Business cycles aggregation and oil shocks. A non-linear econometric approach

Resumen

Understanding the causes of fluctuations or aggregate shocks is one of the primary concerns of macroeconomic analysis. The study of business cycles has been important for economists since the beginning of the 20th century. One of the precursors in business cycle theories was J. A. Schumpeter in his book Business Cycles: A theoretical, historical and statistical analysis of the Capitalist process, published in 1939. He stated that the development process (which, fundamentally, is the activity of innovative businessmen) is not uniform; instead, it is cyclical, due to the fact that innovations are distributed discontinuously in time. According to Schumpeter, the existence of cycles of different periodicity is due to the time passing between the adoption of an innovation and the moment in which it becomes productive, which varies depending on the nature of the innovation. He distinguishes three kinds of cycles according to their periodicity: Kondratieff long cycles (with durations between 45 and 60 years), Juglar cycles (between 9 and 11 years) and Kitchin cycles (from 3 to 5 years). Nevertheless, since Schumpeter proposed this cycle theory -a theoretical-historical study- further econometrical advances have appeared and, nowadays, there are many variables used to examine the behaviour and nature of economic cycles. Romer (1996) considers four essential features that characterise economic cycles. The first is that fluctuations do not exhibit a simple cyclical regular pattern. The macroeconomy has failed in its attempt to interpret fluctuations as combinations of deterministic cycles with different durations because output fluctuations do not occur with regularity. The most generalized hypothesis is that an economy is affected by disruptions of different kinds and amplitude, with variable random intervals. Furthermore, these disruptions spread subsequently throughout the economy. However, there is no unanimity among economists with respect to the mechanism by which the shocks spread. The second feature is that fluctuations are unequally distributed across different output components. The third considers asymmetries in output fluctuations and maintains that large asymmetries do not appear between output increases and decreases; on the contrary, growth is distributed quite symmetrically around its mean. Nevertheless, second order asymmetries appear; consequently, output seems to be characterised by relatively long periods where it is slightly above its regular path, which are interrupted by shorter periods where it is below that path. Fourth, the nature of fluctuations differs between the periods prior to and after the Second World War: in the former period, the movements are broader and slightly less persistent. From a theoretical point of view, amongst the traditional macroeconomic models that consider the business cycle, a variation of the Ramsey model includes aggregate fluctuations that modify the model in two ways. Firstly, through technological shocks (variations in the production function from one period to the next) or changes in public consumption. Both fluctuations are real (as opposed to monetary or nominal ones), so these models are known as Real Business Cycle or RBC models. Secondly, introducing employment variations, which are determined by the intersection between labour supply and demand. However, other theories maintain that technological shocks and propagation mechanisms in RBC models do little to explain current fluctuations and that the key to analysing them are nominal fluctuations and the lack of adjustment of nominal prices and wages to those fluctuations. The models that study these mechanisms are known as Keynesian models and differ from the RBC models not only by including barriers to nominal adjustments but also in the analysis of the behaviour of the economy without those barriers. Consequently, interest in the business cycle has a long-standing history in both theoretical investigations and empirical applications. The important contribution of Burns and Mitchell (1946) paved the way for methods to measure it. Their work is crucial and can be considered the starting point of current works about the analysis and measurement of business cycles, involving the moments of selected variables, which are predicted by parametric models of the cycle. These authors describe the business cycle through a two-stage methodology. First, turning points are located in the series by using graphical methods, thereby defining specific cycles. Second, the specific cycle information is distilled into a single set of turning points that identify the reference (or aggregate) cycle. According to them, "Business cycles are a type of fluctuation found in the aggregate economic activity of nations that organize their work mainly in business enterprises: a cycle consists of expansions occurring at about the same time in many economic activities, followed by similarly general recessions, contractions, and revivals which merge into the expansion phase of the next cycle" They also define concepts such as peak (the high point of an expansion) and trough (the worst moment in a recession period) to determine the cycle length and commonly used in any work about business cycles undertaken after the publication of their work. Furthermore, they develop methodologies to determine the moments at which peaks and troughs appear and that frame economic recession or expansion. Their aggregate cycle was called the business cycle, and their tools were immediately used by the National Bureau of Economic Research (NBER) to deepen the understanding of US business cycles and, afterwards, being a reference for the study of business cycles in other economies. Nowadays, NBER continues publishing a single set of turning points for the US economy. This seminal