This specific segment of the work will help in explaining the significance of the tax incentives in attracting investment in Mauritius. The analysis will be based on the implications of tax rates on the Mauritian Investment.
4.1.1 Introduction
The following studies are the first steps of the analysis part in our dissertation to have a broad idea of the significance of tax incentives, other than double taxation treaty, in attracting investment in Mauritius. The results of the analysis seem credible and reasonable. It entails the different tests explained later and illustrates the outcomes of the tax rates on the different capital inflows.
4.1.2 Unit Root test
It is essential, especially when dealing with time-series analysis, to ensure that all the variables used are individually stationary. The most popular test for testing stationarity is the unit root test. It comprises of the Augmented Dickey Fuller (ADF) test. It regresses the first difference of each variable on its own level form lagged value.
Some variables are already stationary in level form (refer to table C in Appendix); others need to be differenced to become stationary as shown in the table below.
Taking a closer look at the p-value, the ADF test shows the first difference variables are stationary as p<0.1. It can also be identified when the t statistic exceeds the critical values (refer to table D in Appendix). The stationarity of all our variables allows the continuity of our regressions analysis.
4.1.3 OLS Estimation
Empirical analysis based on the OLS estimation shows the significance and impact of each variable.
The following table corresponds to the regression results where the dependent variable is FDI as a percentage of GDP.
The higher the t-value, the lower is the p-value, then the more significant the variable. All the variables have the expected sign, except for GDP, CIT and past FDI as a percentage of GDP. The most important variables to be analyzed throughout this regression are the CIT rates and the investment allowance. The regression provides an unexpected positive relationship between FDI and CIT. However, the p-value indicates that the coefficient is highly insignificant, and thus do not make any economic sense. Therefore, based on this regression, CIT seems not to have any impact on FDI, in the Mauritian context. This can be related to the fact that even though there was limited variation in the CIT in the period 1981 to 2007, shown by the standard error of 0.0335916, FDI of Mauritian had major fluctuations ranging from US$ 671400.66 in 1981 to US$ 340763853.70 in 2007, thus justifying the highly insignificant CIT variable in attracting FDI in Mauritius.
The same trend is applied to investment allowance. Though its sign is theoretically correct, the p-value indicates that investment allowance does not impact on the level of FDI in Mauritius. Both factors therefore seem to contribute poorly to the foreign investment in Mauritius.
Nonetheless, as theory did predict, openness has a positive impact on FDI which is highly significant at 1 percent significance level. It implies that 1 percent increase in the openness proxy leads to a 0.083 percentage point climb on FDI relative to GDP. The result confirms the Singh and Jun's (1995) study showing that there is strong evidence that exports precede FDI flows.
Another highly significant variable is worldwide growth. It positively affects the level of FDI and it deduces that 1 percent increase in world growth impact on FDI through a rise of 0.504 percentage points. It is consistent as investment is mainly concerned by the economical ability of foreign investors to invest abroad. Therefore, worldwide growth and FDI are closely related.
Nevertheless, the remaining variables namely past FDI, GDP growth and inflation are insignificant and do not seem to impact on the Mauritian FDI. The regression yields an unexpected and surprising negative relationship between PAST FDI, GDP growth and FDI, so that it contradicts economic theory. Inflation on the other hand has the expected negative sign, indicating that a rise in macroeconomic volatility will lead to a fall in FDI. Even though those variables are expected to have an influence, in the context of the Mauritian economy, according to the OLS results, these variables do not seem to have any effect.
Our second regression is based on Private investment (PINV) as a percentage of GDP
PINV is perceived as one of the most important category of investment. Its OLS estimation illustrates very intriguing yet interesting results. The signs of the regression are satisfactory except for investment allowance, GDP growth and inflation.
First of all, let us have a closer look to CIT. Compared to FDI, the coefficient estimated on CIT bears an expected negative sign and is statistically significant at 5 percent level. It points out that 1 percent rise in CIT will lead to a decrease in PINV of 0.058 percentage points of GDP. Therefore, it is interesting to note that Mauritius tax base influence PINV as a source of foreign investment.
