THE ESTIMATION MODEL

126

residual does not imply that the model exclusively needs an impulse dummy. A large outlier could indicate a shift in the level of one or more variables. Hence, the appropriate procedure is first to examine whether there has been a shift in the equilibrium mean (using a step dummy) and, if so, to estimate the model with such a shift plus additionally an impulse dummy (blip dummy) in the short-run part of the model. If the step dummy is insignificant, then only the impulse dummy can be included in the model. Given that the interventions in the Iranian economy have been very significant, a priori one would expect to see changes in the equilibrium means. When appropriate, therefore, dummy variables are included in the empirical models to capture their associated effects on the models (see Section 5.5.2). In what follows, the estimation investment model for the oil-based economy of Iran is explained and a number of hypotheses are stated.

127

investment and alternative measures of oil-driven uncertainty. In resource-rich economies, the relationship between natural resources and economic activities is yet subject to debate. Some scholars argue that there is a negative relationship between the abundance of natural resources and economic performance. However, others suggest that natural resources are neither curse nor blessing, and that various factors such as low levels of human capital, debt overhang or poor political and financial institutions are among the core issues causing crises (Sachs and Warner, 1997; Lederman, 2007). However, the capital stock equation could be considered as a capital accumulation identity equation. That is, if the capital stock is measured by cumulating the next investment flows, it is identically true that it will be related to investment and deprecation;

hence it is not a behavioral equation. Accordingly, this study chooses to focus merely on the investment function in the empirical analysis.⁷⁰

A common approach in the early studies was to stipulate a linear relationship between the changes in oil prices and economic performance (Darby, 1982; Hamilton, 1883). The oil price collapse of the 1980s spurred research efforts to derive new specifications that produce a more responsive oil-macroeconomy relationship, one of which was the notion of asymmetry in the economy’s response to positive and negative oil price changes (Hamilton, 1996; Mork, 1989).

Early studies on net oil importing economies show that oil price increases and decreases are associated with significant recession and insignificant boom, respectively (Mork, 1989, 1994;

Mory, 1993; Hamilton, 1996). Asymmetric responses could be different in net oil exporting economies where positive and negative oil shocks may have significant aggregating and insignificant dampening effects, respectively, on the economic activities of these countries (Eika and Magnussen, 2000; Jimenez-Rodriguez and Sanchez, 2005; Korhonene and Ledyaeva, 2010;

Dissou, 2010; Gausden, 2010; Mendora and Vera, 2010).

Intuitively, an increase in oil windfalls in an oil-dependent economy relaxes foreign exchange constraints and stimulates government expenditures. For instance, Talvi and Vegh (2005) argue that in countries where the revenue base is highly fluctuating, budget surpluses create political

70 The capital stock accumulates net investment flows, that is, gross investment net of depreciation, but it must be noted that depreciation could itself be a function of economic variables, as it almost certainly is; this imparts a time subscript to depreciation, which becomes behavioral. The measurement of the capital stock may vary depending on the method of construction, i.e., the perpetual inventory method (PIM) and the capital accumulation identity (CAI).

The PIM usually assumes a constant depreciation rate; the actual capital stock may – or may not – have a constant depreciation rate. For the two, PMI and CAI, to be the same, the depreciation rate for the PIM must match the actual depreciation rate for the CAI.

128

pressures to increase government spending. In this picture, positive oil shocks can induce faster growth in government spending. However, it is often less recognized that this stimulating effect could only be transitory and that such blessing could turn into a curse in the long-run. This is because a sudden inflow of oil windfalls could result in the appreciation of real exchange rate, inflationary pressures on the economy and contraction of tradable sectors including non-oil exports. Furthermore, the notion of asymmetric response to oil revenue changes (rather than oil price changes) may arguably be of greater importance for oil exporting economies where oil revenues have been and are expected to be a crucial feature of their economies. Also, some studies found a stronger economic impact from the volatility of oil prices than the changes in oil prices (Mohaddes and Pesaran, 2013). Therefore, the long-run investment equation is augmented with measures of oil, namely oil revenues and oil price volatility (see Section 5.4.4 for the construction of oil-driven financial constraint measures).

