Permanent and transitory oil volatility and aggregate investment in Malaysia

Energy Policy 67 (2014) 552–563

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

Permanent and transitory oil volatility and aggregate investment in Malaysia Mansor H. Ibrahim a,n, Huson Joher Ali Ahmed b a b

International Centre for Education in Islamic Finance (INCEIF), Lorong Universiti A, 59100 Kuala Lumpur, Malaysia School of Accounting, Economics and Finance; Deakin University, Victoria, Australia

H I G H L I G H T S








Examines the role of oil volatility in Malaysia's aggregate investment. Makes distinction between permanent and temporary volatility using CGARCH. Both volatility components depress investment. Permanent volatility has larger adverse effects. Results are robust to alternative model specifications.

art ic l e i nf o

a b s t r a c t

Article history: Received 13 August 2013 Received in revised form 27 November 2013 Accepted 28 November 2013 Available online 21 December 2013

This paper investigates the relation between aggregate investment and oil volatility and its permanent and transitory components for a developing country, Malaysia. In the paper, the components generalized autoregressive conditional heteroskedasticity (CGARCH) model is utilized to decompose conditional oil volatility into permanent oil volatility and transitory oil volatility. Respectively reflecting fundamental-driven and random shifts in oil volatility, they are expected to exert differential effects on aggregate investment. Adopting a vector autoregression (VAR) framework to allow feedback effects between aggregate investment and its determinants, the paper documents evidence supporting the adverse effects of conditional oil volatility, permanent oil volatility and transitory oil volatility on aggregate investment and real output. Interestingly, contrary to the findings for the developed markets (US and OECD), the real effects of permanent oil volatility tend to be stronger. These findings are reasonably robust to variable specification and measurements in the VAR system. Hence, there is an indication that heightened oil volatility accounts for the slumps in Malaysia's aggregate investment after the Asian financial crisis. & 2013 Elsevier Ltd. All rights reserved.

Keywords: Oil volatility Aggregate investment Malaysia

1. Introduction The importance of domestic investment in a macroeconomy is well noted not only as a driver for long term economic growth but also as a main factor accounting for aggregate fluctuations. Like several other Asian countries, Malaysia witnessed a sudden drop in various spheres of its domestic activities during the 1997/1998 Asian crisis, among which include real output, domestic consumption and domestic investment. After recording a decade-long of miraculous growth from 1987 to 1997, Malaysia's real GDP stumbled and contracted by more than 7% in 1998. Real consumption, which registered an average growth of 6.5% per year from 1991 to 1997, suffered a severe contraction by more than 16% in 1998. The drop in the real investment was even more alarming in terms of its magnitude as well as its n

Corresponding author. Tel.: þ 60 3 7651 4197; fax: þ60 3 7651 4110. E-mail addresses: [email protected], [email protected] (M.H. Ibrahim), [email protected] (H.J.A. Ahmed). 0301-4215/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.enpol.2013.11.072

persistence. During 1991–1997, the annual growth of real investment was 11.6% on average. In 1998, it nosedived by more than 50%. This drastic drop seems to be lasting. While real GDP and real consumption managed to record average growth rates of respectively 4.9% and 6.0% from 1999 to 2011, the real investment only grew at a rate of 3.4% during the same period.1 Thus, in looking ahead at Malaysia's long term growth prospect and outlook, a key question that needs to be addressed is: what explains the slumps in Malaysia's investment? As posited by the neoclassical theory of investment, aggregate investment is a function of real income and cost of capital. The model, however, may not be adequate to account for Malaysia's investment performance post Asian crisis. As the aforementioned macroeconomic

1 Alternatively, we can also manifest the slumps in Malaysia's domestic investment after the Asian crisis based on the ratio of gross fixed capital formation to GDP, or the investment ratio. From 1991 to 1997, the average yearly investment ratio was slightly above 40%. It then dipped to 26.8% in 1998 and has remained low since at an annual average ratio of roughly 22–23% from 1999 to 2011.

M.H. Ibrahim, H.J.A. Ahmed / Energy Policy 67 (2014) 552–563

performance indicators would suggest, the uptrend in real gross domestic product has not been accompanied by a proportionate rise in the nation's aggregate investment. Adding to this, the growth of real investment remains low despite the low interest rate environment that has shaped the country's monetary scene ever since the Asian crisis. In other words, for the case of Malaysia, the explanation of its investment slump may lay elsewhere. Here, we content that the well documented uncertainties arising from particularly oil price escalation and its heightened uncertainty during recent years may provide an explanation. This contention has a basis in the irreversible theory of investment which views (i) investment decisions to be irreversible and (ii) possibility of postponing investments in line with the real option of waiting (Bernanke, 1983; Pindyck, 1991; Ingersoll and Ross, 1991). According to the theory, uncertainty increases the option value of waiting and hence depresses current investment. Accordingly, motivated by this contention, the present paper empirically analyses the roles oil price volatility plays in explaining Malaysia's aggregate investment behavior. The paper adds to the literature by considering potential differential investment implications of permanent and temporary volatilities. The two components are viewed to be driven by different sources, with the permanent volatility to arise from shifts in fundamentals and the temporary volatility from random events or shocks. Since the shifts in fundamentals are generally expected, the permanent volatility should have no discernible impacts on investment (Ahmed et al., 2012). Moreover, insurance should be more readily available for firms to cope with permanent volatility (Byrne and Davis, 2005b). By contrast, the random nature of the temporary volatility is expected to lead the firms to be more conservative or to postpone investments in line with the option value of waiting. Accordingly, making distinction of the nature of uncertainty is important for better understanding the uncertainty – investment links. Apart from attempting to address Malaysia's investment slumps post crisis, the analysis has relevance for various reasons. First, although the role of uncertainty in macroeconomic performance has been widely investigated (Episcopos, 1995; Federer, 1993, 1996; Guo and Kliesen, 2005; Ahmed and Wadud, 2011; Ahmed et al., 2012; Ng, 2012), the macroeconomic importance of oil price uncertainty, permanent and transitory, has been largely absent for a developing country like Malaysia. Thus, the present analysis fills this gap in the literature by providing a developing country's perspective on the uncertainty – investment link. Second, as Malaysia has implemented various policy measures to revive private investment and to again place the private sector at the driver seat of its development process, e.g. through the recently launched economic transformation program (ETP) and recently tabled annual federal budget,2 identifying whether oil price volatility and its components have explained its investment behavior should be essential for designing various relevant policies including stabilization, financial and energy policies. And finally, the findings from the present study should also prove useful for other developing countries that have experienced similar behavior of aggregate investment. In the analysis, we adopt a components generalized autoregressive conditional heteroskedasticity (CGARCH) model developed by Engle and Lee (1999) to decompose conditional oil volatility into its permanent and transitory components. Then, the obtained volatility measures are included as additional determinants of investment behavior. The dynamics of aggregate investment is addressed using a reduced-form vector autoregression (VAR) model. As a preview to the results, we find oil volatility regardless of its nature to have 2 With the main objective to propel Malaysia to become a high-income developed nation, the ETP identifies 12 national key economic areas to be led mainly by the private sector while the public sector provides a facilitating role in the process. The 2013 budget recently tabled by Malaysia's prime minister also sets “boosting investment activity” as one of five focus areas.

