Causality between Foreign Direct Investment and Tourism: Empirical Evidence from India

Saroja Selvanathan, E.A. Selvanathan* Griffith Business School, Griffith University, Queensland, Australia and Brinda Viswanathan Madras School of Economics, Chennai, India

April 2009

Abstract
This paper investigates the causal link between foreign direct investment and tourism in India by employing the Granger causality test under a VAR framework. A one-way causality link is found from foreign direct investment to tourism in India. This explains the rapid growth in the tourism sector as well as FDI in India during the last decade.

_______________________________________________________________________
* The authors would like to thank Dr Sumei Tang for her comments and Ms Tanya Tietze for her research assistance.

Causality between Foreign Direct Investment and Tourism: Empirical Evidence from India

1.

Introduction

Many countries make changes to their economic policies in order to attract foreign investors and India is no exception. Indias liberalization and deregulation policies during the early 1990s have attracted a huge amount of foreign direct investment (FDI) into India in recent years. India has been ranked as the second most favoured FDI destination in the world, just behind China. Policy makers in many countries believe that FDI will lead their countrys overall development, including the tourism sector. For a developing nation like India, FDI could play a significant role in its economic development in general and to the tourism sector in particular by improving Indias infrastructure such as international airports, highways, hotels and modern technologies which are the keystones to tourism development. The significance of the tourism sector to the Indian economy can be seen from the statement issued by the Union Minister of Tourism on the occasion of Economic Editors Conference 2008. Tourism is the largest service industry in the country. Its importance lies in being an instrument for economic development and employment generation, particularly in the remote and backward areas (XI Five Year Plan, 2007/08 2011/12). The National Tourism Policy was introduced in the year 2002, with the specific aim of promoting the tourism industry as it was believed that increased tourism would lead to growth and overall development through employment generation and poverty reduction. New emerging areas like rural tourism, heritage tourism, eco-tourism, health tourism, adventure tourism and wildlife tourism have been given priority. Schemes and programmes were introduced during the X Five Year Plan (2002-2007) to improve finances of the state governments through private partnerships and attracting more foreign direct investment. As mentioned in GOI (2005) the IX Plan expenditure was Rs.589 crores and with a 45% increase in the X Plan outlay the expenditure was about Rs.2635 crores (all estimates are in 2001-02 prices). The amount was spent largely on infrastructure development while development of specific locations and training of personnel in the hospitality sector were also given importance. The XI Plan further emphasizes the need for developing the

industry through rationalization of taxes, reducing the cost of air travel and local transport, procuring land for building hotels, particularly budget hotels, and development of site specific tourism like cultural and heritage sites or eco-tourism.1 Though these plans are drawn by the central government in New Delhi, the tourism sector is the prerogative of the states. Therefore, with the money allocated to the states, the local governments have to provide land and maintain the sites once they are developed. The bright prospects of this industry led to a target of attracting 10 million international tourists by 2011 being set. The importance of the tourism sector in India can also be seen from its contribution (direct and indirect) to the economy, 6.2% to GDP and 8.8% to employment during 2007. Indias tourism earnings increased from US$2.2 billion in 2002 to US$6.6 billion in 2006. This has led to an increase in Indias share of total world receipts from 0.6% to 0.9% during this period. Though a large proportion of the tourists are domestic there has been an increase in foreign tourists as well. About 2.4 million tourists arrived in India in 2002 accounting for 0.34% of the worlds share of tourist arrivals. This number almost doubled to 4.5 million in 2006 accounting for 0.52% of the worlds share. The number of foreign tourist arrivals has increased at a rate of 12.4% between 2006 and 2007 during the 10-month period of January to October. The number of foreign tourist arrivals during the 10-month period of January to October during 2006 was 3.5 million and increased to 3.9 million during the same months in 2007. Currently, FDI into the hotel industry is close to US$12 billion and about 40 international hotel chains are operational in India. Indias ranking in relation to international arrivals and tourism receipts rose from 51st and 37th, respectively, in 2003 to 42nd and 20th, respectively, in 2007. Domestic tourism visits in India also increased from 309 million in 2003 to 527 million in 2007 (Press Information Bureau, Ministry of Information Broadcasting, 2009). Tourism is one of Indias largest net foreign exchange earners and creator of employment at the village level. Due to the increase in foreign tourist arrivals, the foreign exchange earning has also increased from US$5.0 billion in 2006 to US$6.3 billion in 2007, during the same 10-month period, resulting in a growth of 26%2. The total amount of FDI to India

