Effects of public investminvestment: A factor aug

Takao Fujii a,, Kazuki Hiraga b,yogo 6yo 10ces, 3

different categoriesstrate thattitative andpublic inv

reaps different benets in different sectors and that it ca

q An earlier version of this paper was presented at the 8th Modern Monetary Economics Summer Institute, the 68th annualconvention of the Japan Institute of Public Finance. We wish to thank Hiroshi Gunji, Kunihiro Hanabusa, Jun Iritani, KeigoKameda, Shinichi Kitasaka, Yoichi Matsubayashi, Ryuzo Miyao, Tomomi Miyazaki and Etsuro Shioji for their very careful readingof an earlier version of the this paper, and insightful suggestions. We also would like to thank an anonymous referee for manydetailed comments. Needless to say, all remaining errors are our own. Corresponding author.

E-mail addresses: takaofujii1@gmail.com (T. Fujii), khiraga581470@gmail.com (K. Hiraga), Masafumi_Kozuka@red.umds.ac.jp (M. Kozuka).

J. Japanese Int. Economies 27 (2013) 3547

Contents lists available at SciVerse ScienceDirect

Journal of The Japanese andInternational Economies

journal homepage: www.elsevier .com/locate/ j j ie0889-1583/$ - see front matter 2012 Elsevier Inc. All rights reserved.investment is crowded out or crowded in inof industrial investment. Our results demoninvestment confers different effects, both quantive, in individual sectors. This implies thathttp://dx.doi.org/10.1016/j.jjie.2012.11.003publicqualita-estmentn bringKeywords:Fiscal policyInvestmentCrowding-outCrowding-inFAVAR

sibility of crowd-in effects in some industries, such as industriesblessed with public demand. Meanwhile, public investmentdecreases sectoral private investment not only by keeping rentalcost high, but also by differences in the resource misallocationeffect of public investment itself; one sector receives a positivewealth effect while another suffers the opposite. In this paper weuse a factor-augmented VAR (FAVAR), a model capable of analyzinglarge-scale VAR models, to investigate the extent to which publicaKobe University, 2-1 Rokkodai, Nada, Kobe, HbKeio University, 2-15-45 Mita, Minato-ku, TokcUniversity of Marketing and Distribution Scien

a r t i c l e i n f o

Article history:Received 31 January 2012Revised 3 October 2012Available online 8 December 2012

JEL classication:E62H54H32ent on sectoral privatemented VAR approachq

Masafumi Kozuka c

57-8501, Japan8-8345, Japan-1 Gakuen-Nishimachi, Nishi-ku, Kobe, Hyogo 651-2188, Japan

a b s t r a c t

Fujii, Takao, Hiraga, Kazuki, and Kozuka, MasafumiEffects ofpublic investment on sectoral private investment: A factor aug-mented VAR approach

Public investment decreases aggregate private investment in bothneoclassical and Keynesian models. There are no ndings, however,on how public investment affects private investment on a disag-gregated basis, such as sectoral private investment. More speci-cally, previous research has neglected the distinctions of sectoralinvestment behavior in response to public investment and the pos-

the worse effect of resource misallocation on some sectors. J. Japa-nese Int. Economies 27 (2013) 3547. Kobe University, 2-1 Rokko-dai, Nada, Kobe, Hyogo 657-8501, Japan; Keio University, 2-15-45Mita, Minato-ku, Tokyo 108-8345, Japan; University of Marketing

criticized weakness of VAR, its lack of forward-looking components.Several papers have been published o

policy. When Nekarda and Ramey (20

1 Yoshikawa (1991), Miyagawa (1997, 2005) showother advanced countries.

2 In fact, Doi et al. (2011) conclude that the Japan3 Aschauer (1989a) shows that public investment

(1989b) investigates the crowd-in or crowd-out effe

36 T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547n sectoral capital investment behavior in response to scal11) compare which type of model, neoclassical or New

that the capital investment rate is higher and uctuates more in Japan than in

ese scal policy is not sustainable.complements private investment via social capital externality, and Aschauerand Distribution Sciences, 3-1 Gakuen-Nishimachi, Nishi-ku, Kobe,Hyogo 651-2188, Japan.

2012 Elsevier Inc. All rights reserved.

1. Introduction

Capital investment affects not only the supply side via capital accumulation in the long run, butalso the demand side in the short run. This is known as the duality of investment. The relatively highuctuation and instability of capital investment in Japan affects the business cycle.1

There are several papers which insist that capital investment is relevant to Japans lost decade.Hori et al. (2006) and Hayashi (2007), for example, implicate capital investment stagnation as oneof the main causes of this lost decade. Using data from the pre- and middle-bubble period(19801991) and post-bubble period (19912003), Horioka (2007) points to capital investmentstagnation as a principal cause of the lost decade and analyzes its relation to economic growth andthe contribution rates of GDP components. Numerous studies in capital investment in relation to scalpolicy, and to the crowding out problem, have long been developed since Smith (1776). Under thepresent circumstances, this is a critical issue. As stated by Fukuda et al. (2011), beginning inmid-2007, major countries in the world introduced aggressive scal policy in response to the globalnancial crisis. As a result, they ran huge budget decits. In addition, long term interest rates incountries such as Greece and Spain rose dramatically because of sovereign risk (see e.g., Laubach(2010)). Furthermore, though we cannot observe the increase in the long-term interest rate now,the situation in Japan is critical. As stated by Doi et al. (2011) and other papers, the Japanese publicdebt exceeds Italys and is currently the highest among the G7 countries.2 These circumstances suggestthe possibility of revealing the crowding-out problem.