Accordingly, employing the CVAR methodology, the stated relations (equations 4.25-26) are tested for the Iranian sample during the period from 1974 to 2011. Table 5.1 outlines the hypotheses which are divided into two parts. The first part reports hypothesis H1 under the heading of ‘Baseline Investment Equation’. This hypothesis predicts that, in the long-run, investment (it,) is positively related to output (yt) and the sum of the growth rate of capital (g^kt) and capital depreciation (δ), while it is negatively related to the user cost of capital (ct,) as implied by substituting the steady-state condition into the FOC equation.

Table 5.1 Hypotheses of long-run relationships Baseline Investment Equation

H1 Long-run relationships between [it, yt, ct, ln(g^k + δ)t, dpt]

Investment Equation Augmented with Symmetric Oil-driven Measures H1.1 Long-run relationships between [i^t, yt, ct, ln(g^k + δ)t, dpt, orevt] H1.2 Long-run relationships between [i^t, yt, ct, ln(g^k + δ)t, dpt, volot]

Note: it: investment; yt: output; ct: user cost of capital; δ: capital depreciation rate; g^k: capital growth rate; orevt: oil revenues; volot: oil volatility; and dpt: inflation as measured by the changes in the implicit deflator of gross domestic product (percent). Data are in natural log and in real terms (Base Year 2004/05). Source: CBI, Time-series Data; See also Section 5.4.

The second part of Table 5.1 presents two hypotheses H1.1-H1.2, under the heading ‘Investment Equation Augmented with Symmetric Oil-driven Measures’, investigating the long-run relationships between aggregate domestic investment and oil-based measures in the Iranian economy. Inflation (dpt) based on the changes in the implicit deflator of gross domestic product

129

is further included in the models to act ‘as a proxy for the (missing) market interest rate’

(Esfahani, et al., 2009, p.1). As discussed in Chapter Two, ‘the theoretical literature leaves open the sign and persistence of any relationships between investment and uncertainty’ (Bond, et al., 2005, p.10). Thus, it is of interest for this thesis to empirically estimate the relationship between oil-based uncertainty and aggregate investment in Iran, using symmetric oil-driven financial constraint measures as explained in Section 5.4.5.

As mentioned in Chapter Four (Section 4.4.1), the CVAR model used in this chapter is specified in terms of a vector ‘x’, comprising endogenous and exogenous variables, and deterministic terms including constant term and dummies ‘D’. In VECM notation, the cointegrating vectors are included in the dynamic specification given by:

(5.2) Δxt = Πxt-1 + Γi(L)Δxt-i + ΦDt + εt,

where L is the lag operator, Π = αβ', and x is a matrix of I(1). In the macroeconomic analysis of small open (oil-based) economies like Iran, it is plausible to assume that some variables are exogenous, implying that these variables have a direct contemporaneous impact on the endogenous variables, but they are not affected by the error correction terms which are the disequilibria in the economy. Dt is a set of variables weakly exogenous in the long-run cointegration space, and may contain deterministic terms such as constant and trend as well as intervention dummies.⁷¹ The error term εt is thus partitioned to εt = (ε'x*t, ε'zt)'. For instance, in the baseline VECM model, there are five endogenous variables, with two cointegrating vectors (it, yt, ct, ln(g^k + δ)t, dpt) and the long-run matrix can be decomposed into the following reduced rank form:

(5.3) αβ'xt = [

α11 α12 α21 α22 α31 α32 α41 α42 α51 α52]

[β11 β21

β12 β22

β13 β23

β14 β15 β24 β25]

[ i 𝑦c ln(g^k+ δ)

dp ]

,

71 In this chapter, there exist linear trends in the level of variables, but the linear trends in the variables do not cancel in the cointegrating relations, i.e. the models contain trend stationary variables or trend stationary cointegrating relations. Therefore, the trend is restricted only to appear in the cointegration relations, but the constant is unrestricted in the model (see case four in Juselius, 2006, p, 100).

t

130

(5.4) Γi(L)Δxt-i = ∑^k−1_i=1 Γi(L)Δ(i, y, c, ln(g^k + δ), dp)'t-i + εt,

where β defines the cointegrating vectors and α is the response of each variable to the cointegrating vectors as defined above.