These are summarised from the XI plan document and the Working Group on Tourism for the XI plan accessed from http://planningcommission.nic.in/plans/planrel/fiveyr/11th/11_v3/11v3_ch8.pdf and http://planningcommission.nic.in/aboutus/committee/strgrp11/str11_tourism.pdf respectively.
2

Ministry of Tourism (undated), Annual Report, 2007-08 accessed from http://tourism.gov.in/AnnualReport 07-08.pdf in January 2009. 2

in 2001 was US$42 billion which increased to US$113 billion in 2004. The amount of FDI inflows into India differs significantly between industries and between states. However, overall, the tourism sector is still one of the most important sectors attracting a significant amount of FDI. According to World Tourism and Travel Corporation (WTTC), Indias tourism industry is expected to grow at a rate of 9 percent per annum during the next decade. India is rated among the top five travel destination in the world by the Lonely planet magazine and as the most preferred destination on earth by the Association of British Travel Agents (ABTA) magazines. It is expected that further liberalization policies to be introduced by the Indian government in 2009 will further increase the FDI to the tourism sector from the current US$450 million to US$1.5 billion by 2010 and increase the number of foreign tourist arrivals to 10 million in 2011. The recent inflow of FDI to India has helped to create 1980 new hotels with 109,392 rooms. The development of the tourism sector needs investment in many forms and FDI is one such source. This introduces a causal link from FDI (to this sector and hence overall) to tourist arrivals as this attracts greater numbers of visitors due to better amenities. A further indirect link from FDI to tourism is through business tourists. These are entrepreneurs and managers from other countries who, while looking for opportunities to invest in India as well as to promote and sustain business in India visit several tourist destinations. This in turn is likely to boost FDI into this sector as well as other related sectors to improve the quantum and quality of service provided wherever lacking. Consequently there is a reverse causality that links tourism to FDI. Tourism is also one of the few sectors where 100% FDI has been permitted by the government of India recently. A number of empirical studies at individual country level have been published in the last two decades which analyse the link between FDI and the tourism sector (for example, see, Sanford and Dong, 2000; Tisdell and Wen, 1991; Contractor and Kundo, 1995; and Kundo and Contractor, 1999). However, these studies used only a basic regression framework. Our study differs from the existing studies on FDI and tourism in at least two ways: (1) (2) Uses more recent data on FDI and tourism for India; and Applies more recent developments in time series analysis to investigate the link between FDI and tourism in India.

The organisation of the paper is as follows. In Section 2 we present a preliminary timeseries data analysis of FDI and tourism data in relation to India. In Section 3, we investigate the direction of causality under a VAR framework. In the last section, we present our conclusion.

2.

A Preliminary Time Series Data Analysis of Tourism and FDI

We use quarterly time series data for the period 1995(2) to 2007(2) for the two variables in natural log-form, namely the number of foreign tourist arrivals (TOUR) in India and the amount of foreign direct investment (FDI) into India (in rupees core). These data are collected from various issues of the Reserve Bank of India Bulletin (published by the Reserve Bank of India) and Indiastats. Figures 1 and 2 plot the two original series in natural logarithms. As can be seen, both series were relatively stable until early 2003 and increase rapidly with a clear upward trend. Obviously, there is also a clear seasonal pattern in both the original series, especially easily visible in the number of foreign tourist arrivals (TOUR) series. There are several ways a time series can be deseasonalized. If we assume the seasonal pattern to be purely deterministic in a time series {yt}, then we could estimate the model

yt = 0 + 1D1 + 2D2 + 3D3 + yt

where D1, D2 and D3 are quarterly seasonal dummies such that Di = 1 for season i and 0 for

other seasons. Then the residuals yt can be viewed as the deseasonlized values of yt
(Enders, 1995, p229). We follow this approach to obtain the deseasonalized series {yt}. Another way of removing the seasonal components is by testing for seasonal unit roots and applying the relevant seasonal filters to the original series (see Engle et al., 1987). Figures 3 and 4 present the plots for the deseasonalized TOUR and FDI series. As can be seen, there is an upward trend in both series. Therefore, the means of the time series are changing over time indicating that both series in their original form may not be stationary. We plot the first-differenced series of the deseasonalized TOUR and FDI series in Figures 5 and 6. These two plots suggest no evidence of changing means indicating that the TOUR and FDI series may be integrated of order one, that is, both time series are I(1).

