Growth, institutions and oil dependence: A buffered threshold panel approach

Highlights

• We examine the combined effects of oil-dependence and the quality of institutions on growth.

• A new buffered threshold panel data model is applied to 19 oil rent-countries over 1996–2017.

• There is a gradual positive impact of oil rents on growth as the quality of institutions enhances.

• The quality of the institutions has a positive effect on growth when dependence is low or high.

• For intermediate levels of dependence, the quality of the institutions negatively impacts growth.

Abstract

We examine the combined effects of oil dependence and the quality of institutions on economic growth. To do so, we introduce a new buffered threshold panel data model and apply it to 19 oil rent-dependent countries over the period 1996–2017. We show that the relationship between growth and oil-dependence is not linear. More precisely, three categories of oil-dependent countries are identified. Only countries with high-quality institutions are very stable. All the other countries have experienced a transition into a buffer zone and are potentially in a transition between two different regimes. When considering oil dependence as a threshold variable, it appears that the quality of institutions has a positive and significant effect on growth when dependence is either low or high. More interestingly, for countries with intermediate levels of oil-dependence, the quality of the institutions negatively impacts growth. Some of these countries have experienced something of an oil-dependence trap.

Introduction

Dependence on natural resources is the subject of a wide debate in the analysis of economic growth in rentier states. However, in the empirical studies, there is no clear consensus on the negative impact of resource rents on long-term growth (Havranek et al., 2016). In practice, rentier states are characterized by important heterogeneity in their economic performance. The quality of institutions is one major explanation that has been advanced in the literature to explain these disparities. As is now well documented in the literature, natural resource dependence has given rise to some negative phenomena that could hinder growth (rent-seeking behaviours, the contraction of non-resource production activities, corruption, the voracity effect, civil conflicts, social pressure for additional redistribution, increases in public spending in less productive sectors, etc.). In fact, a diversification of the economy and an improvement in the quality of the institutions in natural resource exporting countries seem to be efficient tools for enhancing their growth performance. Indeed, these countries could reach such high levels of dependency that it would become very difficult to sustain good economic or institutional reforms. From an economic policy point of view, it is thus important to understand how institutional reforms could impact economic growth while interacting with natural resource dependence. Indeed, the economic cost of ameliorating the quality of the institutions could be very high before having a positive effect on economic growth.

The empirical literature studying the relationships among natural resources, the quality of institutions and economic growth has not brought about a consensus. Such studies can be roughly classified into three categories. In the first category, natural resources are found to have a negative effect on growth when they are associated with weak institutions (see for example, Leite and Weidman, 1999, Acemoglu et al., 2001, 2002, and Sala-i-Martin and Subramanian, 2013). The second category shows that natural resources interact with the quality of institutions, and their combined effects on growth depend on the nature of their combination (Mehlum et al., 2006a, b, Boschini et al., 2007, Arezki and Van der Ploeg, 2011). The third and last category shows that the observed heterogeneity in economic growth between rentier states is not explained by institutions (Sachs and Warner, 1999; Brunnschweiler, 2008; Alexeev and Conrad, 2009). It is worth noting that this literature generally assumes linearity in the dynamics to address these rather complex relationships. Only a few contributions have insisted on their nonlinearity (Leite and Weidmann, 1999; Sala-i-Martin and Subramanian, 2013).

Since their introduction by Tong (1978), threshold models have been considered to be a very useful and sophisticated way to take into account the nonlinearity exhibited in several financial and macroeconomic phenomena. Indeed, they provide a simplified formulation to mimic nonlinear stylized facts and, more precisely, the dynamics of regime changes. Their structure has been widely used by econometricians in time series analysis. However, many extensions and mathematical developments of threshold models, in particular the panel data treatment framework, have been adopted for the analysis of other data structures. Hansen (1999) proposed a panel threshold regression (PTR) model for the nondynamic panel case. His main contribution lies in the possibility of allowing the individuals constituting the panel to be in different regimes during a given period. This enables the heterogeneity in the panel to be better captured and allows for a visualization of the nonlinearity in the interaction between the dependent variable and the explanatory variables for each panel's component. However, the sudden change in regime that characterizes Hansen's formulation may be problematic in some situations in which the transition is smooth. To capture the absence of sudden jumps, Gonzalez et al. (2017) develop a non-dynamic panel smooth transition regression (PSTR) model with individual fixed effects. The parameters are allowed to change smoothly as a function of the threshold variable. The performance of this model may depend on the choice of the transition function for the studied phenomenon. Overall, this form of modelling turns out to be useful when the number of regimes is sufficiently high.