Public investment crowds out aggregate private investment, following both the transitional neo-classical and Keynesian models. Voss (2002) uses a structural VAR model to show that governmentspending decreases private investment. In contrast, ndings from Aschauer (1989a,b) show the oppo-site effect.3 These papers fail, however, to analyze the transitional sectoral private investment, which re-sponds differently to additional public investment from sector to sector. The construction sector, forexample, often receives orders from government to construct public capital such as roadways, bridges,and dams, while the service industry ordinarily receives no direct orders from government at all. Differ-ences of this type cause differences of investment behavior.

In light of the above, this study investigates the dynamic response of sectoral capital investment topublic investment. Large-scale variables are analyzed by the factor-augmented VAR (FAVAR) model, adynamic factor model that allows us to analyze VAR models even when the independent variables aretoo numerous for the sample size. Estimations by FAVAR combine factor analysis with VAR containingfactor loadings and policy variables, which allows us to decrease the variables within the VAR and ob-tain the large-scale information necessary to introduce many independent variables. FAVAR can alsocontain variables that include expectations, such as stock prices, and thus compensate for the oftenct.

Keynesian, is more consistent with the use of industrial panel data, they conclude that the former isbetter. Hamaaki (2008) inspects the effects of Japanese tax reform on sectoral capital investment byestimating the capital investment function. None of these studies, however, succeed in analyzingthe dynamic behavior of capital investment.

Turning to the dynamic behavior of capital investment, we can cite Pereira (2001), Pereira and

behavpaper

Th

T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547 37VAR to analyze the effects of monetary policy shock in data-rich environments. Only a few studies,however, have used FAVAR to analyze the effects of scal policy, and to our knowledge none have uti-lized FAVAR to analyze sectoral investment. In this paper, we perform the rst attempt to empiricallyanalyze Japanese scal policy with FAVAR, and investigate the relationship between governmentspending and sectoral investment.

The FAVARmodel allows us to obtain the dynamic response of sectoral capital investment to publicinvestment and yields markedly different quantitative and qualitative results. We can cite four rea-sons for this. First, the difference arises from the persistence of public investment. Second, differencesin production externalities alter the dynamic response of sectoral capital investment to public invest-ment, both quantitatively and qualitatively. Third, the ripple effect of public investment differs fromsector to sector, as mentioned before.4 Lastly, FAVAR model enables us to consider comprehensive ef-fects of shocks; such as intra-sectoral effect to public investment shock because FAVAR model can in-volve both macroeconomics variables the policymaker observes and all of sectoral variables.5

The remainder of this paper is structured as follows. In Section 2 we describe the FAVAR modelused for the estimation. In Section 3 we show the estimation results and give interpretations. In Sec-tion 4 we summarize our ndings and conclude.

2. FAVAR model

In this section we explain the econometric framework for the factor-augmented VAR (FAVAR).6 LetYt be an M 1 vector of observable economic variables, where M is small. While Yt is used in a standardVAR, Yt alone does not easily provide additional economic information. We therefore assume that a K 1vector of unobserved factors, where K is small, summarizes this additional information. The joint dynam-ics of (Ft,Yt) are given by the following equation:

FtYt

UL Ft1

Yt1

ut 1

whereU(L) is a matrix of polynomials of nite order d and the error term ut is the mean 0 with covari-ance matrix R.

Eq. (1) is a factor-augmented VAR (FAVAR). There seems to be little difference here between a stan-dard VAR and the FAVAR. Yet Eq. (1) is impossible to estimate because the factors are unobservable.We must therefore assume that the factors affect a large number of variables, in order to estimate Eq.(1). This assumption allows us to infer the factors from these economic time series variables. Let Xt be

4 As a further reason in this regard, rising wages for public servants alter the prot and investment of private rms, as discussedby Alesina et al. (2002).

5 Annala et al. (2008) investigate the sectoral effect to public investment (infrastracture) shock using VECM, which is divided byeach sector. Therefore, ABF cannot consider the comprehensive effects with their method.

6 This section conforms with Bernanke et al. (2005), Shibamoto (2007) and Vargas-Silva (2008). For details, see Bernanke et al.

(2005).ior of aggregate capital investment in response to public investment using a FAVAR model. Thiscombines these two approaches.e FAVAR model was rst proposed by Bernanke et al. (2005). Many papers since have used FA-Andraz (2003), Annala et al. (2008)(ABF) and Forni and Gambetti (2010). Pereira Alfredo M. (2001)and Pereira and Andraz (2003) analyze the dynamic behavior of sectoral capital investment inresponse to public investment using a VAR model containing only those variables. ABF investigatethe effects of public infrastructure, which is nearly equal to public investment, on sectoral output,private capital, and employment using vector error correction model (VECM) in Japan. ABF reportsboth crowding-in-and-out effect of private capital. Forni and Gambetti (2010) analyze the dynamic

an N 1 vector of informational economic variables where N is large, such that K +M N.7 Also, as-sume that Xt is related to both the unobservable factors vector Ft and the observable factors vector Yt,given as follows:

Xt Kf Ft KyYt et 2where Kf, Ky are the N K, N M matrix of factor loadings, respectively, and et is an N 1 vector oferror terms which is weakly correlated with the mean 0.