Number of Foreign Tourist Arrivals

15 LN(NTOUR) 14 13 12 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Year(Q)

Figure 1. Original series of number of foreign tourist arrivals (TOUR), India, 1955:2-2007:2 (in natural logs)

Foreign Direct Investment (Rs Core)

9 8 Ln(FDI) 7 6 5 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Year(Q)

Figure 2. The original series of FDI, India, 1995:2-2007:2 (in natural logs)

Deseasonalized series of Foreign Tourist Arrivals

60 40 20 0 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 -201995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 -40 Year (Q)

Figure 3. Deseasonalized TOUR series, India, 1995:2-2007:2

Deseasonalized FDI Series

200 150 100 50 0 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 Q2 -50 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 -100 Year (Q)

Figure 4. Deseasonalized FDI series, India, 1995:2-2007:2

Deseasonalized Foreign Tourist Arrivals series in First Difference form

20

10

0 Q3 Q3 Q3 Q3 Q3 Q3 Q3 Q3 Q3 Q3 Q3 Q3 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 -10 Year (Q)

Figure 5. The first differenced series of TOUR, India, 1995:2-2007:2

Deseasonalized FDI series in First Differnce form

120 80 40 0 Q3 Q3 Q3 Q3 Q3 Q3 Q3 Q3 Q3 Q3 Q3 Q3 -401995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 -80 Year (Q)

Figure 6. The first differenced series of FDI, India, 1995:2-2007:2

To statistically validate that the two series are I(1), we formally test the stationarity of these two series using the Augmented Dickey-Fuller (ADF) unit-root test in the absence of any structural breaks. We shall also address the issue testing unit roots in the presence of a structural break later, in this section.
6

To perform the ADF test on the deseasonalized series of TOUR and FDI, we estimate the following three regression models (1)-(3) of yt for the presence of unit roots in a time series {yt}: No constant and no trend model yt = yt-1 +

i =1

i yt-i + t

(1)

Constant and no trend model

yt

0 + yt-1 + i yt-i + t
i =1

(2)

Constant and trend model

yt

0 + 2 t + yt-1 + i yt-i + t
i =1

(3)

where yt = yt - yt-1 is the first difference of the series yt, yt-1 = (yt-1 - yt-2) is the first difference of yt-1 etc., and t is a stochastic disturbance term. We apply the ADF test to the TOUR and FDI series separately. The difference among the three regressions is the presence of the deterministic elements 0 and 2t. Equation (2) adds a constant term or drift term 0 to equation (1) and equation (3) includes both a drift and a time trend 0 + 2t. The number of lagged terms is chosen to ensure that the errors are uncorrelated. The sample size used in the estimation is 49. We carry out the estimation of the models using the econometric software SHAZAM and test the presence of unit roots using the systematic procedure described in Enders (1995). The results of the Augmented DickeyFuller (ADF) test for stationarity of the deseasonalized series are presented in Table 1. As can be seen, both series in level form are non-stationary. Now we extend the analysis to the situation of structural breaks. From Figures 3 and 4 it appears that some structural changes may have occurred to the two series around 2002 when the National Tourism Policy was introduced in India. When there are structural breaks, the Dickey-Fuller and Phillips-Perron test statistics are biased towards the nonrejection of a unit root (Enders, 1995). Thus, it is necessary to use the procedure developed
7

Table 1. ADF test results for a unit root on the level form of the original series
Tourism Model
Null hypothesis Critical value at the 10% significance level -3.13 5.34 -2.57 3.78 -1.62 Data-based value of the test statistic -0.51 3.03 2.09 10.52 0.87

FDI Results
Data-based value of the test statistic -0.91 1.14 -0.08 0.84 -0.21

Results

Constant and trend Constant and trend Constant and no trend Constant and no trend No constant and no trend

H0: = 0 H0: 0 = =0 H0: = 0 H0: 0 = =0 H0: = 0

Do not reject H0 Do not reject H0 Do not reject H0 Reject H0 Do not reject H0 {TOUR} has a unit root and the series is nonstationary

Do not reject H0 Do not reject H0 Do not reject H0 Do not reject H0 Do not reject H0 {FDI} has a unit root and the series is nonstationary

Conclusion

by Perron (1989) to test for a unit root in the presence of a structural change. To perform this Perron test we consider the following regression equation (4) for each time series yt and test the null hypothesis of a unit root by testing Ho: 1 =1. The critical values for such hypothesis testing are available in Perron (1989). The model to be estimated for this test is given by yt where = 0 + 1DUt + 2DTt + 1yt-1 + 2t + i yt-i + t
i =1 k

(4)

