IMPACT OF COVID-19 ON FDI INFLOWS INTO INDIA

1. Dual Degree, Mechanical Engineering, IIT Kharagpur. 2. PGDM in Information Technology and Systems Management, NMIMS. 3. BBA, GITAM University. ...................................................................................................................... Manuscript Info Abstract ......................... ........................................................................ Manuscript History Received: 10 July 2020 Final Accepted: 14 August 2020 Published: September 2020

In India, FDI inflows is a key indicator of the confidence investors have in its economy. The key purpose of this study is to evaluate the impact of COVID-19 on FDI inflows into India. A regression model was built using pre-COVID to predict the FDI growth in each month. It has been observed that changes in exchange rate and foreign reserve have a statistically significant impact on FDI growth, where-as changes in IIP (Index of Industrial Production) do not. The predicted values were then compared with the actual FDI growth, during the pandemic. Though the regression model was robust and sound, very large differences in predicted and actual FDI growth was found during the pandemic. This indicates that the pandemic has altered the dynamics of FDI growth in India.

…………………………………………………………………………………………………….... Introduction:-
As the COVID-19 rages in India, it has a telling effect on the Indian economy. Since liberalisation in 1991-92, India has seen a huge growth in FDI. FDI would often be seen as a good omen for Indian economy, as it was a sign of confidence the overseas investors had in India. In recent days, the Indian Govt. has put a strict cap on the Current Account Deficit, thereby reducing its own ability to spend. Private investment has always been low in India, except for the big business families. So, in order to fund the dreams of millions of aspiring youth and to sustain the growth, the reliance on FDI has only grown. Given this background, the pandemic is putting to test the confidence overseas investors have in India. It is especially important to monitor changes in FDI and understand underlying patterns of change, because the other two sources of investment in India namely Govt. spending and private investments have gone down. The Govt. can't really put money in the economy because of a fall in GST revenues and spending on public health during the pandemic. There is little private investment as many people have lost their jobs and the light at the end of the tunnel is not yet visible.
The objective of the study is to analyse the impact of COVID-19 on FDI in India. Data from Jan-2019 to Feb-2020 is used to build the model. Then, this model predicts the FDI growth for months from March-2020 to Jun-2020. The deviation in the predictions and actual FDI values is studied. In her article, Gujrati R. 1 explores qualitatively, the impact of COVID-19 on FDI in India. This study is a further improvisation using quantitative methods like regression analysis, and using it to see if there is a fundamental change in FDI's dynamics.

Data Data Collection
In this study, data from Jan-2019 to Jul 2020 was used for the following variables: 1. Growth in FDI in each month ($ million) was collected from Trading Economics 2 2. Forex Reserves in each month ($ million) was collected from RBI's official website 3 3. Exchange Rate on each month was collected from X-rates website 4 4. IIP for each month was collected from monthly press releases of Indian Govt. 5 Data Transformation and Feature Engineering 1. Natural Logarithm of FDI growth was taken, to correct the imbalance in ranges of variables.
2. IIP -IIP in manufacturing only where FDI is permitted, was used. Since the dependent variable is growth in India, first difference of IIP was calculated. Also, for the purpose of model selection, a lag of 1 month was also introduced in IIP 3. Forex Reserves -As the dependent variable is FDI growth, first difference of Forex Reserves ($ million) was calculated 4. Exchange Rate -As the dependent variable is FDI growth, the first difference of Exchange Rate ($ million) was calculated

Model Specification Preliminary Regression and Selection of Variables
Data from Jan-2019 to Feb-2020 is used to build the model, before building a regression model, first preliminary regression is run to get a fundamental understanding of the relationship between variables Model 1: Y t = β 0 + β 1 X 1t + β 2 X 2t + β 3 X 3t + ε t (1) The dependent variable is 'ln (FDI_gth)' and the independent variables are 'FXRate_change', 'FXReserve_change' and 'IIP_change'. Upon running the regression, the following resultswere seen  From tables 1 & 2, the model is statistically insignificant with Probability of F-stat being greater than 10%, also, none of the independent variables are significant at a level of 5% So, the next candidate model is evaluated, where a one-month lag in IIP is introduced, Model 2: Y t = β 0 + β 1 X 1t + β 2 X 2t + β 3 X 3(t-1) + ε t (2) 635 The dependent variable is 'ln (FDI_gth)' and the independent variables are 'FXRate_change', 'FXReserve_change' and 'IIP_change_lag1'. Upon running the regression, the following resultswere seen

Final Model
Based on the results of run preliminary regressions, the variable involving IIP is droppedfrom the model

Model Validation
For the estimators to be Best Linear Unbiased estimators, we check several conditions and statistics

Mean of residuals
Mean of residuals must be zero.  Fig. 1, the mean of residuals is very close to zero, so, there is no specification error in the model .

Autocorrelation Test
From Fig. 1 we see that auto-correlation is absent, however we confirrm it using Durbin-Watson Test From the following table 9, the Durbin-Watson test statistic is comfortably below (4-d U ). So, we fail to reject the null hypothesis that the residuals are not correlated. So, it can be confidently said that the model has no autocorrelation problem Heteroscedasticity Test From Fig. 1 there is no difference in variance of residualsHowever, we confirm it mathematically by employing White's test for heteroscedasticity

Multicollinearity Test
For the model to be valid, the independent variables should not be correlated, the check for Multi-collinearity is done using VIF (Variance Inflation Factor) test.

Normality of residuals
Normality of residuals is required for the F-test on the model to be valid, so we proceed to conduct the Jarque-Berra Test.

Test for stuctural breaks
As the data is time-series data, there is a chance that there are structural breaks in the data. So, to check for any structural breaks, the Chow Test is done.
There are 14 periods (months), so, the check for structal break is done at periods 6,7,8. Period 0 is Jan-2019

Prediction and Results:-
Predicted and Actual FDI: FDI was predicted using the coefficients obtained from the regression model, these were compared to the Actual FDI inflows.

Interpretation
Though the regression model was robust and without defects, we see that the predictions miss their mark by a mile. So, we can say that the pandemic has altered the dynamics of what determines FDI. This is particularly true as the model has both types of errors (over-prediction and under-prediction) on data from Mar'20 to Jun'20

Conclusion:-
From the study,it was seen that the growth in FDI was a general downward trend as the pandemic progressed.Also, the growth in FDI was lower than predicted.It is seen that growth in FDI varies exponentially with changes in ForexReserves and changes in Exchange Rate. As expected, the coefficients are negative, meaning decrease in these two indicates better investment opportunities. Also, movement of IIP was seen to be a very poor indicator of changes in FDI. A robust regressionmodel was built, to see if the predictions of the modelduring the pandemic are accurate. It was noted that the predictions are off, indicating a significant change in FDI's growth pattern.