However, the impact of investment allowance has remained unchanged. It has an unexpected negative relationship with respect to PINV as a percentage of GDP, but is still statistically insignificant and has no or very little effect on private investment in Mauritius. Klemm and Parys (2009) have also confirmed the inexistence of investment allowance on investment.
Same as for FDI, the degree of openness has been one of the factors affecting PINV. The variable shows a positive sign and is statistically significant. It demonstrates that 1 percent increase in openness leads to 0.07 percentage points increase in PINV. It reflects the willingness of a country to accept foreign investment, and proved to be important in attracting capital [Nonnemberg and Cardoso de Mendonça (2004)].
PINV, contrarily to FDI seems to be more influenced by the variables used. It shows that PINV does not follow a random walk phenomenon; it is partly influenced by its past trend as that 1 percent raise in past PINV leads to a 0.47 percentage points increase in PINV.
Nonetheless, in this new context, inflation seems to have a surprisingly positive relationship and significant impact on FDI relative to GDP. It may be contrasted with the traditional theory like Schneider and Frey (1998) implying that multinational firms invest less in emerging economies with fluctuating inflation. The regression illustrates that 1 percent boost in inflation leads to 0.15 percentage points increase in PINV.
However, as it was the case for FDI, the regression of PINV also points out that GDP growth has a negative relationship with PINV and is statistically insignificant which definitely contradicts economic reasoning.
Moreover, it was unforeseen that worldwide growth would be economically insignificant in determining the level of PINV in Mauritius, but the sign of its coefficient corresponds to what was expected, i.e. it is expected to have a positive relationship.
4.1.4 Multicollinearity and Autocorrelation
The multicollinearity test depicts the existence of a "perfect" relationship between some or all variables. The presence of multicollinearity should be corrected because it designates that the regression coefficients stay undefined, the standard errors are unlimited and a singular solution is not obtained for the individual regression coefficient. R2 in PINV shows that 99% of the changes on PINV are due to the variables used in the above model. It outlines no multicollinearity compared to R2 in the FDI model indicates that 38% of the variations in FDI are determined by the variables in the model may show presence of multicollinearity.
1 - R2 is known as the Tolerance of the X (or 1/VIF). A tolerance near to 1 means there is small multicollinearity, whereas a value near to 0 implies that multicollinearity can be prevailing. The reciprocal of the tolerance is known as the Variance Inflation Factor (VIF). The VIF demonstrates the extent to which the variance of the coefficient estimate is being inflated by multicollinearity.1
However, both regressions entails that there is no multicollinearity, as shown by the Variance Inflating Factor (VIF), which happens to be less than 10 for all the variables and reciprocal of Variance of Inflating Factor being not close to zero. As shown in the tables below, both mean VIF are less than 10, demonstrating no multicollinearity.
Another important test should be undertaken to ensure the proper use of our variables, and define the correlation between the variables ordered in time. Autocorrelation in time series data is defined as the correlation of members of series of observations ordered in line. If the data include autocorrelation, it implies that the OLS estimators are inefficient. Therefore the standard t, F and R2 cannot be effectively used. In order to detect for autocorrelation the Durbin-Watson (DW) test is carried out.
For the regressions of FDI and PINV, the Durbin Watson test shows 1.92 and 1.61 respectively. These results imply that both regressions approximate 2 showing that there are no serial correlations and that our OLS estimations have been well interpreted. There is no need to correct for autocorrelation using the Cochrane- Orcutt method.
4.1.5 Cointegration and Error Correction Mechanism (ECM)
Test of cointegration shows whether there is a long-term relationship between the variables of this regression model using the Engle Granger test. If a relationship is cointegrated, the estimates of these relationships are very consistent.
Granger (1986) noted that a cointegrating test is a pre-test to avoid spurious regression. The test used for cointegration is the cointegrating regression Durbin Watson (CRDW).
According to Granger and Newbold (1974) if R2 < d, it represents no spurious results, consequently, the results are cointegrated. Therefore, as both regressions follow the R2 < d rule of thumb, the variables seem to share a stable long-run relationship between variables between them. Nevertheless, in the short-run, there may be disequilibrium.