Fir

While Bernanke et al. (2005) estimate the FAVAR by both the two-stage approach and a Bayesian method based on Gibbssamplin

9 As Lagana and Sgro (2011) point out, quarterly data are more likely to capture the effects of any scal policy change. We

38 T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547therefore use quarterly data.10 Numbers in parentheses imply the numbers of variables in Table 1.11 In most cases, our empirical results show that impulse responses are not signicant. Yet most previous studies also show thatestimated values are insignicant and discuss the following point estimate. In this paper we do not present the standard error band

and weg, they suggest that the two-stage approach tends to produce more plausible responses.position. Data on ve variables are examined: public investment, nominal long-term interest rate(114), private investment (135), real GDP (120) and tax revenue (130).10 The empirical results for vevariables are compared with those for four variables, excluding tax revenue. Then the results are veriedover the sample period from 1983:Q2 to 2008:Q1. The logarithms for all variables but the interest rateare calculated and multiplied by 100. The length of lags is 4. The error term is orthogonalized by Choleskidecomposition. Fig. 1 presents the impulse response functions obtained from these settings.11 These re-sults show increases of the long-run interest rate and decreases of investment, movements that togetherindicate a crowding out of investment. We can therefore corroborate the existence of the crowding-outeffect demonstrated in earlier reports on the Japanese macroeconomy.

7 As Bernanke et al. (2005) point out, it is acceptable for N to be greater than T.8st, we perform the empirical analyses using the standard structural VAR with Choleski decom-For the estimation procedure we follow the two-stage approach proposed by Bernanke et al.(2005).8 This means that we identify Ft in the rst stage and estimate Eq. (1) in the second stage. Spe-cically, we perform the following steps in the rst stage. To begin, the common components, Ct, are esti-mated using the rst K +M principal components of Xt. Second, following the example from Bernankeet al. (2005), variables are classied into two categories, namely, slow-moving and fast-moving.Slow-moving variables are those predetermined in the current period, such as output and employment.Fast-moving variables are those sensitive to contemporaneous economic news or shocks, such as assetprices. Next, a principal component analysis is applied to the slow-moving variables in order to get a vec-tor of slow-moving factors, FSt . Lastly, the following regression is estimated,

bCt bFS bFSt bYYt et 3and the estimated factors, bF t , are obtained from bCt bYYt . In the second stage we estimate the VAR inbFt and Yt and compute the impulse response function using a Choleski decomposition.

In this paper we use the FAVAR model to analyze how public investment impacts each type ofindustrial investment. The dataset used to estimate the factors contains 137 quarterly macroeconomictime series for the period from 1983Q2 to 2008Q1 (such that T = 100).9 The selected series basicallyfollow those in Shibamoto (2007). These series are transformed to induce stationarity, and the season-ality is removed by Census X12 ARIMA for all series except the interest rate. The data are taken fromthe Nikkei NEEDS database. Public investment is assumed to be the only variable included in the vectorof observable variables, Yt. The VAR includes four lags of each variable, and the estimation is conductedusing two or three factors. Table 1 lists all the data series included in the dataset and the classicationinto two categories of variables, slow-moving and fast-moving.

3. Empirical analyses

3.1. Empirical analyses by standard structural VARperform the verication with the point estimate.

Table 1Data.

