the highway at the time and t is the error term. The estimated equation is as follows: T it =c+αd g+x it β+ δ t D g +ε it t where c is the constant term, and D g is the dummy variable that the regions affected by the highway are set as 1 and the other regions are set as 0. If δ t is significantly positive, the tax revenue of affected areas would increase as compared with that of unaffected areas. Table 8 shows estimates of the increase in business tax of affected regions as compared to that of unaffected regions before and after the completion of the STAR highway, using results for equation (10). represents the completion year of the STAR highway and its sub- Table 8 Increase in business tax in affected regions as compared to unaffected regions of Philippine STAR highway (10) Source: Yoshino and Pontines (2015) 6 Yoshino and Abidhadjaev (2015) estimate the impact of GDP by the railway network development of Uzbekistan. Its estimation method is same as Yoshino and Pontines (2015).

17 Policy Research Institute, Ministry of Finance, Japan, Public Policy Review, Vol.12, No.1, March script shows the reference time relative to completion time (0) of the STAR highway. In other words, negative subscript represents the number of years before its completion, and positive subscript represents the number of years elapsed after its completion. According to Table 8, it is suggested that business tax has increased in all periods. After the completion of the highway, the increase in business tax is caused by cost reduction and an increase in sales of offices along the highway through shortening time to carry materials and products. In addition, before the completion of the highway, the increase in business tax is due to the increase in sales of construction materials. This increase in business tax and property tax can be said to be an spillover effect due to the highway. If it is possible to confirm the increase in tax revenue is due to the spillover effects of the highway, it might be possible to return the increase in tax revenue to private investors and the public sector. A specific example applying this to highway projects is a fund returning to private investors a fraction of tax revenue created by the spillover effect as well as toll revenue. (Figure 4) By doing so, the rate of return to private investors is increased, and as a result, it will become possible to lead private funds in various infrastructure projects. Figure 4 Injection of a fraction of tax revenues gained from spillover effect

18 64 M Nakahigashi, N Yoshino / Public Policy Review IV-3 Incentive mechanism of infrastructure operating entities In order to enhance efficiency and increase the rate of return in infrastructure development, it is necessary that the dividend for private investors varies tied to the revenue from those infrastructure projects, and it is necessary that infrastructure-operating entities make efforts to devise higher levels of income. Table 9, as an example, shows the payoff table by the presence or absence of efforts by investors and an infrastructure-operating entity. If neither the entity or investors make efforts, the entity gains 50 in revenue and investors receive the dividend income r. It is supposed that the entity is able to increase operating income by 100 through improvements in the salary system in the entity, such as by paying bonuses for staff of the entity tied to the increase in the revenue of the entity. Furthermore, investors make various efforts to maximize their dividends in order to try to obtain the dividend income by ar (a >1) through cost reduction and revenue increase from an infrastructure, such as through an increase in the number of highway turnoffs and in available cars. The lower right of the payoff table (Table 9) represents the revenue in the case in which both the entity and investors make various concerted efforts to increase its revenue and improve its service. This indicates that the income obtained in this case is higher than in the normal case. (The income of the entity increases from 50 to 100 and the income of investors is from r to ar.) In Table 9 we show that it is important to shape the institutional design of the dividend policy and salary system in an infrastructure-operating entity so that it leads to an improvement in the revenues of the entity and investors. In the public-private partnership, as described above, it is necessary to improve the efficiency of infrastructure projects through private funds and to introduce mechanisms in order to benefit the staff of an infrastructure-operating entity, through for example paying bonus for staff tied to the increase in profit. Table 9 Payoff table for infrastructure operating entity and investors