work generated a great deal of literature, the main differential aspect being more the level of sophistication of the statistical tools than the definition of the business cycles. In the last few decades, many alternative procedures have been suggested. In this context, the aim of the present thesis is twofold. The first is to identify regional business cycles in Spain. There is great interest among economists in analysing business cycle behaviour. In recent years, the methodological literature focused on it has been prolific, with the application of non-linear approaches. Notwithstanding, these analysis are aimed at knowing the business cycle of countries but there is no great progress at more disaggregate levels; and depending on the purpose of the analysis, national disaggregation might not go far enough. We should point out that the cycle of a particular economic area represents the average cyclical behaviour of the different regions that compose it. Thus, it is possible that the adoption of particular economic policies has opposite effects to those desired for some regions, due to their cyclical behaviour being very different from the average pattern. Thereby, in spite of the growing economic and monetary integration, such as the recent creation of EMU in 1998, there is great controversy about whether the economic cycles of each country co-move enough to carry out these supranational projects. However, traditionally, only the degree of coordination between the business cycles of countries is taken into account in integration processes, without paying any attention to the links inside a national economy. But, how do the cyclical movements inside a country behave? Do they really determine a country's business cycles? The second objective is to analyse the effect of oil price shocks on the economic evolution of an area. One of the main aims in business cycle analyses is the characterization and explanation of the frequent fluctuations of real economic activity and their relationship with movements in other macroeconomic variables. A great deal of research has focused on the causes of business cycles and the identification of the shocks responsible for the fluctuations in output (see Cochrane (1994) for a review of these papers). Changes in oil prices are one of the most common aggregate shocks and the reason of our interest in examining the relationship between this variable and business cycle fluctuations . Furthermore, since the mid 1960s, oil has been the most important primary energy source all over the world. Almost all economic activities are based on oil, which represents around 40% of the total energy needs. The price of a barrel of crude is considered a reference point in the world economic system. Besides, movements in oil prices do not only affect energy markets but they also have an impact on the rest of the economy, even causing increases in inflation rates, modifying stock exchange markets and hindering economic growth. Bearing in mind these statements, this thesis "Business cycles aggregation and oil shocks. A non-linear econometric approach" intends to analyse two subjects that are very outstanding nowadays. On the one hand, it studies the possible existence of different business cycles inside a country, Spain, and how they are linked. On the other, it tries to understand the relationship between oil price shocks and the economic behaviour within two geographical areas: Spain (as a whole and its regional disaggregation) and the G7 countries. This PhD thesis is organised as follows. In the first chapter, the first objective mentioned above is developed, identifying the Spanish NUTS-2 regional business cycles and presenting their main features. In the second and third chapters, the links between an oil price shock and the macroeconomy are analysed. The second chapter goes deeper into the study of this relationship in the Spanish economy but also at a more disaggregate level: Spanish NUTS-2 regions, with all the restrictions that the scarcity of data availability imposes. The third chapter focuses on the G7, for a wider period than the previous chapter, to shed additional light on the aggregate behaviour of these developed economies when experiencing an oil price shock. The thesis presented intends to follow the new doctoral thesis guidelines, where the encyclopaedic knowledge of a particular theme is not the main aim. On the contrary, following the new guidelines, and because the publication of papers in international journals is of great importance, our aim is to present three original pieces of research with a common feature. The existing link between them is that they share a common methodology, that is, a non-linear time series approach. Moreover, we try to go more deeply, step by step, into the identification and better knowledge of Spain and its NUTS-2 regions' cycles through Markov Switching methodology and synchronization measures. Then, we focus on the impact of the oil prices on the economic evolution of Spain and its regions using the Qu and Perron (2007) and the Bai and Perron (1998, 2003a and 2003b) techniques. Afterwards, the results are generalized into a wider context, the G7 countries, using the previous methodology, but also computing two different procedures for impulse response analysis. This gives us useful information on the effect of oil price shocks on the rest of the economy. In the following lines, we explain each of the questions addressed in the three chapters that compose this thesis and the main conclusions drawn from them. REGIONAL BUSINESS CYCLES AGGREGATION: THE CASE OF SPAIN The objective of this first chapter is to detect the possible existence of regional business cycles in Spain and to analyse their relationships with the aggregate national cycle. The work tries to identify the cyclical patterns of Spanish NUTS-2 regions through a non-linear method and using, for