Number Data name Period Code

Slow moving variables1 Industrial Production Index (Mining and Manufacturing, 2005 = 100) 1983Q2-2008Q1 52 Industrial Production Index (Construction Goods, 2005 = 100) 1983Q2-2008Q1 53 Industrial Production Index (Capital Goods, 2005 = 100) 1983Q2-2008Q1 54 Industrial Production Index (Durable Consumer Goods, 2005 = 100) 1983Q2-2008Q1 55 Industrial Production Index (Nondurable Consumer Goods, 2005 = 100) 1983Q2-2008Q1 56 Industrial Production Index (Consumer Goods, 2005 = 100) 1983Q2-2008Q1 57 Industrial Production Index (Final Demand Goods, 2005 = 100) 1983Q2-2008Q1 58 Industrial Production Index (Investment Goods, 2005 = 100) 1983Q2-2008Q1 59 Industrial Production Index (Producer Goods, 2005 = 100) 1983Q2-2008Q1 510 Producers Shipment Index (Mining and Manufacturing, 2005 = 100) 1983Q2-2008Q1 511 Producers Shipment Index (Construction Goods, 2005 = 100) 1983Q2-2008Q1 512 Producers Shipment Index (Capital Goods, 2005 = 100) 1983Q2-2008Q1 513 Producers Shipment Index (Durable Consumer Goods, 2005 = 100) 1983Q2-2008Q1 514 Producers Shipment Index (Nondurable Consumer Goods, 2005 = 100) 1983Q2-2008Q1 515 Producers Shipment Index (Consumer Goods, 2005 = 100) 1983Q2-2008Q1 516 Producers Shipment Index (Final Demand Goods, 2005 = 100) 1983Q2-2008Q1 517 Producers Shipment Index (Investment Goods, 2005 = 100) 1983Q2-2008Q1 518 Producers Shipment Index (Producer Goods, 2005 = 100) 1983Q2-2008Q1 519 Capacity Utilization Index (Machinery, 2005 = 100) 1983Q2-2008Q1 520 Capacity Utilization Index (Chemicals, 2005 = 100) 1983Q2-2008Q1 521 Capacity Utilization Index (Ceramics, Clay and Stone Products, 2005 = 100) 1983Q2-2008Q1 522 Capacity Utilization Index (Electrical Machinery, 2005 = 100) 1983Q2-2008Q1 523 Capacity Utilization Index (Fabricated Metals, 2005 = 100) 1983Q2-2008Q1 524 Capacity Utilization Index (General Machinery, 2005 = 100) 1983Q2-2008Q1 525 Capacity Utilization Index (Manufacturing, 2005 = 100) 1983Q2-2008Q1 526 Capacity Utilization Index (Nonferrous Metals, 2005 = 100) 1983Q2-2008Q1 527 Capacity Utilization Index (Pulp, Paper and Paper Products, 2005 = 100) 1983Q2-2008Q1 528 Capacity Utilization Index (Transport Equipment, 2005 = 100) 1983Q2-2008Q1 529 Capacity Utilization Index (Textiles, 2005 = 100) 1983Q2-2008Q1 530 Department Store Sales (per Square Meter, 10,000 yen) 1983Q2-2008Q1 531 Department Store Sales (per Worker, 10,000 yen) 1983Q2-2008Q1 532 Large-Scale Retail Store Sales (million yen) 1983Q2-2008Q1 533 Index of Wholesale Price in Small and Medium Sized Enterprises

(2005 = 100)1983Q2-2008Q1 5

34 Index of Sales in Small and Medium Sized Enterprises (Manufacturing,2005 = 100)

1983Q2-2008Q1 5

35 Index of Shipment in Small and Medium Sized Enterprises (2005 = 100) 1983Q2-2008Q1 536 Index of Total Worked Hours (All Industries, 30 or More Persons,

2005 = 100)1983Q2-2008Q1 5

37 Household Disposable Income (yen) 1983Q2-2008Q1 538 Electric Power Consumption of Large Users (million KWH) 1983Q2-2008Q1 539 New Job Offers (person) 1983Q2-2008Q1 540 New Job Offers (Part-Time, person) 1983Q2-2008Q1 541 Consumption Expenditure (yen) 1983Q2-2008Q1 542 Consumption Expenditure (Food, yen) 1983Q2-2008Q1 543 Employment Index of Regular Workers (All Industries, 30 Employees or

more, 2005 = 100)1983Q2-2008Q1 5

44 Unemployment Rate (%) 1983Q2-2008Q1 545 Consumer Price Index (Clothes and Footwear) 1983Q2-2008Q1 546 Consumer Price Index (Food) 1983Q2-2008Q1 547 Consumer Price Index (Fuel Light and Water Charges) 1983Q2-2008Q1 548 Consumer Price Index (General) 1983Q2-2008Q1 549 Consumer Price Index (General, Exclude Fresh Food) 1983Q2-2008Q1 550 Consumer Price Index (General, Exclude Imputed Rent) 1983Q2-2008Q1 551 Consumer Price Index (Furniture and Household Utensils) 1983Q2-2008Q1 552 Consumer Price Index (Miscellaneous) 1983Q2-2008Q1 553 Consumer Price Index (Reading and Recreation) 1983Q2-2008Q1 554 Consumer Price Index (Transportation and Communication) 1983Q2-2008Q1 5

(continued on next page)

T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547 39

Table 1 (continued)

Number Data name Period Code

55 Corporate Goods Price Index (Manufacturing Industry Products) 1983Q2-2008Q1 556 Corporate Goods Price Index (Mineral Produce) 1983Q2-2008Q1 557 Corporate Goods Price Index (All Commodities) 1983Q2-2008Q1 558 Real Wage Index (Contractual Cash Earnings in All Industries), 2005 = 100 1983Q2-2008Q1 559 Real Wage Index (Contractual Cash Earnings in Manufacturing, 2005 = 100) 1983Q2-2008Q1 560 Wage Index (Contractual Cash Earnings in Manufacturing, 2005 = 100) 1983Q2-2008Q1 561 Wage Index (Contractual Cash Earnings in All Industries, 2005 = 100) 1983Q2-2008Q1 562 Import Volume Index (Total) 1983Q2-2008Q1 563 Export Volume Index (Total) 1983Q2-2008Q1 564 Value of Exports, Customs Clearance Basis (million) 1983Q2-2008Q1 5

Fast moving variables65 Total Floor Area of Building Construction Started (Grand Total, 1000 m2) 1983Q2-2008Q1 566 Total Floor Area of New Housing Construction Started (Built for Sale,