553

adverse effects on aggregate investment and real output. Contrary to the findings for the developed markets, the adverse implications of permanent volatility seem stronger. We also find evidence that real interest rate tends to drop following oil price shocks but is likely to increase in responses to oil volatility shocks. These results appear reasonably robust to variable specification and measurements in the VAR model. The rest of the paper is organized as follows. As a precursor to the analysis, we provide a brief background information. Then, Section 3 reviews related literature. The next two sections detail the empirical approach and present estimation results. Finally, Section 6 contains a summary of the main findings and some concluding remarks.

2. Background information Malaysia is one of the vibrant and open economies within ASEAN, whose positive macroeconomic performance was interrupted only by the 1985 and 1997/1998 recessions. At the time of its independence in 1957, Malaysia was an agricultural-based economy. Since the early 1980s, it has developed its way to be an industrial-based economy and aimed at reaching a developed country status by 2020. In the process, Malaysia has depended heavily on imported intermediate and capital goods for its industrialization process. As the oil price may play a role through a nation's trade matrix as demonstrated by Abeysinghe (2001), oil price variations may have repercussions on Malaysia's capital formation, an issue that deserves investigation. Moreover, Malaysia's current status as a net exporter of oil but with narrowed gap between domestic oil consumption and oil production over the years provides further motivation to undertake this study. More specifically, with the ratio of domestic oil consumption to oil production to increase steadily from 0.42 in 1990 to 0.60 in 2000 and to 0.80 in 2009 (Ibrahim and Said, 2012), there is a need to look at Malaysia's energy policy such that the adverse impacts of oil price volatility can be curtailed. In their early stage of economic development, Malaysian was reliant primarily on fossil fuel for most of energy need. When Malaysia formulated its first energy policy, concern over efficient utilization of energy and the need for energy development to take account of environmental issues were fundamental. The Energy Policy of 1979, the National Depletion Policy of 1980 and the Four Fuel Diversification Policy of 1981, have provided the impetus for the development of energy supply. Since then the focus in the energy sector has shifted to the sustainable development of nonrenewable resources and the diversification of energy sources. The Four Fuel Diversification Policy identified the country's preferred energy mix as oil, natural gas, coal and hydro power. In 2001, the government articulated the Five Fuel Policy, adding renewable resources and linking this to sustainability and efficiency. More recently, Malaysia has set itself a key strategic objective of becoming a regional energy hub. Speaking at the opening ceremony of the 17th Asia Oil and Gas Conference in Kuala Lumpur, Malaysia's Prime Minister Najib said the energy sector plays a pivotal role in the Malaysian economy with oil and gas contributing to more than 40% of the country's national income. Based on a report from Malaysian Petroleum Resources Corp., Malaysia has its role as an energy consumer and a regional hub for energy trading. In cooperation with Singapore, with its strong financial and IT sectors, Malaysia – a key exporter of oil and gas to Japan – is striving to become a global center of the energy industry. As suggested in the report, Malaysia is a bridge between the markets of Europe and those of China and the rest of Asia. Until now, Malaysia has developed as an oil and gas producer, but economic growth has pushed domestic demand to the point where supply capacity will

554

M.H. Ibrahim, H.J.A. Ahmed / Energy Policy 67 (2014) 552–563

probably decline in the future. Against this backdrop, assessment of the macroeconomic implications of oil price volatility is much needed given that the success of any energy plan would depend critically on how it can insulate the economy from global oil market disturbances.

3. Related literature The investment behavior is well founded on such theories as the neoclassical investment theory, the accelerator model of investment and q theory of investment (see discussion in Chirinko, 1993). However, as noted by Pindyck (1991) and Laopodis and Sawhney (2007), these theories have had limited success in accounting for variations in investment. Pindyck (1991) further notes that the inability of these traditional models to explain investment behavior stems from their failure to recognize (i) the irreversibility of investment decisions and (ii) firms' ability to postpone investment in anticipation of getting new information along the line of a real option problem. Considering these characteristics of investment, Bernanke (1983), Pindyck (1991), Pindyck and Solimano (1993) and Dixit and Pindyck (1994) demonstrate theoretically a negative impact of uncertainty on firms' investment. By contrast, assuming perfect competition, constant returns to scale and symmetric adjustment costs, Hartman (1972) and Abel (1983) show that an increase in uncertainty may instead boost firms' real investment by raising the marginal value of capital. These contradicting theoretical predictions have attracted a great deal of empirical studies and the recurrence of various forms of uncertainty in recent years has further heightened interest in the subject. In the empirical literature, uncertainty related to various financial and macroeconomic variables have been considered. These include, among others, output uncertainty, inflation uncertainty, exchange rate uncertainty and stock market uncertainty.3 As regards to the oil price, the focus has been mostly on the macroeconomic effects of oil price shocks. An illustrative list of the recent studies includes Aydin and Acar (2011), Iwayemi and Fowowe (2011), Jayaraman and Keong (2009), Jbir and ZouariGhorbel (2009), Oladosu (2009) and Yeh et al. (2012). Federer (1993), Episcopos (1995) and Goel and Ram (1999) are among early studies evaluating the investment – uncertainty relations for the US and other OECD countries. Federer (1993) utilizes risk premium embedded in the term structure of interest rate as a measure of uncertainty, which is argued to be forward looking and to better capture future uncertainty, and examines its relation to investment in durable equipment as well as contracts and orders of new plant and equipment for the US case. He finds their relations to be negative and significant. Meanwhile, Episcopos (1995) considers various sources of uncertainty including uncertainty in leading indicator index, stock price, consumption, interest rate and the price level. He relies on an autoregressive conditional heteroskedasticity (ARCH) model to construct the uncertainty measures. The inverse investment – uncertainty relations are again documented. Goel and Ram (1999) widen the analysis to a panel of 12 OECD countries but focus specifically on inflation uncertainty measured by 5-year moving standard deviation. Their analysis is distinct from the previous studies in that an attempt is made to relate the dependence of uncertainty effect on the degree of investment irreversibility by looking at investments in producer durables, non-residential structures and residential structures. They find evidence that the adverse 3 Carruth et al. (2000) provide a comprehensive empirical survey on investment and uncertainty, to which interested readers may refer. Many recent studies have also focused on the impacts of uncertainty on investment at the firm level; see for instance Bulan (2005), Bloom et al., (2007), Haque and Shaoping (2008), Fuss and Vermeulen (2008) and Demir (2009).