In some circumstances, an interesting phenomenon happens when a past temporary change in a relevant forcing variable leads to a change in the economic behaviour of the analysed variable but a return to the initial value of the forcing variable does not induce a return to the initial behaviour (i.e., the state of a system is dependent on its history). This so-called hysteresis phenomenon, originally stemming from physics, has been widely used in labour theory and foreign trade to explain the persistent effects of temporary stimuli (see, e.g., Göcke, 2002). For example, in foreign trade, temporary exchange rate shocks could induce persistent consequences for quantities and prices due to sunk market-entry costs. Indeed, to sell in a foreign market, a firm incurs some entry costs that cannot be recovered after exit (e.g., distribution and service networks). If the domestic currency temporarily depreciates, entering this foreign market becomes profitable for some domestic firms. However, even if the exchange rate regains its initial level, it is still profitable to sell in the foreign market if the variable costs are recovered. This simple microeconomic hysteresis can thus be aggregated and gives rise to a continuous macro-level loop in overall exports (Borgersen and Göcke, 2007). This effect has been widely documented in the empirical literature (see Belke and Kronen, 2019, for a recent study).

This smooth switching between different equilibria (finite configurations or finite states) of the studied variable (system) could thus be usefully mobilized to analyse the dynamics of its evolution. This is why in this article, we propose an alternative model based on this idea of hysteresis by defining a new smooth and flexible regime switching mechanism. To illustrate and highlight this point, let us limit ourselves, without loss of generality, to the case of a two-regime model. Instead of assuming a single threshold parameter, we consider an interval consisting of a lower and an upper threshold that acts as a buffer zone. If the threshold variable is below the lower boundary of the buffer zone, then the observation is from the first regime. Conversely, the observation comes from the second regime when the threshold variable is above the upper boundary. When the threshold variable falls within the buffer zone, the regime indicator keeps the value of its most recent past. This makes the transition dynamics smoother and more flexible than those of the classical PTR model. Even though this idea is still in its infancy, it provides a new way to understand and explain the nonlinearity observed in the data. In addition to this new modelling approach, our paper provides several interesting empirical results. First, we clearly show that the relationships between growth and dependence on oil rents are not linear: there is a gradual positive impact as the quality of institutions is enhanced. More precisely, our analysis identifies three categories of oil-dependent countries with respect to the quality of institutions. It is worth noting that except for three countries in the sample with high-quality institutions that are very stable, all the other countries have experienced a transition into a buffer zone. They are thus potentially in a transition between two different regimes, and the impact of oil resource dependence on their growth has not yet stabilized. Moreover, while considering dependence on oil rents as a threshold variable, it appears that the quality of institutions has a positive and significant effect on growth when dependence is low or high. More interestingly, it turns out that for intermediate oil-dependent countries, the quality of their institutions negatively impacts growth. Some of these countries have experienced something of an oil-dependence trap.

The remainder of the paper is organized as follows. In Section 2, we provide the analytical framework for our model: the definition of our buffered threshold panel data (BTPD) model, our estimation methodology, different test procedures and a simulation study. In Section 3, we first provide and discuss the results of our empirical study, which is devoted to the analysis of the combined interaction effects of natural resource dependence and the quality of institutions on economic growth in rentier countries. We thus compare our results to those provided by some alternative models and show how our model gives better results. Section 4 concludes.