1000 m2)1983Q2-2008Q1 5

67 Total Floor Area of New Housing Construction Started (Owned, 1000 m2) 1983Q2-2008Q1 568 Total Floor Area of New Housing Construction Started (Total, 1000 m2) 1983Q2-2008Q1 569 Total Number of New Housing Constuction Started (Built for sale) 1983Q2-2008Q1 570 Total Number of New Housing Constuction Started (Owned) 1983Q2-2008Q1 571 Total Number of New Housing Constuction Started (Rented) 1983Q2-2008Q1 572 Total Number of New Housing Constuction Started (Total) 1983Q2-2008Q1 573 Amount of Clearing (Value: million yen) 1983Q2-2008Q1 574 Amount of Clearing (Number: thousand bills) 1983Q2-2008Q1 575 Nikkei Stock Average 225 Selecteed Stocks (Average of Month, yen) 1983Q2-2008Q1 576 Nikkei Stock Average 500 Selecteed Stocks (Average of Month, yen) 1983Q2-2008Q1 577 Arithmetic Stock Price Average (First Section of the Tokyo Stock Exchange) 1983Q2-2008Q1 578 Tokyo Stock Price Index (TOPIX) 1983Q2-2008Q1 579 Foreign Exchange Rate (Yen per US Dollar) 1983Q2-2008Q1 580 Foreign Effective Exchange Rate (Nominal, 2005 = 100) 1983Q2-2008Q1 581 Money Stock (M1, Average, 100 million yen) 1983Q2-2008Q1 582 Money Stock (M2, Average, 100 million yen) 1983Q2-2008Q1 583 Monetary Base (Average, 100 million yen) 1983Q2-2008Q1 584 Bank of Japan Accounts, Assets, Loans (100 million yen) 1983Q2-2008Q1 585 Banking Accounts, City Banks, Assets, Cash and Due from Banks (100

million yen)1983Q2-2008Q1 5

86 Banking Accounts, Regional Banks, Assets, Cash and Due from Banks (100million yen)

1983Q2-2008Q1 5

87 Banking Accounts, City Banks, Assets, Call Loans (100 million yen) 1983Q2-2008Q1 588 Banking Accounts, Regional Banks, Assets, Call Loans (100 million yen) 1983Q2-2008Q1 589 Banking Accounts, City Banks, Assets, Loans and Bills Discounted (100

million yen)1983Q2-2008Q1 5

90 Banking Accounts, Regional Banks, Assets, Loans and Bills Discounted (100million yen)

1983Q2-2008Q1 5

91 Banking Accounts, City Banks, Assets, Investment Securities (100 millionyen)

1983Q2-2008Q1 5

92 Banking Accounts, Regional Banks, Assets, Investment Securities (100million yen)

1983Q2-2008Q1 5

93 Banking Accounts, City Banks, Liabilities, Borrowed Money (100 millionyen)

1983Q2-2008Q1 5

94 Banking Accounts, Regional Banks, Liabilities, Borrowed Money (100million yen)

1983Q2-2008Q1 5

95 Banking Accounts, City Banks, Liabilities, Call Money (100 million yen) 1983Q2-2008Q1 596 Banking Accounts, City Banks, Liabilities, Deposits (100 million yen) 1983Q2-2008Q1 597 Banking Accounts, City Banks, Liabilities, Negotiable Certicates of Deposits

(100 million yen)1983Q2-2008Q1 5

98 Banking Accounts, Regional Banks, Liabilities, Negotiable Certicates ofDeposits (100 million yen)

1983Q2-2008Q1 5

99 Banking Accounts, Regional Banks, Liabilities, Deposits (100 million yen) 1983Q2-2008Q1 5100 Index of Industrial Inventories (Mining and Manufacturing, 2005 = 100) 1983Q2-2008Q1 5101 Index of Industrial Inventories (Construction Goods, 2005 = 100) 1983Q2-2008Q1 5102 Index of Industrial Inventories (Capital Goods, 2005 = 100) 1983Q2-2008Q1 5

40 T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547

Table 1 (continued)

T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547 41Number Data name Period Code

103 Index of Industrial Inventories (Durable Consumer Goods, 2005 = 100) 1983Q2-2008Q1 5104 Index of Industrial Inventories (Nondurable Consumer Goods, 2005 = 100) 1983Q2-2008Q1 5105 Index of Industrial Inventories (Consumer Goods, 2005 = 100) 1983Q2-2008Q1 5106 Index of Industrial Inventories (Final Demand Goods, 2005 = 100) 1983Q2-2008Q1 5107 Index of Industrial Inventories (Investment Goods, 2005 = 100) 1983Q2-2008Q1 5108 Index of Industrial Inventories (Producer Goods, 2005 = 100) 1983Q2-2008Q1 5109 The Basic Discount Rate and Basic Loan Rate (%) 1983Q2-2008Q1 5110 Yield of Interest-Bearing Government Bonds (10 years, %) 1983Q2-2008Q1 1111 Yield of Government Guaranteed Bonds (10 years, %) 1983Q2-2008Q1 1112 Yield of Local Government Bonds (10 years, %) 1983Q2-2008Q1 1113 Yield of Interest-Bearing Bank Debentures (5 years, %) 1983Q2-2008Q1 1114 Yield of Government Bonds (10 years, %) 1983Q2-2008Q1 1115 Tokyo Interbank Offered Rate (TIBOR) (3 months, %) 1983Q2-2008Q1 1116 Deposit Rates of Postal Savings, Ordinary Savings (%) 1983Q2-2008Q1 1117 Spread between Long and short term interest rates (%) 1983Q2-2008Q1 1118 Call Rate (Collateralized Overnight, Month Average, %) 1983Q2-2008Q1 1Our priority, however, is to investigate the response of investment in each industry. To this end, thenext subsection describes our empirical analyses by the FAVAR model.

