An Econometric Investigation of
\nForecasting GDP, Oil Prices, and
\nRelationships among GDP and
\nEnergy Sources

In order for a policy to be effective, the links between the policy tools and the sub- sequent targets must be known, understandable, stable, and predictable. In this respect, this thesis is composed of three separate yet related empirical studies, that target important macroeconomic variables, which play a central role in the conduct of macroeconomic policies.
\n
\n
\nFirst, simple regression estimates and a factor-based model are utilized to produce forecasts for Bahrain quarterly GDP growth using quarterly data-set that cover the period from 1995: Q1 to 2008: Q3. To do so, we run a simple regression model using a small data-set including six explanatory variables carefully selected based on in-sample correlation with the target variable. Alternatively, a factor model based on 65 indicators is employed to forecast Bahrain GDP growth. Using simulated out- of-sample experiments we assess and compare the performance of both approaches. The main finding from our forecasting experiment is that the best forecasting performance can be reached using simple regression estimates with a carefully selected small set of variables. In particular, by looking at point and density nowcasts, we find that a simple regression model that use industrial production as an indicator are more accurate than a static factor approach that uses over 65 variables. The official flash estimates of Bahrain quarterly GDP are published with a delay of 90 days after the end of the reference quarter. Our flash estimates, based on simple regression model, reduce the lag to 54 days.
\n
\n
\nSecond, we aim to forecast West Texas Intermediate (WTI) crude oil prices using a large monthly data-set that cover the period from March 1983 to December 2011. To achieve this aim, forecasting with factor models offer a usual approach that utilizes large data-sets, however; a forecasting model which simply includes all factors in state space equation and do not allow for time varying may be not suitable with a highly volatile market such as oil market. To overcome these limitations, we employ an approach that accounts both for parameter and model uncertainty. In particular, we implement the the Dynamic Model Averaging (DMA) approach suggested by Koop and Korobilis (2012). The key element of the DMA approach is that it allows both for model and parameter to vary at each point of time. By doing so, the DMA is robust to structural breaks. Empirical findings show that DMA approach outperforms any other alternative model used in the forecasting literature. We show that there is model but not parameter variation. Finally, we find that the DMA
\napproach provides a better proxy of expected spot prices than future prices.
\n
\n
\nThird, the Johansen cointegration technique is used to examine the long-run relationship between oil consumption, nuclear energy consumption, oil price and economic growth for eight countries over the period from 1965 - 2010. The countries investigated are divided into two groups. The first group includes four industrialized countries: USA, Canada, Japan and France, while the second group includes four emerging economies: Russia, China, South Korea and India. Results suggest that there is a long-run relationship between the four variables. Exclusion tests show that at least one energy source enter the cointegration space significantly, which im- plies that energy a long-run impact on economic growth. The emerging economies found to be heavily dependent on both oil and nuclear energy consumption. We also examine the causal linkage between the variables through exogeneity test. There is evidence of a unidirectional causality between energy consumption (oil or nuclear) and economic growth in all investigated countries. Our findings have important pol- icy implications that should be taken into account in designing appropriate energy policies. Energy conservation policies might have drawbacks or damaging repercus- sions on economic growth for this group of countries.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.