DUt

1 0

if

t + 1
DTt =

otherwise

t { 0-

if t + 1 otherwise

where = 29 is the structural break which took place in 2003(1). DUt is a level dummy variable; DTt is a slope dummy variable; 0 is a vector of intercept term; t is a deterministic trend; i, i and i are the parameters; k is the lag length; and t is the disturbance term. The value of the test statistics of the Perron (1989) test at various lag lengths of each time series are reported in Table 2. The Perron test results presented in the table indicate that the null hypothesis of a unit root is not rejected by both the series in the presence of a structural break at all lag lengths. This confirms the previous ADF test results and observations made from Figures 3 and 4 that the series in level form may be nonstationary.
8

Table 2. Perron test for a unit root in the presence of a structural change Value of the test-statistic T FDI TOUR 49 49 0.6 0.6 K=1 K=2 K=3 K=4 K=5 K=6 Critical value at 1% -4.24 -4.24 5% 4.88 4.88

-2.72 -2.29 -2.91 -3.00 -2.93 -3.25 -3.98 -2.85 -3.03 -2.48 -2.98 -2.57

Notes: T = number of observations, = proposition of observations occurring before the structural change and K = lag length.

The results so far confirm that both time series have at least one unit root and hence are non-stationary in its original form. We now test the first difference of both series for stationarity by applying the ADF test on the first difference series. The results are reported in Table 3. As can be seen, the results show that both series are stationary in their first difference form. This means both series are I(1).
Table 3. ADF test results for a unit root on the first difference of the original series
Tourism Model
Null hypothesis Critical value at the 10% significance level -3.13 Data-based value of the test statistic -3.22

FDI Results
Data-based value of the test statistic -3.85

Results

Constant and trend

H0: there is a unit root

Reject H0

Reject H0

Conclusion

D{TOUR} has no unit roots and the series is stationary

D{FDI} has no unit roots and the series is stationary

Even if the two variables TOUR and FDI individually are I(1), it may be possible that a linear combination of the two variables may be stationary. If we are modelling a linear relationship between TOUR and FDI, even if each of them are individually non-stationary (i.e. I(1)); as long as they are cointegrated, the regression involving the two series may not be spurious. Thus, we now investigate whether the two series are cointegrated and having a long run equilibrium relationship. We employ the Engle and Granger (1987) procedure, which is based on testing for a unit root in the residual series of the estimated equilibrium relationship by employing the Dickey-Fuller test. Therefore, the null and alternative hypotheses are:

H0: HA:

The residual series has a unit root (or TOUR and FDI are not cointegrated) The residual series has no unit root (or TOUR and FDI are cointegrated)

Rejection of the null hypothesis in both cases would mean that the two series TOUR and FDI are cointegrated. The critical values for the unit root test on the residuals of the cointegrating regression are not the same ones used in the ADF test as the test statistics are not invariant to the number of variables included in the regression. The appropriate critical values are given in Davidson and MacKinnon. (1993). The residual unit root test results are presented in Table 4. The results on the table clearly show that both the least squares residual series are non-stationary and hence the series TOUR and FDI are not cointegrated indicating that there is no long-run equilibrium relationship between FDI and the number of foreign tourist arrivals in India.
Table 4. Test for co-integration of Tour and FDI on the residuals
Critical value at the 10% significance level -3.04 Data-based value of the test statistic

Model

Null hypothesis H0

Results
Do not reject H0

Tour = + FDI + e FDI = + Tour +e

H0: The residuals series has a unit root H0: residuals series has a unit root H0: The residuals series has a unit root H0: residuals series has a unit root

-2.34

-3.04

-2.08

Do not reject H0

Tour = + FDI + t + e FDI = + Tour + t +e

-3.50

-3.06

Do not reject H0

-3.50

-2.52

Do not reject H0

Conclusion

{Residual series} has a unit root and the two variables TOUR and FDI are not cointegrated.

3.

Testing Granger Causality

From the analysis so far, we found that both series TOUR and FDI are I(1) and are not cointegrated. Therefore they have no long term relationship. They may nevertheless be related in the short-run. Their short-run fluctuation can be described by their firstdifferences, which are stationary. The interactions in the short-run fluctuations may therefore be described by a VAR system in first differences.
10

We determine the optimal lag length for the VAR system by using the Schwarz (1978) Criterion (SC) and the Akaike (1974) Information Criterion (AIC). We use a VAR system of k lags and estimate it for various lag lengths. We found that the optimal lag lengths for both the FDI and TOUR series to be 3 lags. Therefore the final system to be used is a VAR(3). We estimate the VAR(3) system in the following form with all variables in firstdifference form and test various hypotheses.
TOURt = 01 + 11TOURt-1 + 21TOURt-2 + 31TOURt-3 + 11FDIt-1 + 21FDIt-2 + 31FDIt-3 + e1t FDIt = 02 + 12TOURt-1 + 22TOURt-2 + 32TOURt-3 + 12FDIt-1 + 22FDIt-2 + 32FDIt-3 + e2t