3.2. Estimations by FAVAR

Figs. 25 show the results of estimations with the FAVAR model we describe in the previous sec-tion.12 To conrm robustness we investigate two cases, an estimation with two factors and another withthree factors. Here, the shock of the public investment implies a disturbance of growth by 1%. First, inFigs. 2 and 3, we present the case with private investment, as has been done in the earlier studies. Figs. 2and 3 both show the existence of a crowding-out effect in every case: the interest rate increases andinvestment decreases. And the response of GDP is positive. These results imply that the multiplier effectsdecrease, which is consistent with the results of previous papers.

Next we consider the response of private investment in each industry, as plotted out in Figs. 4 and5. Here we nd that the response of the mining industry is unstable, with alternations in the number

119 Government Fixed Capital Formation (billion yen) 1983Q2-2008Q1 5120 GDP (billion yen) 1983Q2-2008Q1 5121 Real New Investment, Agriculture, Forestry and Fisheries (billion yen) 1983Q2-2008Q1 5122 Real New Investment, Mining (billion yen) 1983Q2-2008Q1 5123 Real New Investment, Construction (billion yen) 1983Q2-2008Q1 5124 Real New Investment, Manufacturing (billion yen) 1983Q2-2008Q1 5125 Real New Investment, Wholesale and Retail Trade (billion yen) 1983Q2-2008Q1 5126 Real New Investment, Real Estate (billion yen) 1983Q2-2008Q1 5127 Real New Investment, Transport and Communications (billion yen) 1983Q2-2008Q1 5128 Real New Investment, Electricity, Gas and Water Supply (billion yen) 1983Q2-2008Q1 5129 Real New Investment, Service activities (billion yen) 1983Q2-2008Q1 5130 Real Tax and Stamp Revenue (billion yen) 1983Q2-2008Q1 5131 Real Effective Exchange Rate Index (2005 = 100) 1983Q2-2008Q1 5132 Real Long-term Interest Rates 1983Q2-2008Q1 1133 Real Short-term Interest Rates 1983Q2-2008Q1 1134 Public Debt 1983Q2-2008Q1 5135 Gross Private Fixed Capital Formation 1983Q2-2008Q1 5136 Gross Private Fixed Capital Formation (Housing) 1983Q2-2008Q1 5137 Gross Private Fixed Capital Formation (Facilities) 1983Q2-2008Q1 5

The selected series basically follow those in Shibamoto (2007). The last column represents the transformation code followingBernanke et al. (2005). Variables from 121 to 129 are represented by including the construction in progress and at market pricesin calendar year of 2000.

12 We refer to the Matlab code opened in Bernanke et al. (2005) and modify this code when estimating the FAVAR model.

long-term interest rate

.3

.4

.5

.6

.7

.8GDP

.10

.15

.20

.25Private Investment

-.4

-.3

-.2

-.1

.0

42 T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547of factors. The results for all of the other industries, however, are robust. This shows an observablecrowding-out effect in most industries, and the effect is positive in agriculture and utilities. Thus,we can assert that the effect of public investment on private investment differs from industry toindustry not only in quantity, but also in quantity. Compared with the result of Annala et al.(2008), utility, construction, estate and agriculture sectors are consistent with the result of them.On the other hand, mining, manufacturing, transport and communications and wholesale and retailtrade sectors are not consistent with their result. It is thought that these differences ascribe to the dif-

-.1.0.1.2

1 2 3 4 5 6 7 8 9 10-.05

.00

.05

1 2 3 4 5 6 7 8 9 10-.7

-.6

-.5

1 2 3 4 5 6 7 8 9 10

-.1.0.1.2.3.4.5.6.7.8

1 2 3 4 5 6 7 8 9 10-.05

.00

.05

.10

.15

.20

.25

1 2 3 4 5 6 7 8 9 10-.7

-.6

-.5

-.4

-.3

-.2

-.1

.0

1 2 3 4 5 6 7 8 9 10

Fig. 1. Impulse responses in the standard VAR model.ferences in sample period and econometric method.13

Next, we theoretically consider the causes of the differences in the effects of government expendi-ture on private investment.

On the crowding-out effect, the positive demand shock caused by public investment can be seen toraise the interest rate and lower the level of private investment. This mechanism is mentioned innumerous papers on Keynesian and neo-classical models.