effect of uncertainty does depend on the degree of investment irreversibility. More specifically, being most irreversible, investments in producer durable tend to be most adversely affected by inflation uncertainty. Meanwhile, inflation uncertainty does not seem to influence investments in residential structures, which are argued to possess low degree of irreversibility. The analysis for the developed economies is further undertaken recently by Byrne and Davis (2004, 2005a, 2005b) and Stockhammer and Grafl (2010). Like Goel and Ram (1999) and Byrne and Davis (2004) examine the impacts of inflation uncertainty on real investment but only for the US case. However, unlike Goel and Ram (1999), they make distinction between temporary and permanent components of uncertainty using a switching regime model. Both components of uncertainty exert adverse effects on the US real investment but with stronger effect from the temporary component. In a subsequent study, Byrne and Davis (2005a) extend the analysis to G-7 countries. Adopting GARCH-generated measures of uncertainty, they highlight the importance of sources of uncertainty. More specifically, they document consistent evidence supporting the significance of exchange rate uncertainty. Meanwhile, interest rate uncertainty turns significant only for the later sample and uncertainties in consumer prices, industrial production and stock prices are found to be insignificant. Built upon this finding, they further investigate whether temporary and permanent exchange rate volatilities have different implications on investment using the same data set (Byrne and Davis, 2005b). Their results suggest the adverse investment effect of only temporary exchange rate uncertainty. Looking at five developed countries (i.e. Germany, France, US, UK and the Netherlands), Stockhammer and Grafl (2010) find the adverse effects of financial uncertainties, measured by 12-month moving average of squared changes in exchange rate, stock price and price level, only for the US and the Netherlands. While these recent studies add further evidence supporting the adverse effects of uncertainty on firms' investment behavior, they bring up a notable dimension on the subject. Namely, the nature of uncertainty matters. As Byrne and Davis (2005b) point out, firms should be able to cope with permanent volatility since insurance should be readily available. By contrast, short-term sparks in volatility would make firms to be more conservative in their investment decision. Ahmed et al. (2012) further emphasize that the sources of permanent volatility are shifts in fundamentals while those of temporary volatility stem from random events such as sudden disruptions in oil markets and political unrests. They view the former to constitute expected fluctuations in oil prices and the latter to be unexpected changes in oil volatility. Based on these arguments, firms should be affected more by the transitory uncertainty. Departing from the aforementioned studies that have a main focus on developed markets, several studies investigate the uncertainty – investment link for developing countries. Notable among these studies are Serven (2003), Pradhan et al. (2004), Kumo (2006), Bhandari and Upadhyaya (2010), Ang (2010), Fatimah and Waheed (2011) and Ibrahim (2011). Collaborating Byrne and Davis (2005a) and Serven (2003) documents evidence supporting the adverse effect of exchange rate uncertainty on investment in developing countries. He further notes that a negative exchange rate uncertainty – investment link is stronger for a country with higher openness and weaker financial system. In a country-specific study, Kumo (2006) evaluates the issue for South Africa by considering GARCH-generated measures of uncertainty of output, exchange rate, interest rate, inflation and terms of trade. The negative investment implication of exchange rate volatility is further highlighted. Moreover, for the case of South Africa, uncertainty arising from output variations is also noted to be significant. In a more recent study on Pakistan, Fatimah and Waheed (2011) look at a variety of macroeconomic uncertainties and document significant and negative investment effects of

M.H. Ibrahim, H.J.A. Ahmed / Energy Policy 67 (2014) 552–563

uncertainties stemming from budget deficit, government consumption, total trade, exports and FDI. Like many other studies discussed above, they apply the GARCH methodology for volatility or uncertainty measurement. Pradhan et al. (2004), Bhandari and Upadhyaya (2010), Ang (2010) and Ibrahim (2011) all look at a country or group of countries in Southeast Asia. Except Ang (2010), these studies employ GARCH-generated measures of uncertainty. Pradhan et al. (2004) assess the implications of exchange rate volatility on aggregate private investment in Indonesia, Malaysia, the Philippines and Thailand using annual data from 1972 to 2001. Their results are rather inconclusive with the findings to be different across countries. Namely, while they note a positive influence of exchange rate uncertainty on investment in Malaysia, the evidence for Thailand suggests otherwise. Meanwhile, no significant relation is found for Indonesia. As for the Philippines, the coefficient of contemporaneous exchange rate volatility is significantly positive and that of once-lagged volatility is negative. However, when Bhandari and Upadhyaya (2010) pool these countries together and apply basic panel estimation methods, the exchange rate uncertainty is found to negatively affect aggregate investment in these countries. Ang (2010) incorporates 3-year moving standard deviation of real output changes in his analysis to identify factors accounting for aggregate investment slumps in Malaysia. In line with many studies, he finds its impact to be negative. Adding to these findings for the developing markets, Ibrahim (2011) finds the negative impacts of stock market uncertainty on Thai real investment in the long run as well as short run. While these studies differ in terms of countries covered (developed versus developing countries), data structure (time series versus panel data), time periods, and measurement and sources of uncertainty, they seem to provide a fairly consensus view that uncertainty depresses real investment. Still, looking back at the aforementioned studies, we may realize the absence of oil markets as a source of volatility and its implication on investment. In the literature, numerous studies have evaluated the oil uncertainty – real output relationship and, as noted in Elder and Serletis (2009), real investment's responses to oil price uncertainty is one of the reasons underlying the dampening output effect of oil volatility. In other words, with escalation and sharp swings in oil prices in recent years, oil market volatility deserves empirical attention. We believe that the oil market could be a better source of uncertainty during recent period as compared to exchange rate uncertainty considered by Pradhan et al. (2004) and Bhandari and Upadhyaya (2010). While Malaysia's exchange rates were highly volatile during the Asian crisis, they exhibited relative stability in most years as a result of “fear of floating” prior to the Asian crisis and officially fixed exchange rate from September 1998 to June 2005. By contrast, sharp swings in oil price have been obvious in years after the Asian crisis. Hence, we add to the literature by looking at whether oil volatility and, in line with Byrne and Davis (2004, 2005b), transitory and permanent components of volatility account for Malaysian aggregate investment dynamics.