(5a) (5b)

We also test whether FDIt-1, FDIt-2, FDIt-3 do not appear in the TOURt equation to test H0: FDI>TOUR and TOURt-1, TOURt-2, , TOURt-3 do not appear in the FDIt equation to test H0: TOUR>FDI. Hence the null hypothesis to test non-causality that FDI does not cause TOUR is that H0: 11 = 21 = 31 = 0. Rejection of the null hypothesis H0 means that FDI causes TOUR in the Granger sense. Similarly the null hypothesis to test non-causality that TOUR does not cause FDI is that H0: 12 = 22 = 32 = 0. We perform the above estimation in SHAZAM and Table 5 presents the results. As can be seen from row 1 of Table 5, for testing the null hypothesis, H0: FDI>TOUR, the p-value is 0.08, which is less than the level of significance, 0.10. Hence we reject the null hypothesis that FDI does not cause TOUR in favour of the alternative HA: FDI causes TOUR in the Granger sense at the 10% level of significance. Looking at row 2 of the table, for the testing of H0: TOUR>FDI, the p-value for this test is 0.24, which is larger than the level of significance 0.10. Therefore, we do not reject the null hypothesis H0: TOUR does not cause FDI.

11

Table 5. Results of Granger Causality Test between Tourism and FDI in India, 1995-2007
Null hypothesis (1) (2) H0: FDI > TOUR (11 = 21 = 31= 0 ) H0: TOUR > FDI (12 = 22 = 32= 0) p-value of the F-test statistic 0.08 0.24 Conclusion at the 10% significance level Reject H0 That is, FDI=>TOUR Do not reject H0: That is, TOUR>FDI

4. Conclusion In this paper we have investigated the causal relationship between foreign direct investment (FDI) and the number of foreign tourist arrivals (TOUR) in India using the quarterly data for the period 1995:2 to 2007:2. For this investigation we employed various time series econometric techniques such as unit root test, cointegration and causality. The analysis reveals that the two time series TOUR and FDI are both I(1) and are not cointegrated. We then use the VAR system in first-difference of the two variables to investigate the causality between TOUR and FDI. The results show that there is only a one-way causal relationship from FDI to tourism. That is FDI has a causal effect on the number of foreign tourist arrivals in India. As we pointed out in the introduction, FDI plays a significant role in expanding the tourism sector in India. This shows that appropriate policy to explore tourism resources and plans to develop new tourist venues and facilities may need to be considered in order to meet the increasing demand of tourism in India expected as a result of continued strong foreign direct investment.

12

References Akaike, H. (1974) A new look at statistical model identifications, IEEE Transitions on Automatic Control 19: 716-723. Contractor, F.J. and S.K. Kundu (1995) Explaining variation in the degree of internationalisation across firms: the case of the hotel industry, Journal of International Management 1(1): 87-123. Enders, W. (1995) Applied Econometric Time Series, John Wiley & Sons, Inc, New York. Engle, R.F., and C.W.J. Granger (1987) Dynamic model specification with equilibrium constraints: co-integration and error correction, Econometrica 55(3): 251-276. Davidson, R. and J.G. MacKinnon. (1993) Estimation and Inference in Economics. Oxford University Press. GOI, (2005) Mid term Appraisal of the Tenth five year Plan (2002-2007) Planning Commission, New Delhi. http://planningcommission.nic.in/plans/mta/midterm/english-pdf/chapter-13.pdf in January 2009 accessed

Kundu, S.K. and F.J. Contractor (1999) Country location choices of service multinationals an empirical study of the international hotel sector, Journal of International Management 5(4): 299-317. Perron, P. (1989) The great crash, the oil price shock, and the unit root hypothesis, Econometrica 57: 1361-1401. Sanford, D.M. and H. Dong (2000) Investment in familiar territory: tourism and new foreign direct investment, Tourism Economics 6(3): 205-19. Schwarz, G. (1978). Estimating the dimension of a model, Annals of Statistics 8: 461464. Tisdell, C. and J. Wen, (1991) Investment in Chinas tourism industry: its scale, nature, and policy issues, China Economic Review 2(2) pp175-93. Toda, H.Y. and P.C.B. Phillips (1993) Vector autoregressions and causality, Econometrica, 61(11):1367-1393.

13