On the other hand, our data also support three possible mechanisms of crowding-in. The rst ofthese factors is the extreme persistence of shock brought about by public investment. As shown inBaxter and King (1993) and Campbell (1994), the shock of public investment (or government expen-diture) leads to a negative income effect, decreasing both consumption and leisure. At the same time,labor increases. When the shock of public investment is more persistent, people recognize the morepersistent negative income effect and supply their labor for longer periods of time, and productionalso increases. Here, the intertemporal elasticity of substitution on the labor supply must be largerthan 1. The persistent negative income effect discourages people from consuming, and the higher sav-ings gained from non-consumption ow into private investment.14 Thus, public investment crowds-inthe private investment. From the empirical results we obtain in this study, we can regard the shock ofpublic investment as persistent. The views of previous studies are not in total accord, however, with re-

13 Annala et al. (2008) uses the data from 1970 to 1998. On the other hand, we employ the data from 1983 to 2008. The data weutilize include the period of Japanese long-run recession.14 The details are presented in Table 6 of Campbell (1994). Campbell (1994) shows that initial private investment is crowded inunder following conditions: (1) existence of a large response of labor supply to the negative income effect, and (2) existence ofpersistent shock of government expenditure.

Public Investment1.04

T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547 430.950.960.970.980.991.001.011.021.03

1 2 3 4 5 6 7 8 9 10

long-term interest rategard to the scale of the intertemporal elasticity of the labor supply. As such, we cannot support thishypothesis here.

The second point is the external effect on production by social stock. Baxter and King (1993) devel-op an RBC (Real Business Cycle) model considering the effect of social stock on production.15 Theirmodel explains that the shock of public investment crowds-in private investment with the large exter-nality of social stock. In their model, people forecast the positive shock of production, observing the accu-mulation of social stock. And to catch up with high productivity in the future, they increase theirinvestment.

.020

.022

.024

.026

.028

.030

.032

1 2 3 4 5 6 7 8 9 10

GDP

.015

.016

.017

.018

.019

.020

.021

1 2 3 4 5 6 7 8 9 10

Private Investment

-.160-.156-.152-.148-.144-.140-.136-.132-.128-.124

1 2 3 4 5 6 7 8 9 10

Fig. 2. Impulse responses in the FAVAR model with two factors.

15 They employ the production function inuenced by the external effect of social stock. This specication is developed inAschauer (1989a).

Public Investment

0.950.960.970.980.991.001.011.021.031.04

1 2 3 4 5 6 7 8 9 10

long-term interest rate

44 T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547The third point is the existence of private rms producing public goods. It is important that mostpublic investment is carried out by private rms. The increase of public investment leads to demandshock, and the rms producing public goods increase their investment in order to cope with the addi-tional demand.

We can therefore reach the following conclusions. The effect of the public investment on the pri-vate investment of each industry is inuenced by the interest rate elasticity of investment demand,the scale of the external effect brought by social stock, and the effect of demand shock on rmsproducing public goods. With the larger effect of interest rate elasticity, public investment has acrowding-out effect. In this case, the increase of capital cost brought by rising interest rates for publicinvestment has a larger effect. And when the external effect of social stock is larger, investment spend-ing is crowded in. In this case, the external effect of social stock is larger and the increased demand

.000

.004

.008

.012

.016

.020

.024

.028

.032

1 2 3 4 5 6 7 8 9 10

GDP

.008

.010

.012

.014

.016

.018

.020

.022

1 2 3 4 5 6 7 8 9 10

Private Investment

-.17

-.16

-.15

-.14

-.13

-.12

1 2 3 4 5 6 7 8 9 10

Fig. 3. Impulse responses in the FAVAR model with three factors.

long-term interest rate.032.028

.030Public Investment

0.950.960.970.980.991.001.011.021.031.04

1 2 3 4 5 6 7 8 9 10

T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547 45promotes public investment. The construction, wholesale, and trans-communication sectors come un-der the former case; the agriculture and utility sectors come under the latter.

4. Conclusion

Private investment is a major source of economic growth. At the same time, it uctuates more thanany other component on the demand side. The effect of public investment on private investment isthus important to investigate, as the governments ability to stabilize or stimulate the private invest-ment is one of the most important issues to know. The effects of sectoral capital investment are espe-cially divergent from industry to industry, with different policy implications. Here, therefore, we haveused the FAVAR model to investigate the extent to which public investment affects sectoral capitalinvestment. FAVAR allows us not only to analyze large-size VAR models, but also to consider variablesconnoting forward-looking components (expectations), such as stock prices.

.020

.022

.024

.026

1 2 3 4 5 6 7 8 9 10

Sectoral Private Investment

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

1 2 3 4 5 6 7 8 9 10

Fig. 4. Impulse responses in the FAVAR model with two factors.

0.96

0.95

1 2 3 4 5 6 7 8 9 10

long-term interest rate

.000

.004

.008

.012

.016

.020

.024

.028

.032Public Investment

0.970.980.991.001.011.021.031.04

46 T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547As a result, public investment crowds out aggregate investment. We can also, however, obtain re-sults indicative of both crowd-out and crowd-in effects on sectoral capital investment. This impliesthat public investment has differential effects in each industry, such as productive externality of pub-lic capital consisting of accumulating public investment and differences in wealth or demand shock.