4. Empirical approach While empirical analyses on investment behavior rely mainly on a well-specified investment function as hinted by theories, the present analysis takes a pragmatic approach by focusing on dynamic interactions between investment and its various determinants. In specific, we infer the relations between real investment and oil volatility as well as other investment determinants on the basis of a reduced-form vector autoregressive (VAR) framework. The approach allows all variables, i.e. real investment and its determinants, to be

555

potentially endogenous and hence the possibility of feedback effects among them and it captures empirical regularities in theoretically related time series with minimal a priori theoretical restrictions, which are reasons often cited for its wide adoption in econometric analyses of time series. Besides, the adoption of VAR is based on the following reasons. When modeling the dynamic behavior of investment, a oneequation error-correction representation of investment is normally applied to capture short-run responses of investment to its determinants and its adjustment towards a long-run equilibrium. However, the one-equation approach is statistically valid only if the other included variables are weakly exogenous. Accordingly, inferences that have been made would be more likely invalid if this is not the case, which we find in the present context, and hence the necessity for multi-equation modeling. And second, by allowing all variables to be potentially endogenous, the framework provides broader perspectives on macroeconomic implications of oil market development. For instance, in addition to looking at the influences of oil volatility on investment, we may gain additional insights on whether real investment is a main channel through which oil volatility exerts dampening output impacts (see Beetsma and Giuliodori, 2012 for stock market volatility) and whether adjustments of interest rate, due to for example the changes in monetary policy stances, play a role in ameliorating the real implications of oil prices and oil volatility (Bernanke et al., 1997). Given our main theme, we first discuss measurement of oil price volatility and its transitory and permanent components and then elaborate the VAR modeling and its related statistical issues.

4.1. Measurement of oil price volatility The generalized autoregressive conditional heteroskedasticity (GARCH), which has an appealing statistical property for its ability to capture the normally observed volatility clustering, is perhaps the most common approach to characterize volatility of a time series. Since its introduction by Engle (1982) and later generalization by Bolleslev (1986), the GARCH methodology has been extended to incorporate various other features of volatility. An interesting extension of the basic GARCH model is the components GARCH or CGARCH model by Engle and Lee (1999). As its name suggests, the CGARCH decomposes the volatility into its transitory and permanent components. Hence, given its relevance to the present analysis, we adopt the CGARCH model. We write the CGARCH model for the oil price (ot) as: p

Δot ¼ μ0 þ ∑ μi Δot  i þ εt

ð1Þ

s2t  qt ¼ w þ αðε2t  wÞ þ βðs2t  1 wÞ

ð2Þ

qt ¼ γ þ ρðqt  1  γ Þ þ ϕðε2t  1  s2t  1 Þ

ð3Þ

i¼1

where Δ is the first difference operator and εt is the error term with a time-varying variance (i.e. a measure of conditional volatility) s2t . Eq. (1) is the conditional mean equation specified to follow an autoregressive process, the order (p) of which is selected based on un-correlated errors. Eqs. (2) and (3) capture respectively the transitory and permanent components of conditional volatility. Note that the conditional volatility is meanreverting around the permanent volatility (qt), the latter converges to its mean of γ =ð1  ρÞ. Accordingly, s2t  qt measures the transitory component of volatility with the speed of mean reversion of (α þ β). Since the permanent volatility is more persistent than the transitory volatility, it is expected that 0 o(α þ β)o ρ o1.

556

M.H. Ibrahim, H.J.A. Ahmed / Energy Policy 67 (2014) 552–563

4.2. Investment modeling As noted, we adopt the VAR methodology to discern dynamic interactions between aggregate investment and oil volatility. More specifically, we form a 5-variable VAR model consisting of real aggregate investment (inv), real GDP (y), real interest rate (r), real oil price (op), and oil volatility or uncertainty (var) to serve as a baseline specification. Real aggeragte investment and oil volatility are our focal variables. The inclusion of real GDP and interest rate is in line with many existing studies that build upon the neoclassical theory of investment. The inclusion of oil price, however, requires some explanation. We believe that, in looking at the implication of the oil market, both the level (first moment) and variability (second moment) of oil price should be included for the following reason. Being a small open economy, Malaysia is heavily dependent on international trade. From the investment perspective, intermediate and capital goods form a major part of Malaysian imports. As demonstrated by Abeysinghe (2001), despite being a significant oil-exporting country, oil prices tend to exert negative influences on Malaysia's output and the influences work mainly through the trade matrix of the country. Accordingly, oil prices are likely to have bearings on aggregate investment. Following a common approach in the VAR methodology, especially when evaluating dynamic interactions among the variables, we write the VAR in level form as: p

X t ¼ A0 þ ∑ Ai X t  i þvt i¼1

ð4Þ

where X ¼[inv, y, r, op, var], A0 is a 5  1 vector of constant terms, Ai is a 5  5 matrix of coefficients, vt is a 5  1 vector of error terms and p is the optimal lag order selected such that the error terms are serially uncorrelated. Although some variables employed in the analysis may contain a unit root or are likely to be nonstationary, the level VAR estimation is consistent when there is cointegration among the variables (Sims et al., 1990). Ramaswamy and Slok (1998) further outline several advantages of using level VAR instead of vector error-correction modeling (VECM) for evaluating causal interactions among the cointegrated series. As they note, if there is no a priori theory to suggest the number of cointegrating vectors or to interpret them, which would be difficult if two or more cointegrating vectors are identified, imposing wrong long-run restrictions in the VECM leads to bias short-run dynamics. Moreover, while the VECM shocks and their impacts are implied to be persistent, those from the level VAR allow the data to speak for themselves. Based on these arguments, we opt for the level VAR in our analysis. In the empirical implementation, three VAR systems are estimated each with oil uncertainty (var) represented by conditional uncertainty (cvar), temporary uncertainty (tvar) and permanent uncertainty (pvar). From the estimated VAR, we simulate impulse–response functions (IRF) as a basis for inferences. The impulse–response functions trace temporal responses of a variable of interest to its own innovations and innovations in other variables. Thus, from the IRF, we can note temporal responses of real investment (inv) to innovations in oil volatility (cvar, pvar, and cvar), the main theme of the paper, and to innovations in other investment determinants. At the same time, the VAR enables us to disentangle the dynamic impacts of oil market variables on other variables in the system. A reduced-form VAR as written in (4) is taken to represent an unknown structural system and thus its shocks cannot be construed as structural shocks. Since the reduced-form shocks are composite of structural shocks, an important consideration in generating the impulse–response functions is identification of structural shocks. The traditional strategy is to adopt Sims' (1980) approach by using the so-called Cholesky orthogonalization. The approach has a major

