An Empirical Investigation into the Relationship between Investment and Economic Output / Growth

For the years 1970-2004 we obtained data from the Penn World Tables (PWT) for the following variables in Greece:

• Population (000s) – POP
• Real Gross Domestic Product per Capita ($) – cgdp
• Investment Share of cgdp (%) – ci

The Real GDP/Capita data was obtained by PWT from the World Bank which had calculated the data. We chose to use Real GDP/capita as it takes into account the effects of inflation.

Final Data Table Modifications:

From the original data we removed the country and isocode columns as we felt they were superfluous to our requirements.        

We then did the following equations:

adj. POP = POP*1000.

total gdp = adj. POP x cgdp

total Investment = total gdp x adj. ci

%Change in GDP: %Change=((gdp1/gdp0)–1)x100

%Change in Investment=((i1/i0) – 1) x 100

LN(Investment)=LN Investment

LN(GDP)=LN GDP

These were done to give us better data with which to analyse and reach more effective conclusions.

Data Reliability:

The PWT gave the data quality from Greece a Grade B suggesting the data is fairly reliable.

The reason that Greece’s rank is not an A is because in the income category it is quite low. However because Greece scored highly on the other measures used for data quality we will assume it is valid for our research. In addition, when compared with other data, the results appeared to be similar, suggesting our data is relatively reliable. Furthermore, in the data measurement methods remained constant, were contiguous, and without missing entries. Nevertheless, it may be noted though when we calculated the percentage change in both GDP and Investment they were slightly higher than expected, so there may possibly be overestimations in our data.

Data Description

Summary of data

---

From these graphs our data appears to demonstrate a very strong correlation. Firstly fig. 1 demonstrates a general increase in both Investment and GDP figures since 1970. Perhaps the most notable graph though is fig. 2, as clearly changes in both Investment and GDP appears to change in remarkably similar directions. That Investment appears more subjective to these changes is most likely due to the ‘accelerator effect’ taking place in the Greek economy. Using Stigliz’s (2000) definition, during periods of high growth an Investment bubble occurs (such as the period before 1973) which causes more growth; therefore sparking further Investment. However the reverse effects caused by the bubble bursting are also evident in 1975 and 1988. This is why appears GDP to follow a similar pattern, be it less extreme.

Looking at fig. 4, the data appears to generally correlate very well, with the two variables increasing proportionately.  There are exceptional outliers such as in 1987 and 1992, when the GDP change is higher than Investment change. Both of them have a higher GDP than the previous year, but lower Investment – Investment was $19.9 million and $16.3 million in 1986 and 1987 respectively. Similarly, in 1991, Investment was $24.1 million whereas in 1992 it dropped to $22.9 million. Using figure 3 this is supported as in 1991, the growth of GDP was 2.81% but the Investment was -4.87%, and 1987 experienced the lowest Investment growth of -18.05%, but GDP grew at a rate of 1.15%. Despite these outliers (fig.3), the general data correlates very well. Low/High Investment growth rates are concurrent with low/high GDP growth. An example of this is 1973, where Investment grew at a rate of 21.73% and GDP grew at 14.45%.

Fig. 5 and 6 also shows that a lower lending rate attracts higher Investment growth and GDP, due to lower cost of borrowing; hence attracting business to expand. Investment, as identified in the introduction, is only part of the AD equation. Consumers spending will also increase under lower interest rates. Using an IS/LM analysis it shows that a lower interest rate – a monetary policy.

The extent to which Investment growth would encourage GDP growth depends largely on Investments share of GDP. (Greece’s ranges from around 25% to 40% - a substantial proportion of GDP). There are other factors which affect Investment; corruption, cost, consumers’ spending and government spending. Government spending is negatively correlated with Investments. Higher Investment is linked to lower government spending, and vice versa, agreeing to the idea of crowding out – when government borrowing displaces private Investments, causing high interest rates and less Investment.

Correlations and Regression

 

 

        

The figure of 0.959 from the correlation coefficient shows there is clearly a strong positive relationship between Investment and GDP. The trendline above reinforces this pattern; an increase in Investment coincides with increases in output. The imperfections in the correlations are likely due to the few years when Investment falls and GDP grows: as there are other factors affecting output, namely government expenditure and consumption. Both the logged and non-logged data correlate to almost the same degree; suggesting that the data is almost perfectly linear.