This research can be expanded and elaborated in two possible directions. First, the selection offast-moving or slow-moving variables may alter the results. In this paper we consider the sectoralcapital investment as a control variable with good control in the current period. Second, we can alsoanalyze other government policy effects, such as the effects of regional output and investment onsome forms of regional public investment.

References

Alesina, Alberto., Ardagna, Silvia., Perotti, Enrico., Schiantarelli, Fabio., 2002. Fiscal policy, prots, and investment. AmericanEconomic Review 92 (3), 571589.

Annala, Christopher N., Batina, Raymond G., Feehan, James P., 2008. Empirical impact of public infrasructure on the Japaneseeconomy. The Japanese Economic Review 59 (4), 419437.

Aschauer, David., 1989a. Is public expenditure productive? Journal of Monetary Economics 23, 177200.Aschauer, David., 1989b. Does public capital crowd out private capital? Journal of Monetary Economics 24, 171188.

1 2 3 4 5 6 7 8 9 10

Sectoral Private Investment

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

1 2 3 4 5 6 7 8 9 10

Fig. 5. Impulse responses in the FAVAR model with three factors.

Baxter, Marianne, King, Robert, 1993. Fiscal policy in general equilibrium. American Economic Review 83 (3), 315334.Bernanke, Ben S., Boivin, Jean., Eliasz, Piotr S., 2005. Measuring the effects of monetary policy: a factor-augmented vector

autoregressive (FAVAR) approach. The Quarterly Journal of Economics 120 (1), 387422.Campbell, John Y., 1994. Inspecting the mechanism: an analytical approach to the stochastic growth model. Journal of Monetary

Economics 33 (3), 463506.Doi, Takero, Hoshi, Takeo, Okimoto, Tatsuyoshi, 2011. Japanese government debt and sustainability of scal policy. Journal of

the Japanese and International Economies 25 (4), 414433.Forni Mario, Gambetti Luca, 2010. Fiscal Foresight and the Effects of Government Spending. CEPR Discussion Papers 7840.Fukuda, Shin-ichi., Hoshi, Takeo., Leeper, Eric., 2011. Fiscal policy and crisis: editors introduction. Journal of the Japanese and

International Economies 25, 355357.Hamaaki, Junya., 2008. Investment responses to Japanese tax reforms: a cross-industry comparison. Japan and the World

Economy 20 (4), 542562.Hayashi, Fumio, 2007. General statement. In: Hayashi, Fumio (Ed.), The Source and System of the Economic Stagnation. Keiso

Shobo, pp. 318 (in Japanese).Hori, Keiichi, Saito, Makoto, Ando, Koichi, 2006. What caused xed investment to stagnate during the 1990s in Japan? Evidence

from panel data of listed companies. The Japanese Economic Review 57 (2), 283306.Horioka Charles Yuji, 2007. Japanese economy and macroeconomics. In: Hayashi Fumio (Ed.), The Source and System of the

Economic Stagnation. Keiso Shobo, pp. 1936 (in Japanese).Lagana, Gianluca, Sgro, Pasquale M., 2011. A factor-augmented VAR approach: the effect of a rise in the US personal income tax

rate on the US and Canada. Economic Modelling 28 (3), 11631169.Laubach, Thomas, 2010. Fiscal policy and interest rates: the role of sovereign default risk. In: NBER International Seminar on

Macroeconomics 2010. National Bureau of Economic Research, Inc., pp. 729.Miyagawa, Tsutomu., 1997. The development of investment theory and empirical analysis. In: Asako, Kazumi, Masayuki, Otaki.

(Eds.), Macroeconomic Dynamics. University of Tokyoyo Press, pp. 283322 (in Japanese).Miyagawa, Tsutomu, 2005. Stagnation Economics. University of Tokyoyo Press (in Japanese).Nekarda, Christopher J., Ramey, Valerie Ann, 2011. Industry evidence on the effects of government spending. American

Economic Journal: Macroeconomics 3 (1), 3659.Pereira Alfredo M., 2001. On the effects of public investment on private investment: what crowds in what? Public Finance

Review 29 (1), 325.

T. Fujii et al. / J. Japanese Int. Economies 27 (2013) 3547 47Pereira, Alfredo M., Andraz, Jorge Miguel, 2003. On the impact of public investment on the performance of US industries. PublicFinance Review 31 (1), 6690.

Shibamoto, Masahiko, 2007. An analysis of monetary policy shocks in Japan: a factor augmented vector autoregressiveapproach. The Japanese Economic Review 58 (4), 484503.

Smith, Adam, 1776. The Wealth of Nations. In: Strahan , Cadell (Eds.), Bantam Classics. London, New title edition, 2003.Vargas-Silva, Carlos., 2008. The effect of monetary policy on housing: a factor-augmented vector autoregression (FAVAR)

approach. Applied Economics Letters 15 (10), 749752.Voss, Graham M., 2002. Public and private investment in the United States and Canada. Economic Modelling 19, 641664.Yoshikawa Hiroshi, 1991. Japanese Economy and Macroeconomics. Toyo Keizai Shinposha.

Effects of public investment on sectoral private investment: A factor augmented VAR approach1 Introduction2 FAVAR model3 Empirical analyses3.1 Empirical analyses by standard structural VAR3.2 Estimations by FAVAR

4 ConclusionReferences