weakness in that, due to the recursive structure of shocks in the Cholesky orthogonalization, it requires a pre-specified causal ordering of the variables and the results of the IRF will be affected by the way the variables are ordered unless the correlations among the VAR errors are low or insignificant. To circumvent this problem, Koop et al. (1996) and Pesaran and Shin (1998) develop the generalized impulse–response functions by incorporating historical patterns of correlations among different shocks. In this way, the impulse–response functions are unique and invariant to alternative orderings of the variables. Due to this advantage, we adopt the generalized impulse–response functions in the analysis. The data are quarterly covering the period 1991.Q1 to 2012.Q2, a total of 86 observations. We employ gross fixed capital formation and gross domestic product, both deflated by GDP deflator, to represent respectively real aggregate investment and real income. Both real investment and real GDP are seasonally adjusted using the X12 procedure. The real interest rate is captured by the average lending rate less year-to-year inflation rate. We use the Brent spot crude oil price multiplied by the ringgit-USD rate as a measure of oil price in domestic currency. The price is then deflated by the consumer price index to get the real oil price. Monthly Brent oil prices in domestic currency are used in the estimation of CGARCH. Then, a quarterly volatility measure is constructed by taking average of volatility measures within the quarter. Real investment, real GDP and real oil price are expressed in natural logarithm while real interest rate and volatility measures are used in their original level form. These data are obtained from Datastream International.

5. Estimation results In this section, we first briefly discuss estimation results of CGARCH model. This is followed by the baseline results. Finally, to further add credence to the baseline results, we perform several robustness checks. These include the inclusion of crisis-dummy variables, employment of alternative measures of variables, addition of other variables, and incorporation of other sources of uncertainty. 5.1. CGARCH results The results for the CGARCH model are given below in Eqs. (5) through (7) while the plots of conditional variance and its permanent and transitory components are presented in Fig. 1,4: Δot ¼ 0:002 þ 0:198Δot  1 þ 0:082Δot  2 þ 0:104Δot  3  0:115Δot  4  0:131Δot  5

ð0:464Þð0:001Þ

ð0:149Þ

ð0:041Þ

ð0:028Þ

ð0:011Þ ð5Þ

s2t  qt ¼ 0:314ðε2t  1  qt  1 Þ  0:017ðs2t  1  qt  1 Þ ð0:000Þ ð0:920Þ

ð6Þ

qt ¼ 0:082 þ 0:998ðqt  1  0:082Þ þ 0:141ðε2t  1 s2t  1 Þ ð0:927Þ ð0:000Þ ð0:032Þ

ð7Þ

The mean equation for oil return is specified to follow an autoregressive process of order 5 on the basis of the un-correlated errors. From the estimated mean equation, changes in oil price tend to be followed by further changes in the same direction for the next three quarters. Thereafter, their movements are corrected or reversed. The transitory oil volatility does not seem to exhibit 4 In the estimation, we also allow for asymmetric volatility but find no evidence for the presence of asymmetry.

M.H. Ibrahim, H.J.A. Ahmed / Energy Policy 67 (2014) 552–563 .05 CVAR PVAR TVAR

.04 .03 .02

.025

.01

.020

.00

.015 .010 .005 .000 -.005 -.010 1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

Fig. 1. Graphical plots of volatility measures. Notes: CVAR, PVAR and TVAR represent respectively conditional volatility, permanent volatility and temporary volatility.

the GARCH effect. Instead, it stems principally from shocks in oil price. We may also note that the permanent component of oil volatility is accounted by both past volatility and past shocks or new arrivals of information. The high value of the autoregressive term (i.e. GARCH term) relative to the estimated coefficient of shocks (i.e. ARCH term) in the permanent volatility suggests high volatility persistence. The graphical plots of oil volatility and its components reveal two interesting phases of oil market volatility. While the conditional volatility had been below its permanent component prior to the 1997/1998 Asian crisis, we may observe periodic rise in the permanent volatility as well as frequent sparks in the transitory volatility since then. The upswings in transitory volatility and rise in permanent volatility tend to coincide with various major financial and political events such as the 1997/1998 Asian crisis, Dot Com bubble in early 2000, 9/11 attack and subsequent war on terrorism in 2001, Enron bankruptcy in 2001, and the latest subprime crisis from 2007 to 2009. Based on the volatility pattern after the Asian crisis, it would thus be interesting to see whether it accounts for sluggish investment performance in Malaysia and, if it does, whether the temporary or permanent components play a major part. We address this main theme next. 5.2. Baseline results As noted, our baseline VAR model includes real investment, real GDP, real lending rate, real oil price and a measure of oil volatility (i.e. conditional volatility, permanent volatility or transitory volatility). We first subject the time series to standard ADF and PP unit root tests and Johansen–Juselius cointegration test (Johansen, 1988; Johansen and Juselius, 1990). Since these tests are standard, we do not detail them here. The results of these tests are not reported to conserve space. From the ADF and PP unit root tests, we find all variables except transitory volatility to be nonstationary integrated of order 1, or I(1). The cointegration tests further indicate the presence of long run relations among the variables regardless of which measure of oil volatility is used. The finding of cointegration prevails even when we drop the I(0) transitory volatility from the system. However, the evidence tends to suggest the presence of two cointegrating vectors in all systems. Moreover, the weak exogeneity tests fail to support the weak exogeneity of some investment determinants. With these findings, the adopted level VAR approach is justifiable. Accordingly, three baseline level-VAR systems, each with alternative measures of oil volatility, are estimated. Based on the criterion that the error terms in each of the VAR equations are serially uncorrelated, the lag order is set to 4 for the systems with conditional volatility and transitory volatility and to 3 for the system with permanent volatility. Given our focus on the

