![Sample our Area Studies journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5922/TOC-Subject-Sample-2021-Area-Studies.jpg"})
Abstract
The similar evolution of GDP and savings, the high level of natural resources, and the high financial surplus of Saudi Arabia motivate us to examine the linkages between economic growth and savings from 1980 to 2012 in the ARDL framework. By incorporating relevant determinants, we find cointegration among the variables and positive two-way Granger-causality between economic growth and savings over the long- and short-run. Authorities should thus pursue policies that promote economic growth and savings, to achieve higher levels of both. The results are robust to using alternative specifications and econometric procedures, and show stable economic growth and savings functions.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes
1. The advantages of the ARDL procedure compared to other techniques are discussed in Section 3.
2. In that regard, Lin (Citation1992) argues that savings leads to economic growth if resources are translated into capital accumulation.
3. Gross domestic savings is deflated by the consumer price index to get the real measure.
4. Inclusion of these channels enables us to avoid biased results, as the study is based on a more generalized specification.
5. Considering three different financial development indicators aims to check the robustness of the economic growth-savings nexus to alternative specifications, implying that Saudi authorities could make sound policies and strategies to enhance economic growth and improve savings.
6. In this situation, Bahmani-Oskooee and Nasir (Citation2004) recommend using the error correction term to establish cointegrating linkages.
7. The data on the consumer price index used to compute RGDS are gathered from the United Nations Conference on Trade and Development and are not available before 1980. This leads to the non-availability of RGDS over a longer period, which dictates the choice of the data sample. In contrast, the time span we consider seems to be long enough to conduct a cointegration and causality analysis in the ARDL framework.
8. Narayan and Popp (Citation2013) find that the NP test performs better than other extant unit root tests with two unknown breaks in terms of size, power, and selection of the two structural changes. Note that for the first time in the literature on the economic growth-savings nexus, a break detection process is considered when testing for unit root in the variables.
9. This conclusion is in line with that of Nelson and Plosser (Citation1982), who point out that most macroeconomic aggregates become stationary in first-differences.
10. For the cointegration analysis and computation of the long- and short-run coefficients in the ARDL framework, the orders of the corresponding models are selected optimally by the Schwarz Bayesian criterion (SBC) by considering a maximum permitted number of lags of three. The choice of three lags avoids the loss of freedom degrees, as we use annual data over the 1980–2012 period.
11. The existence of cointegration among the variables excludes eventual estimated spurious regressions.
12. In the long-run, economic growth and savings can drive each other, and the auxiliary determinants are considered as forcing variables for them.
13. The application of the Johansen (Citation1988) cointegration approach relies on I(1) variables, which is possible in our case if we refer to the results of the NP test, which is more effective than standard tests when structural changes exist in the variables.
14. Within the same context, Ozcan, Gunay, and Artac (Citation2003) and Kelly and Mavrotas (Citation2008) show evidence of positive linkages between savings and financial development.
15. In the same context, Kónya (Citation2005) finds evidence of causality from economic growth to savings in Saudi Arabia.
16. For a variable where the number of coefficients is higher than 1 in the error correction model, the impact of such a variable on the dependent variable is computed as the sum of its lagged coefficients.
17. This may be explained largely by the high correlation between the three indicators, as shown in the initial data assessment.
18. The statistical significance and negativity of the error correction terms confirm the stable long-term linkages among the variables, which are depicted by the ARDL bounds testing approach to cointegration (see Table ).
19. The CUSUM and CUSUMSQ stability tests are computed on the basis of recursive regression residuals.
20. The exception is that for the case where the RGDP is used as a dependent variable for the last financial development indicator, there is evidence of serial correlation at the 5% significance level.
21. The exception is that for the case where the RGDS is considered as a dependent variable for the first measure of the financial sector, the CUSUMSQ test is slightly above the upper 5% critical bound.
22. The causality helps implement economic policies and strategies according to the causal direction (see Deaton Citation1995).
23. This empirical finding is consistent with some of the above-mentioned theory-based hypotheses about the positive connection between economic growth and savings such as, for example, Modigliani (Citation1970) and Carroll, Overland, and Weil (Citation2000).