the first time, the Industrial Production Index (IPI) series instead of the labour market variables that are traditional in these studies. It is expected that IPI data better reflects business cycle behaviour and so our results present a high level of information content. Once regional business cycles have been identified, the degree of synchronization is analysed in order to establish groups of regions and compare them with the aggregate Spanish cycle. We attempt to contribute to a better understanding of business cycle fluctuations within a country, explained through their regional cycles. The econometric tools are based on non-linear models with a changing regime, Markov Switching (MS). The use of these techniques has become more and more interesting for the identification of cyclical patterns of different economic aggregates. However, there are few studies in spite of the advantages of this methodology. The methodology has been mainly applied to countries, and hardly at all to more disaggregate levels, mostly due to the lack of available data. This work concentrates on the case of Spain, a country with suitable size characteristics, divided into seventeen regions, with a high degree of fiscal federalism and where the previous literature is really scarce. On the one hand, there are studies that analyze the Spanish business cycle compared to other countries, basically its main European neighbours (Camacho et al., 2005). On the other hand, there are studies that try to measure the cycle in Spain as a whole without going into regional behaviours and mainly using GDP or employment variables, the latter being less accurate indicators of economic activity than the industrial production index (Doménech and Gómez, 2005; Dolado et al., 1993; Dolado and María-Dolores, 2001). The exception -considering regional activity- are the papers of Cancelo and Uriz (2003) and Cancelo (2004) applied to Spanish regions but also using employment data. A first MS univariate analysis allows us to capture the stylized facts -length, probabilities and mean growth in each regime- characterizing the economic activity in Spain and its individual regions. A three-regime model, corresponding to recession, normal and high growth phases, was needed in Spain and in most of its regions to capture the cyclical paths accurately. Based on these results, we have studied the degree of synchronization among regions in order to identify common cycles and we have found three groups of Spanish regions. The first, a larger group formed by ten regions, exhibits a business cycle with the more probable and longest-lasting normal growth regime and two other regimes representing recessions and high growth, both with a lower and similar probability. The second group of Spanish regions, which traditionally present slow rates of growth and that receive cohesion and structural funds, is clearly identified. Meanwhile, the high growth period of the Cantabrian coast in the last fifteen years has determined the third group, characterized by the high weight of its industrial sector. The application of the MS-VAR model gives support to the previous results and estimates different business cycles with common regime shifts in the stochastic process of the industrial activity growth of each of the three groups. The Spanish cycle shows a high concordance with the first group and can be seen as a synthesis of these three regional groups, combining the more stable behavior of group 1 with the more extreme behavior of groups 2 and 3. Therefore, the result supports the view that not all regional economies are related to the national economy in the same way. Our findings have two policy implications. Firstly, the possibility of forecasting the Spanish cycle on the basis of regional analysis: once regional cycles are identified, the most representative regions, or those with more advanced cycles in each group, might be used to forecast Spanish economic activity. Secondly, the results are also important for economic policy because, if we distinguish different regional cycles, we should not apply a nationwide scope without focusing on particular areas. In other words, a centralized approach to policy making that does not consider localized measures to smooth regional fluctuations might not be the most suitable one. THE EFFECT OF OIL PRICE SHOCKS ON THE SPANISH ECONOMY According to the International Energy Agency (IEA), the oil dependency of Spain is stronger than that of other industrialised economies. In 2006, oil represented 58.1% of Spain's energy needs (12 pp above the European OECD countries), which, nevertheless, was a reduction of almost 20 pp with respect to the beginning of the 1970s. Furthermore, oil price shocks have differential effects on the different sectors of the economy, affecting more strongly the industrial sector, followed by the transport sector (which consumes other petroleum derivatives). Spain can be spatially disaggregated into 17 NUTS-2 regions with different weights of their industrial sectors. This situation together with the greater importance of the country's use of oil led us to study the influence of oil price shocks on the Spanish economy and its regions. In this chapter, we want to show the existence of a relationship between oil prices and economic activity for Spain as a whole and for its NUTS-2 regions. The relationship between oil price shocks and economic activity has been taken into account since Hamilton's (1983) contribution, where he demonstrates that oil price increases in the US have caused a reduction in the growth rate for the period 1948-1980. The results of this work have been tested and empirically confirmed by other authors in different scenarios. Nevertheless, the assumption that oil price shocks directly contribute to economic recessions remains controversial, in part because the correlation between oil prices and economic activity