557

implications of oil price and oil market volatility on Malaysia, we report only the responses of real investment, real GDP and real lending rate to one standard deviation shocks in other variables in the systems. These impulse–response functions are given in Fig. 2. As may be observed from fig., the interactions among domestic or non-oil variables (real investment, real GDP and real interest rate) are generally in conformity across the three baseline systems. The causal pattern between real investment and real GDP is bi-directional. A one standard deviation shock in real GDP is noted to exert significant and quite lasting responses from real investment. Likewise, a positive shock in real investment tends to boost real GDP significantly roughly up until 4 quarters. These results conform well to the prediction of standard growth and investment theories such as the neoclassical growth theory and neoclassical investment theory. Moreover, in line with conventional wisdom, positive shocks in the real interest rate also lead to contraction in both real investment and real output. However, the real impacts are significantly felt only after some lags, i.e. roughly after 3–6 quarters depending on which measure of oil volatility is used. These lagged effects are in line with the notion that a rise in the policy interest rate, which is then passed through to the lending rate, may not immediately contract bank loans due to the availability of capitals as a buffer against loanable fund shortfalls. Note also that, while investment shocks do not result in significant responses from the real lending rate, the output shocks do. The pattern of responses tend to suggest that, following output shocks, inflation increases as a faster pace than the lending rate does resulting in a drop in the real interest rate. The impulse–response functions of the three domestic variables to real oil price shocks are also noteworthy. If we take real lending rate to reflect banks' pricing behavior in responses to changes in monetary stance, the results tend to suggest the role of monetary policy in ameliorating the adverse effects of oil price shocks, in line with the findings by Bernanke et al. (1997) for the US and Lee et al. (2001) for Japan. The decline in the real lending rate following positive oil shocks tends to be quite immediate and it remains significant for roughly more than 1 year. This may explain why real investment and real output are not largely affected or positively but temporarily influenced by oil price shocks. Our contention that oil volatility may play a role in accounting for real investment performance seems to receive empirical supports from the impulse–response functions of the three domestic variables to shocks in oil volatility (panel a). The adverse impacts of conditional volatility shocks on real investment and real output are significant and quite persistent. Following oil volatility shocks, real investment and real output contract significantly and the contraction remains significant to more than 2 years. In a recent work for Malaysia, Ahmed and Wadud (2011) note that oil volatility shocks tend to depress real industrial production. We reaffirm this finding using real GDP. A further insight that we gain from the present analysis is that heightened oil volatility is likely to increase financial risk as reflected by significant increases in the real interest rate after the oil volatility shocks. This may have further exacerbated and explained quite persistent real impacts of oil volatility. When we decompose oil volatility into its transitory and permanent components and enter them alternatively in the VAR model, we again observe significant influences of both components on real investment and real output. When we compare the responses of real investment and output to both transitory and permanent oil volatility, which are graphed in Fig. 3 for ease of comparison, we may note roughly similar impacts at 1–2 quarter horizons. However, after 2 quarters, the adverse influences of permanent volatility tend to be stronger especially on real output. These results depart from the ones documented by Byrne and

558

M.H. Ibrahim, H.J.A. Ahmed / Energy Policy 67 (2014) 552–563

Fig. 2. Baseline results. (a) Conditional oil volatility (Lag order¼ 4), (b) permanent oil volatility (Lag order¼ 3) and (c) temporary oil volatility (Lag order¼ 4).

Davis (2004, 2005b) for the US and OECD. To recall, Byrne and Davis (2004) document the significant adverse effects of both permanent and temporary inflation uncertainty for the US case

but find the impact of temporary uncertainty to be stronger. Meanwhile, in Byrne and Davis (2005b), only temporary exchange rate volatility exerts negative bearings on OECD investment.

M.H. Ibrahim, H.J.A. Ahmed / Energy Policy 67 (2014) 552–563

559

-.006

.00

-.007 -.01

-.008 -.009

-.02

-.010

-.03

-.011 -.012

-.04

-.013

-.05

PVAR TVAR

-.014

PVAR TVAR

-.015

-.06 1

2

3

4

5

6

7

8

9

10

Responses of Investment

1

2

3

4

5

6

7

8

9

10

Responses of Output

Fig. 3. Investment and output effects of permanent and temporary volatility. Responses of investment and responses of output respectively.

Although the arguments underlying their findings seem appealing, we should not expect that their results would also hold for a developing country. Indeed, on the basis of their reasoning, it is plausible that for a developing country, the breadth and depth of its financial markets may not be adequate in providing insurance against any nature of uncertainty. Moreover, the developing country tends to lack managerial expertise, a reason often related to the Asian crisis (Mishkin, 1999), and consequently the ability to identify the nature of risk or uncertainty. As a consequence, oil volatility regardless of its nature tends to dampen aggregate investment and real output. 5.3. Robustness In this subsection, we perform robustness checks to further add credence to our baseline results. Given our main interest and to conserve space, we report only the impulse–response functions of real investment and GDP to only oil volatility shocks under various cases. In the robustness exercise, we consider 9 different cases based on modification or extension of the baseline systems. In the first case, in line with such studies as Byrne and Davis (2005a, 2005b), we drop the real oil price from the systems. Case 2 involves the use of real money market rate instead of the real lending rate. The money market rate is normally used to represent monetary policy (Ibrahim, 2005). In case 3, we incorporate two crisis dummies to account for the influences of the Asian and global financial crises on real activities. We specify the Asian crisis dummy to take the value of 1 for 1997.Q3 to 1998.Q4 and 0 otherwise. Meanwhile, the global financial crisis dummy assumes the value of 1 for 2007–2009. In case 4, we experiment with the West Texas Intermediate (WTI) crude oil price to represent the oil markets. As in the baseline case, the WTI is expressed in domestic currency and deflated by the consumer price index. Monthly WTI price is used for the CGARCH estimation to derive conditional volatility, permanent volatility and transitory volatility. In cases 5 and 6, we add alternatively real credit and real government spending. The real credit is represented by commercial banks' loans and advances deflated by the consumer price index. As argued by Serven (2003), interest rates are to a large extent controlled in developing countries and the presence of credit-rationing in their financial markets is common. He suggests the inclusion of some measures representing credit market conditions, the suggestion that justifies the inclusion of real bank credit. The government consumption as a determinant of aggregate investment is also widely investigated in the context of “crowding-out” and “crowding-in” debate and hence the inclusion of real government expenditure. The remaining three cases add alternatively measures of stock market volatility, exchange rate volatility and inflation volatility, which have been considered in various existing studies. For the construction of these volatility measures, we use the market-wide Kuala Lumpur composite index to represent stock price, the nominal effective exchange rate, and the consumer

price index to construct inflation and apply the GARCH methodology. Data required for these robustness checks are again retrieved from Datastream International. Fig. 4 plots the impulse–response functions of both real investment and GDP to shocks in conditional oil volatility, permanent oil volatility and temporary oil volatility. As may be observed from the plots, our baseline results generally prevail in all cases examined. Namely, the negative bearings of conditional oil volatility hold for all cases. We may also observe consistently negative influences of permanent volatility on real output in the modified or extended VAR systems. Likewise, while real investment effects of permanent oil volatility tend to be marginally significant when the crisis dummies or inflation volatility is incorporated, they are significant in all other cases. As regards to shocks in temporary oil volatility, the significantly negative responses from output also hold for all cases. Its implication on real investment; however, depends on VAR specifications. More specifically, it turns insignificant when real oil price is dropped and the crisis dummies, real government spending and inflation volatility are added. Looking across the impacts of the two oil volatility components, the stronger effects of permanent volatility are further substantiated. In a nutshell, our baseline results appear reasonably robust to modifications and extensions of the VAR systems.