The correlation coefficient for % change GDP against % change Investment is lower but still positive at 0.635. This is probably due to the more dynamic changes in Investment, such as business confidence; whereas GDP is unlikely to increase at such varying rates. This can be seen in the data tables where GDP growth remains positive throughout the period 1970-2004, whilst Investment has high fluctuations from positive to negative growth.

Unfortunately, there are negative changes in Investment so we couldn’t log this data to see if any linearity occurs.

We can see that the LN GDP and LN Investment have an 0.91 R Square, marking a strong correlation. The positive intercept shows even when no Investment is present, GDP would still be positive (from other aforementioned factors). The coefficient of Investment being greater than 1 also dictates an increase in Investment has a multiplied increase in GDP; from the theory that both the AS and AD curves will shift.

The data below suggests that GDP would be below 0 if there was no Investment. As this is unlikely, we suspect is a mere anomaly in the tables as other variables such as consumption are not accounted for in the regression analysis.

 

The regression analysis for growth of both GDP and Investment provides a much weaker R squared value of 0.403. Similarly the coefficient of 0.283 suggests GDP growth is only weakly affected by Investment. However, the high intercept value of 5.29 suggests that there is always a high level of growth in Greece even if Investment does not increase. I suspect the reason this analysis being less accurate is because of Investment fluctuation; a reduction in Investment does not immediately lead to a proportional fall in GDP. By assessing the causality we’ll be able to get a view of how strongly Investment affects GDP.

Assessing Causality

We have decided to use 10% significance level, since it is commonly used and provides a good insight into the quality of the data. We will also analyse the R Square Value to see how closely our values follow the regression line.

By looking at the GDP Growth/Investment Growth against Time graph (fig. 2, section 3); there is an indication GDP Growth peaks sooner than Investment Growth, at a lag of approximately 1 year, but they rise and fall at approximately the same point. We have, therefore, decided to formulate three years lag for both factors to see which is the lead variable.

We first ran the regression of GDP Growth against lagged Investment Growth. We placed GDP Growth as the Y-value and the Investment Growth as X-values. This shows a correlation (R Square Value) of 0.37149; implying a positive correlation but is weak. When we compare the P-Value of the different lags, it soon becomes apparent the best lag is 0 year with a P-Value of 0.006382, the others exceed the 10% levels. Subsequently, we lagged GDP Growth to assess any correlation. The correlation is a medium 0.56516; 0, 1 and 3 years lag have the best Significance Level, suggesting a higher possibility there is a relationship between lagged GDP Growth and Investment Growth. However both regression models suggest that 0 year Lag is the best fit, as in both cases it has the best significance level.

We have also done a similar test with GDP and Investment. We first tested the lagged GDP(X-value) with Investment(Y-value). It gives a correlation of 0.976425, which is very close to perfect correlation. Both the 0 year lag and the 3 years lag are shown to be inside the 10% significance level, however the 0 years lag has the better significance. We also did regression on GDP and lagged Investment. The correlation gives 0.961267, with only 3 years of lag being within the 10% significance level. This has shown that it is more likely a GDP lag. However, the significance level of 0 years lag of GDP is the highest, suggesting GDP has the best relationship with Investment when there is no lag.

Conclusion

We feel (although perhaps certain outliers weaken our conclusions) generally our data for Greece supports our hypothesis that Investment and economic growth are strongly related. Though we have focused this essay on whether there is indeed a correlation, we also feel it is likely the reasoning behind the relationship is largely due to the economic arguments we gave at the start of the essay. It would be fair to suggest the strength of the accelerator effect in Greece has also helped in compounding this relationship. What perhaps is more interesting is that our data is too inconclusive to suggest any causality between GDP and Investment. This may be because modern economies such as Greece and the firms inside them react much quicker than our year lags to changes in Investment or GDP. That additional research was able to suggest Investment in Greece has a stronger correlation with GDP than government spending, consumer spending, and openness, does imply the previous Greek governments, whose budget deficit is currently 104.6% of GDP, should have left firms in the economy to invest, rather than crowd them out with excessive government spending.

---