becomes weaker when considering data from the mid 1980s onwards (see Hooker, 1996). This relation has also been reported, mainly for the US economy, by Barsky and Kilian (2002, 2004), Bernanke et. al (1997), Blanchard and Galí (2008), Bohi (1989, 1991), Bruno and Sachs (1982), Davis and Haltiwanger (2001), Hamilton (1983, 1996, 2003, 2005), Hamilton and Herrera (2004), Hooker (1996, 2002), Lee and Ni (2002), Lee et. al (1995), Mork (1989), Mork et. al (1994), Raymond and Rich (1997) and Shapiro and Watson (1988), amongst others. We establish the effect of oil price shocks on macroeconomic variables (GDP and CPI inflation) for the longest available period (1970-2008 for the Spanish economy and 1980-2008 for its regions). Allowing for the presence of different periods, for Spain, we use recent methodological advances for finding structural breaks such as the Qu and Perron (2007) procedure where the breaks are selected endogenously considering all the model parameters and, for its regions, we apply the methodology of Bai and Perron (1998, 2003a, 2003b) that tests for the presence of structural breaks in the relationship between oil prices and each of the two macroeconomic variables considered. The use of these procedures confirms the existence of a non-linear relationship between oil price shocks and the macroeconomic variables. The breaks obtained for Spain as a whole and for the NUTS-2 regions fit the historical economic record quite well. The influence of oil shocks has been estimated though the use of long-term multipliers for the different periods identified and for each geographical unit. Our results reveal two facts. Firstly, in Spain, after 1970s, the estimated response of macroeconomic fluctuations to oil price shocks decreases but, in the late 1990s for the GDP and, especially, in the 2000s for the CPI, the influence of oil shocks recovers some of its initial importance, although the impact is smaller than in the 1970s. Secondly, in Spain's regions, the effect of oil price shocks on production and inflation loses importance progressively. However, for inflation, commencing in approximately 1995, the influence recovers some importance. This also occurs in Spain as a whole, but beginning about five years later. Thus, the most outstanding result of our research is that the oil price fluctuations could explain at least some of the recent inflation. The level of disaggregation and the sectoral distribution appear to be important for the understanding of the economic behaviour when faced with an oil shock, at least for the GDP. This analysis is important for the design of adequate economic policies in the face of new oil price shocks that consider the differential geographical behaviour of some regions or that are more orientated to specific sectors. ECONOMIC GROWTH, INFLATION AND OIL SHOCKS: ARE THE 1970S COMING BACK? The origin of our interest in examining the relationship between oil prices and business cycle fluctuations is that changes in oil prices are one of the most commonly identified aggregate shocks, as we have mentioned earlier. On the theoretical side, some papers, such as Rotemberg and Woodford (1996), Finn (2000) or Kilian (2006), have constructed models to account for the effects of oil price shocks. More recently, the study of the relationship between oil prices and the economy has been carried out through DGSE models that incorporate global and domestic energy markets (see Crucini et. al (2008), Nakov and Pescatori (2007) or Blanchard and Galí (2008)). On the empirical side, much of the literature has focused on the US, for example Hamilton (1983, 1996), Bernanke et. al (1997), Baumeister and Peersman (2008) or Kilian (2008b). Another line of research focuses on the evaluation of the relationship between oil price shocks and inflation and output since the 1970s, the common main finding being a waning effect of oil price shocks in the most important economies since 1980 (see Jiménez-Rodríguez and Sánchez (2005), Kilian (2008a) or Blanchard and Galí (2008)). In this context, the main objective of the last chapter is to determine the influence of oil shocks on G7 economies (on inflation and economic growth) between 1970 and 2008, considering the presence of different periods. We apply the recent technique of Qu and Perron (2007) and we also systematically assess the magnitude, the length and the differences and similarities in the response of the G7 economies to exogenous oil price shocks. This analysis is carried out in order to understand the historical record and to design adequate policy measures to control or smoothen the effects of future energy shocks on an economy. We find that, whereas the evidence of a temporary reduction in the impact of oil prices on GDP and CPI is consistent across all seven countries until the late 1990s, from then on, the impact on inflation (and, to a lesser extent, on GDP) is less clear-cut. The results suggest that the response of output and prices becomes weaker from 1970 (when it shows its greatest responses) until the late 1990s. This confirms the results previously obtained in the literature while, in clear contrast to the previous research, in the 2000s, the impact of oil prices on the macroeconomic variables recovers some of its initial importance. Notwithstanding, the impact is smaller than in the 1970s, although it does allow us to confirm that the recent strong variability of oil prices has transmitted a minimum effect, mainly on prices. Our results could open up a line of future research focused on the identification of the possible causes of this renewed effect of oil prices. An adequate and precise characterization of the features studied in this paper (magnitude, length and differences in the response of G7 growth and prices to oil price shocks) is crucial for the implementation of policy measures to control the effects of future oil price shifts.