6. Conclusion and policy implications The present paper empirically examines the relations between oil market volatility and Malaysia's aggregate investment behavior by means of a vector autoregressive (VAR) model. Applying the VAR methodology containing real investment, real GDP, real lending rate, real oil price and nature of oil volatility (total oil volatility, permanent oil volatility and temporary oil volatility), i.e. our baseline system, we generate generalized impulse–response functions as a basis of inferences. Conforming to the prediction by the irreversible theory of investment, the baseline results indicate the negative influences of oil volatility on real investment and real output. In addition, countering the documented finding for the US and other developed markets, the real investment and real output impacts of permanent oil volatility are found to be stronger as compared to those of temporary oil volatility. The real interest rate also tends to increase following positive shocks in oil market volatility. Hinting on the increase of financial risk, the increase in the interest rate perhaps further explains the quite persistent influences of oil volatility on real investment and real output. These baseline results appear reasonably robust across measurements and specifications of the variables in the VAR system. Based on these results, we incline to attribute the slumps of Malaysia's aggregate investment since the 1997/1998 Asian crisis to recurring and heightened uncertainty originating from the global oil market.

560

M.H. Ibrahim, H.J.A. Ahmed / Energy Policy 67 (2014) 552–563

Fig. 4. Robustness analysis. (a) Dropping real oil price, (b) using real money market rate, (c) adding Asian crisis and global financial crisis dummies, (d) using WTI, (e) adding real credit, (f) adding real government spending, (g) adding stock market volatility, (h) adding exchange rate volatility and (i) adding inflation volatility.

M.H. Ibrahim, H.J.A. Ahmed / Energy Policy 67 (2014) 552–563

561

Fig. 4. (continued)

In the light of our findings, we can emanate a number of implications. Firstly, in view of the increased volatility of oil prices and the resultant surge of uncertainty, Malaysian firms could

formulate their short and long run investment based on the anticipated length of uncertainty. Since the permanent volatility has been found to have a dampening effect on aggregate

562

M.H. Ibrahim, H.J.A. Ahmed / Energy Policy 67 (2014) 552–563

Fig. 4. (continued)

investment, it could be worthwhile for the government to further diversify its energy needs. Further, to cope with such volatility, and to reduce the importance of oil consumption relative total energy

demand for oil, the government can adopt a framework to improve energy efficiency and regulations to restraint demand for oil by encouraging the use of alternative or renewable energy. Hence, the

M.H. Ibrahim, H.J.A. Ahmed / Energy Policy 67 (2014) 552–563

present energy strategic objective to become an energy hub in the region is a step in the right direction. At micro level, firms may take possible hedging positions to neutralize long lasting effect of permanent oil price volatility on investment. In this regards, further development to broaden and deepen financial markets is needed. At the same time, focus should also be given to upgrading human capitals equipped with managerial as well as analytical expertise to analyze risk and identify their nature and sources. Finally, it is evident that monetary policies are needed to offset the detrimental effect of oil volatility on investment. It is expected that as a newly industrialized net oil exporting economy, Malaysia would be able to further fine-tune its development course by adopting a host of these policies to subside the adversities of oil price volatility. References Abel, A., 1983. Optimal investment under uncertainty. Am. Econ. Rev. 73, 228–233. Abeysinghe, T., 2001. Estimation of direct and indirect impact of oil price on growth. Econ. Lett. 73, 147–153. Ahmed, H.J.A., Bashar, Omar H.M.N., Wadudd, Mokhtarul I.K.M, 2012. Thetransitory and permanent volatility of oil prices: what implications are there for the US industrial production? Appl. Energy 92, 447–455. Ahmed, H.J.A., Wadud, Mokhtarul I.K.M, 2011. Role of oil price shocks on macroeconomic activities: an SVAR approach to the Malaysian economy and monetary responses. Energy Policy 39, 8062–8069. Ang, J.B., 2010. Determinants of private investment in Malaysia: what causes the postcrisis slumps? Contemp. Econ. Policy 28 (3), 378–391. Aydin, L., Acar, M., 2011. Economic impact of oil price shocks on the Turkish economy. Energy Policy 39 (3), 1722–1731. Beetsma, R., Giuliodori, M., 2012. The changing macroeconomic response to stock market volatility shocks. J. Macroecon. 34, 281–293. Bernanke, B.S., 1983. Irreversibility, uncertainty and cyclical investment. Q. J. Econ. 98, 85–106. Bernanke, B.S., Gertler, M., Watson, M., 1997. Systematic monetary policy and the effects of oil price shocks. Brook. Pap. Econ. Act. 1, 91–142. Bhandari, R., Upadhyaya, K.P., 2010. Panel data evidence of the impact of exchange rate uncertainty on private investment in South-east Asia. Appl. Econ. 42, 57–61. Bloom, N., Bond, S., Van Reenen, J., 2007. Uncertainty and investment dynamics. Rev. Econ. Stud. 74, 391–415. Bolleslev, T., 1986. Generalized autoregressive conditional heteroskedasticity. J. Econom. 31, 307–327. Bulan, L.T., 2005. Real options, irreversible investment and firm uncertainty: new evidence from U.S. firms. Rev. Financ. Econ. 14, 255–279. Byrne, J.P., Davis, E.P., 2004. Permanent and temporary inflation uncertainty and investment in the United States. Econ. Lett. 85, 271–277. Byrne, J.P., Davis, E.P., 2005a. Investment and uncertainty in the G7. Rev. World Econ. 141 (1), 1–32. Byrne, J.P., Davis, E.P., 2005b. The impact of short- and long-run exchange rate uncertainty on investment: a panel study on industrial countries. Oxf. Bull. Econ. Stat. 67 (3), 307–329. Carruth, A., Dickerson, A., Henley, A., 2000. What do we know about investment under uncertainty? J. Econ. Surv. 14, 119–153. Chirinko, R.S., 1993. Business fixed investment spending: modeling strategies, empirical results, and policy implications. J. Econ. Lit. 31 (4), 1875–1911. Demir, F., 2009. Macroeconomic uncertainty and private investment in Argentina Mexico and Turkey. Appl. Econ. Lett. 16, 567–571. Dixit, A., Pindyck, R.S., 1994. Investment Under Uncertainty. University Press, Princeton: Princeton. Elder, J., Serletis, A., 2009. Oil price uncertainty in Canada. Energy Economics 31, 852–856. Engle, R.F., 1982. Autoregressive conditional heteroskedasticity with estimates of the variance of UK inflation. Econometrica 50, 987–1008. Engle, R.F., Lee, G.J., 1999. A long run and short run component model of stock return volatility. In: Engle, R.F., White, H. (Eds.), Cointegration, Causality and Forecasting: A Festschrift in Honour of Clive W.J. Granger. University Press, Oxford: Oxford, pp. 475–497.