19 Policy Research Institute, Ministry of Finance, Japan, Public Policy Review, Vol.12, No.1, March V. Conclusion This paper analyzes whether the productivity effect of public capital, which is one of the economic effects of public capital, has changed using a trans-log production function, and discusses how one might achieve the goals of promoting the spending of private-sector money for the maintenance of infrastructure and of enhancing the economic effect of infrastructure while reducing government spending. In the secondary and tertiary industries, the productivity effect of public capital has been significantly positive throughout the estimated period and has been present in recent years. In the 2000s, although inter-regional allocation of public investment has changed, the structure of the regional differences in its productivity effect has not changed over time. When comparing the magnitude of its productivity effect at FY1990 to that at FY2010 in regard to the output elasticity of public capital, it can been seen that there are statistically different directions in some industries and regions, but marginal productivity has reduced in almost all regions. In order to promote spending of private funds in future infrastructure development, the following three measures are required: firstly, regarding essential infrastructure for residents with high risk, the government must inject a fund. Secondly, measures must be taken in order that there is a reflection in dividends by way of including the spillover effects of infrastructure, which are represented in the increase in tax revenue through economic activity based on said infrastructure development. Finally, in order to increase incentives for infrastructure-operating entities, incentive mechanisms must be introduced for their staff, for example by paying bonuses tied to increases in its profits. References Annala, C. N., R. G. Batina, and J. P. Feehan (2008), Empirical Impact of Public Infrastructure on the Japanese Economy, Japanese Economic Review, 59 (4), pp Cabinet Office Economic and Social Research Institute (2011), Prefectural Capital Stock of Private Enterprises (at CY2000 price, SNA-based, May 2011), go.jp/jp/sna/data/data_list/kenmin/files/contents/main_h21stock.html Cabinet Office Director General for Economic, Fiscal and Social Structure (2012), Public Capital Stock in Japan 2012, Cabinet Office PFI Promotion Office (2015), On the Current State of PFI (May, 2015), (in Japanese) Nakahigashi, M. (2003), Productivity Effect of Public Capital in Japan, Mitsubishi Economic Research Institute (in Japanese) Nakahigashi, M. (2015), Aging of Public Capital and the Productivity Effect, in N. Yoshino, K. Kameda, M. Nakahigashi and M. Nakata (ed) Challenges and Course of Japanese Economy: Theoretical and Empirical Analyses of Economic Policy, Keio Univer-

20 66 M Nakahigashi, N Yoshino / Public Policy Review sity Press (in Japanese) Hayashi, M. (2009), Productivity Effect of Public Capital: Reconsideration by the Dynamic Panel, Public Finance Studies, Vol. 5, pp (in Japanese) Miyagawa, T., K. Kawasaki and K. Edamura (2013), Re-examination of the Productivity Effect of Public Capital, RIETI Discussion Paper Series, 13-J-081 (in Japanese) Yoshino, N., T. Nakajima and M. Nakahigashi (1999), Productivity Effect of Public Capital, in N. Yoshino, T. Nakajima (ed) Economic Effect of Public Investment, Nihon Hyoron-sha (in Japanese) Yoshino, N. and H. Nakano (1994), Public Investment Allocations to the Tokyo Metropolitan Area, in T. Hatta (ed) Economic Analysis of Tokyo Over-concentration, Nihon Keizai Shimbun, Inc. (in Japanese) Miyara, I. and M. Fukushige (2008), Types of Public Capital and Their Productivity in Japanese Prefectures, Japanese Economic Review, 59 (2), pp OECD (2008), Public-Private Partnership: in Pursuit of Risk Sharing and Value for Money, OECD Publishing. Okubo, M. (2008), Public Capital and Productivity: a Nonstationary Panel Analysis, Applied Economics Letters, 15, pp Revankar, N. S. (1974), Some Finite Sample Results in the Context of Two Seemingly Unrelated Regression Equations, Journal of the American Statistical Association, 69 (345), pp Yoshino, N. (2010), Financing Transport Infrastructure Investment, OECD (ed.), Southeast Asian Economic Outlook 2010, OECD Publishing Yoshino, N. (2012), Global Imbalances and the Development of Capital Flows among Asian Countries, OECD Journal: Financial Market Trends, Vol. 2012/1, pp Yoshino, N. and F. Taghizadeh-Hesary (2015), Effectiveness of the Easing of Monetary Policy in Japanese Economy, Incorporating Energy Prices, Journal of Comparative Asian Development, 14(2), pp Yoshino, N. and F. Taghizadeh-Hesary (2016), Causes and Remedies of Japan s Long-lasting Recession: Lessons to China, China and World Economy, 24 (2), forthcoming Yoshino, N. and M. Nakahigashi (2004), The Role of Infrastructure in Economic Development, ICFAI Journal of Managerial Economics, 2, pp Yoshino, N. and V. Pontines (2015), The Highway Effect on Public Finance: Case of the STAR Highway in the Philippines, Asian Development Bank Institute (ADBI) Working Paper, No. 549 Yoshino, N. and U. Abidhadjaev (2015), An Impact Evaluation of Investment in Infrastructure: The Case of the Railway Connection in Uzbekistan, Asian Development Bank Institute (ADBI) Working Paper, No Appendix The appendix describes data on public capital stock, private capital stock, output, and la-