The ECM reconciles the short term behavior of an economic variable to its long run behavior. Engle and Granger (1987) prove that if two or more variables are cointegrated, then the short run equilibrium relationship between the two variables can always be represented by an ECM. In our analysis, the two regressions have been added a new variable named as r (refer to table E in Appendix). It is known as the equilibrium error term.
The Engle and Granger test confirms that from the two regressions, there has been evidence, as shown below, that the variables share a short-run relationship as r in both OLS regression is negative and significant.
4.1.5 Interpretation
Let us now have a more in depth analysis of the significance of the tax rates in attracting investments in Mauritius.
The inefficiency of Investment Allowance
The regressions showed that tax incentives in the form of Investment Allowance do not have any impact on FDI and also on PINV relative to GDP; this might imply that the government of a jurisdiction such as Mauritius should aim at providing other types of fiscal incentives so as to influence foreign investors' decision about where to invest.
For instance, tax holidays might be a very attractive argument in boosting up the level of FDI and new capital formation.
In effect, many studies show that countries (mostly the developing ones) tend to offer long period of tax exemption to companies investing in the country. Investors possibly also have other concerns or are interested in other aspects of their investment rather than considering the relief on capital expenditure, explaining why investment allowance is economically insignificant. For instance, questions may arise to whether this allowance will be sufficient to outweigh any extra costs or losses that are encountered. Therefore though investment allowance might seem to increase the attractiveness of a country in which to invest, it has severe weaknesses and these do not help in raising the level of investments.
Reduced CIT & FDI
As seen in table 2, reduced CIT is not significant in determining FDI in the country. This is quite surprising since according to literature we can easily deduce that an investor's priority is to maximize his after tax returns. Therefore, what explains the fact that in Mauritius, CIT does not impact on FDI relative to GDP? Firstly, we can assume that in the case where the actual return is largely superior to the expected return then the tax payment is clearly not an issue for the investor. Secondly, it might imply that before investing, investors consider other factors apart from tax incentives such as human capital, economic clusters, exchange regimes, political stability and geographical location of the host country. These factors weigh a lot more in their decision than the reduced taxes that they might obtain. Thirdly, it can be that the profit margins of companies investing in Mauritius are so low that tax incentives do not have any effect (Clark 2002). Recently, the OECD pointed out that the sensitivity of FDI to tax is all a matter of the host country and the mobility of business activities underlying tax base. Given that our regression produced accurate results, then tax incentives do not actually promote high levels of FDI relative to GDP in Mauritius. What is therefore the rationale behind the low tax rates that Mauritius offers (knowing that this might create intense competition among countries to attract investments)?
Reduced CIT & PINV
Although it may be subject to controversy to keep CIT low when it is not helping in attracting FDI, we must nevertheless consider its impact on PINV which is also one of the main sources of foreign income and vital in maintaining encouraging growth rates. In effect, according to our regression, CIT helps in improving the level of PINV.
A priori negative relationship between CIT and PINV exists. Therefore if you decrease your corporate tax rates, it will help in increasing the level of PINV relative to GDP. However, it will not be by a considerable amount. It is interesting to note that CIT has a significant impact on PINV and not on FDI. It seems that CIT tends to be beneficial only to new investment. When considering the economic literature about factors susceptible of making PINV more sensitive to tax rates, we see that Mauritius has a good business environment, and has over time consolidated its fiscal policy. According to our empirical findings, tax rates will not affect private investment significantly. This may be because investors tend to consider the structure of the tax regime of the country first, i.e. whether it is simple or complex, transparent and stable.
Limitations
The empirical implications of our model are limited. For instance, we are aware that aggregate investment depends on many other factors such as government expenditure, political stability, democracy and many others. Also, we have tested only two types of tax incentives, but there are other types such as tax credits, accelerated investment allowances and tax holidays. Finally, our model offers few clues to how Mauritius can best lower its tax rates to attract FDI.
The next section is the second part of the analysis which will assess the impact of double taxation treaty in depth, compare Mauritius offshore centre with other jurisdictions and the future prospects of Mauritius.