563

Episcopos, A., 1995. Evidence on the relationship between uncertainty and irreversible investment. Q. Rev. Econ. Financ. 35 (1), 41–52. Fatimah, A., Waheed, A., 2011. Effects of macroeconomic uncertainty on investment and economic growth: evidence from Pakistan. Transit. Stud. Rev. 18 (1), 112–123. Federer, J.P., 1993. The impact of uncertainty on aggregate investment spending: an empirical analysis. J. Money Credit Bank. 25 (1), 30–48. Federer, J.P., 1996. Oil price volatility and the macroeconomy. J. Macroecon. 18 (1), 1–26. Fuss, C., Vermeulen, P., 2008. Firms' investment decisions in response to demand and price uncertainty. Appl. Econ. 40 (18), 2337–2351. Goel, R.K., Ram, R., 1999. Variations in the effect of uncertainty on different types of investment: an empirical investigation. Aust. Econ. Pap. 38 (4), 481–492. Guo, H., Kliesen, K.L., 2005. Oil price volatility and US macroeconomic activity. Federal Reserve Bank St. Louis Rev. 87 (6), 669–683. Haque, A., Shaoping, W., 2008. Uncertainty and investment evidence from a panel of Chinese firms. Struct. Change Econ Dyn. 19, 237–248. Hartman, R., 1972. The effects of price and cost uncertainty on investment. J. Econ. Theory 5, 258–266. Ibrahim, M.H., 2005. Sectoral effects of monetary policy: evidence from Malaysia. Asian Econ. J. 19 (1), 83–102. Ibrahim, M.H., 2011. Level and volatility of stock prices and aggregate investment: the case of Thailand. Glob. Econ. Rev. 40 (4), 445–461. Ibrahim, M.H., Said, R., 2012. Disaggregated consumer prices and oil price passthrough: evidence from Malaysia. China Agr. Econ. Rev. 4 (4), 514–529. Ingersoll Jr., J., Ross, S., 1991. Waiting to invest: investment and uncertainty. J. Bus. 65 (1), 1–29. Iwayemi, A., Fowowe, B., 2011. Impact of oil price shocks on selected macroeconomic variables in Nigeria. Energy Policy 39 (2), 603–612. Jayaraman, T.K., Keong, C.C., 2009. Growth and oil price: a study of causal relationship in small pacific island countries. Energy Policy 37 (6), 2182–2189. Jbir, R., Zouari-Ghorbel, S., 2009. Recent oil price shock and Tunisian economy. Energy Policy 37 (3), 1041–1051. Johansen, S., 1988. Statistical analysis of cointegrating vectors. J. Econ. Dyn. Control 12, 231–254. Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inferences on cointegration – with application to the demand for money. Oxf. Bull. Econ. Stat. 52, 169–210. Koop, G., Pesaran, M.H., Potter, S.M., 1996. Impulse response analysis in nonlinear multivariate models. J. Econom. 74, 119–147. Kumo, W.L., 2006. Macroeconomic uncertainty and aggregate private investment in South Africa. S. Afr. J. Econ. 74 (2), 190–204. Laopodis, N.T., Sawhney, B.L., 2007. Dynamic interactions between private investment and the stock market: evidence from cointegration and error-correction models. Appl. Econ. 17, 257–269. Lee, B.R., Lee, K., Ratti, R.A., 2001. Monetary policy, oil price shocks, and the Japanese economy. Jpn. World Econ. 13, 321–349. Mishkin, F.S., 1999. Lessons from the Asian crisis. J. Int. Money Financ. 18, 709–723. Ng, W., 2012. Oil price volatility and the Singapore macroeconomy. Singap. Econ. Rev. 57 (3), 1–26. Oladosu, G., 2009. Identifying the oil price- and macroeconomy relationship: an empirical mode decomposition analysis of US data. Energy Policy 37 (12), 5417–5426. Pesaran, M.H., Shin, Y., 1998. Generalized impulse–response analysis in linear multivariate models. Econ. Lett. 58, 17–29. Pindyck, R.S., 1991. Irreversibility, uncertainty and investment. J. Econ. Lit. 29, 1110–1148. Pindyck, R.S., Solimano, A., 1993. Economic instability and aggregate investment. NBER Macroecon. Annu. 8, 259–303. Pradhan, G., Schuster, Z., Upadhyaya, K.P., 2004. Exchange rate uncertainty and the level of investment in selected South-east Asian countries. Appl. Econ. 36, 2161–2165. Ramaswamy, R., Slok, T., 1998. The real effects of monetary policy in the European Union: what are the differences? IMF Staff Pap. 45 (2), 374–396. Serven, L., 2003. Real exchange rate uncertainty and private investment in LDCs. Rev. Econ. Stat. 85, 212–217. Sims, C.A., 1980. Macroeconomics and reality. Econometrica 48, 1–48. Sims, C.A., Stock, J.H., Watson, M.W., 1990. Inference in linear time series models with some unit roots. Econometrica 58 (1), 161–182. Stockhammer, E., Grafl, L., 2010. Financial uncertainty and business investment. Rev. Polit. Econ 22 (4), 551–568. Yeh, F.-Y., Hu, J.-L., Lin, C.-H., 2012. Asymmetric impacts of international energy shocks on macroeconomic activities. Energy Policy 44, 10–22.