21 Policy Research Institute, Ministry of Finance, Japan, Public Policy Review, Vol.12, No.1, March bor share that are used in order to estimate the productivity effect of public investment. Public capital stock The public capital stock used in our paper is prefectural gross public capital stock, which is estimated by the Cabinet Office Director General for Economic, Fiscal and Social Structure (2012). Our public capital stock consists of roads, ports, airport facilities, sewerage, waste disposal, water supply, urban parks, school facilities, social education facilities, flood control, forest conservation, coastal, agriculture, forestry, fishing, national forests and industrial water supply. Private capital stock The private capital stock in our analysis is based on trial estimates published by Economic and Social Research Institute (ESRI), Cabinet Office Prefectural Private Capital Stock (at 2000 price, SNA-based, May 2011). Trial estimates by ESRI, however, have data discontinuities because the coverage of private enterprise companies has changed by the privatization of public enterprises. The amount of capital stock of privatized public enterprises at the beginning of their privatization can be derived through use of the data for capital stock, investment and net depreciation. In order to clear the problem of data discontinuities, we use backward estimation using the growth rate from the private capital stock at the end of FY2010. In particular, before and after the time when the public enterprise has privatized, we calculate the growth rate of private capital stock including privatized public enterprises as well as that excluding privatized public enterprises. And we assume that the growth rate of private capital stock remains unchanged before and after their privatization, the backward estimation can be held. Output Output by industry and region is real Gross Regional Product (GRP) calculated by dividing the nominal GRP by the GRP deflator. The nominal GRP is the GRP at producers prices excluding taxes on production and imports less subsidies. After FY2001, the GRP deflator by economic activity is directly used by Annual Report on Prefectural Accounts. Before FY2000, the deflator is regarded as the same in all prefectures for each industry, and then we use a GDP deflator by economic activity of Annual Report on National Accounts as the deflator. Note that all of the data is performed with backward estimation using the growth rate from FY2010. This is because there is a possibility that the definition of SNA of Japan is discontinuous for each reference year.

22 68 M Nakahigashi, N Yoshino / Public Policy Review Labor input Labor input is the product of working hours and the number of workers. The number of employed persons and the number of workers by prefecture are estimated using the Ministry of Internal Affairs and Communications Population Census, which is published every 5 years as a benchmark, and in the estimation of interpolate years, the Ministry of Internal Affairs and Communications Labour Force Survey and Population Census are used. 7 When estimating working hours it is assumed as taking the same value by industry in all regions. Working hours of the agriculture, forestry and fisheries industries are estimated using the Labour Force Survey. Working hours of other industries, on the other hand, are estimated using the Ministry of Health, Labour and Welfare Monthly Labor Statistics Survey. Labor Income of Workers Compensation of employees in the Annual Report on National Accounts is only the remuneration for employees. In this paper, the labor income of workers is estimated using the number of workers and workers wage per capita. When estimating employed income of agriculture, forestry and fisheries, it is assumed that although there are regional differences between the employee s income and other worker s income, the ratio of these is the same among regions, and then employed income is estimated using this ratio. On the other hand, when estimating labor income in other industries, it is assumed that labor income per capita of employees and workers is the same, then labor income is estimated by multiplying the number of workers by the ratio of the number of employees to the compensation of employees. Note that the estimation method to connect different compensation of employees is by backward estimation using the growth rate of compensation of employees from FY The number of workers and number of employees in our paper does not consider the number of workers with a sideline, unlike estimates of the SNA of Japan.