Macroeconomic Impacts of Electricity Generation on Croatian Real GDP: Causality Analysis

According to Davis (2003), the non-farm enterprise (NFE) may be defined as comprising all those activities associated with waged work or self-employment in income generating activities (including income in-kind) that are not agricultural but which generate income (including remittances etc.). Rao (2013) uses this indicator because NFE’s are a significant generator of income in both rural and urban areas.

For the initial capital stock, we divided real fixed investment in the first period (1966—the first year of our analysis) with the sum of depreciation rate (5 %) and average growth rate of investment (Kyriacou 1991; Hall and Jones 1999). In order to increase the realism of the estimates, we divided the depreciation of new investment (namely, δ) in the standard linear PIM equation by 2 since new investment is assumed to be placed in service at midyear instead of at the end of the year. According to Kamps (2004), investment typically occurs throughout the year, not only at the end of the year.

One of the main indicators of country’s commitment to technological progress are expenditures for research and development (R&D). In addition, good output measures of R&D process (in terms of legal protection of the innovation process) are registered and approved patents (OEDC 2010). Accordingly, these indicators should be used as proxy variable for technological progress. Since the data on patent activity and R&D expenditures in Croatia are available since 1990 and 1997 respectively, we have taken into account several other variables that approximately reflect Croatia’s research and development capacity. As an input indicator of R&D activities, according to OECD (2002), we used the figures on legal scientific and research entities and total research, technical and other supporting staff as a convenient proxy for R&D personnel and R&D organizations. Especially for R&D personnel since this indicator perfectly complements R&D expenditures which are not available prior to 1997. In addition, we used the figures on published research works as an output indicator of R&D process (a sort of proxy for patents). Furthermore, we used the variables such as average years of schooling (as an aggregate measure of national economy’s level of education) and public expenditure on education with an argument that (highly) educated and skilled workforce is important in creating, acquiring, transferring and using relevant knowledge (Sundac and Fatur Krmpotic 2009). These proxies had to be statistically processed using several tests (Kolmogorov-Smirnov test for normality, linear regression and VIF (variance inflation factor) test for multicolinearity) to determine whether indeed all the variables could be used in the PCA analysis. In conclusion, a combination of PCA analysis and Cronbach alpha (c-alpha) test of internal consistency (Nardo et al. 2005) resulted in a homogeneous construct comprised of remaining two variables (total research, technical and other supporting staff and published research works) and with 83.04 % of explained variance. Detailed results on PCA analysis are available from the authors upon request.

We use several test statistics and most of them, namely Log LR (likelihood ratio) and Wald statistic, confirm the existence of a structural break in the year 1990. For lnGDP, the results of the test statistics are: F-statistic = 1.657946; Log LR = 79.59803; Wald statistic = 29.84303. For lnELECGEN, the results of the test statistics are: F-statistic = 1.707184; Log LR = 80.62788; Wald statistic = 30.72932.

The null hypothesis of no cointegration (H₀) against the alternative (H₁) for each equation is as follows: \( {\mathrm{H}}_0:\ {\delta}_{11} = {\delta}_{12} = {\delta}_{13} = {\delta}_{14} = 0\ \mathrm{and}\ {\mathrm{H}}_1:\ {\delta}_{11}\ \ne\ {\delta}_{12}\ \ne\ {\delta}_{13}\ \ne\ {\delta}_{14}\ \ne\ 0 \).

Most widely used methods to empirically analyze the long-run relationship and dynamic interactions between two (or more) variables include Engle and Granger’s (1987) two-step procedure and multivariate maximum likelihood based approach of Johansen (1988) and Johansen and Juselius (1990). These procedures however have some limitations: (a) Engle-Granger’s procedure is only appropriate when the analysis is carried out on two variables; (b) Johansen’s multivariate approach has a problem with the degree of freedom when applied to a small sample (Toda 1994).

Detailed results regarding lag order selection criteria prior to multivariate VAR estimates are available from the authors upon request.

The results remained unchanged when the analysis was repeated with the optimum lag set to 2. To conserve space we do not present these results but are available from the authors upon request.

According to the information given at the Energy Institute Hrvoje Pozar, last vulnerability analysis of the entire Croatian energy sector was carried out at the end of 2008, it included the time horizon from 1995 to 2006 and according to that analysis, the vulnerability indicator had an increasing trend.

Both big projects, Russia’s South Stream and Nabucco (which had the support of both the EU and USA but in the meantime will be replaced by the so-called Trans Adriatic Pipeline (TAP), which will deliver gas from the Caspian basin using 500 km shorter route) were designed to pass through Croatia. Participation in the South Stream was offered in 2007 but no answer was given to the Russians. Subsequently an agreement was concluded on Croatia’s connection on South Stream but outside the main route leaving Croatia without the income from gas transfer. Meanwhile, the representatives of Azerbaijan, Albania, Bosnia and Herzegovina, Croatia and Montenegro at the end of 2013 signed a Memorandum on cooperation regarding the realization of the Southern Gas Corridor in southeastern Europe. They also committed to work on the realization of all the preconditions necessary for the realization of the Adriatic-Ionic Pipeline (as a part of TAP).