Johansen Fisher Panel Co Integration Test Results

Published Date: 02 Nov 2017

To begin with, the stationarity of the variables is tested using 4 panel unit root tests. Lag length selection is based on Akaike information criterion (AIC).Table 3 below summarise the results of the stationarity test of the ln values of the variables, in level form and in first difference respectively.

Table 2: Summary Results of Unit Root Tests:

Levin, Lin and chu t*

Im, Persaran and Shin W-stat

ADF-Fisher Chi-square

PP- Fisher Chi-square

Variable

Level

1st Dif.

Level

1st Dif.

Level

1st Dif.

Level

1st Dif.

Ln_fdi

0.0004

0.00

0.1607

0.00

0.0070

0.00

0.0217

0.00

Ln_gdppc

0.9147

0.00

0.9604

0.00

0.8280

0.00

0.9736

0.00

Ln_hc

0.0473

0.00

0.9236

0.00

0.0152

0.00

0.0722

0.00

Ln_opns

0.0071

0.00

0.0401

0.00

0.0026

0.00

0.0103

0.00

Ln_tel

0.0000

0.00

0.5188

0.00

0.0010

0.00

0.0851

0.00

Ln_xtra

0.0097

0.00

0.3631

0.00

0.4433

0.00

0.0000

0.00

Note: The figures in the tables are probability values and the probabilities for fisher tests are computed using an asymptotic chi‐Square distribution.

Null: Unit root

The null hypothesis of a unit root is accepted or rejected based on the p-values of the LLC, IPS, ADF-Fisher and PP-Fisher tests. It is to be noted that, as long as one of the tests rejects the null hypothesis of a unit root at the level form, but accepts same when the variable is first differenced. Therefore, the stationarity test shows that all our variables are integrated of order 1 (I (1)) and are thus stationary at first difference.

5.2 Johansen Fisher Panel Co-Integration Test Results

The first differenced form is use for the VAR model to check for co integration and further, causality. Using the Johansen test for Co integration confirm the existence of co integration as all the variables are integrated of order 1. A summary of the results of the Johansen Fisher panel Co-integration test is reported in Table 4 below.

Table 3: Test result from Johansen Fisher panel procedure

Johansen Maximum Likelihood procedure of the co-integrating regression fdi= (gdppc,hc,opn,tel,extra): number of co-integrating vector(s) using the co-integration likelihood ratio.

Hypothesized

No. of CE(s)

Fisher Stat. *

(from trace test)

Prob.

Fisher Stat. *

(from max-eigen test)

Prob.

None

518.4

0.0000

278.0

0.0000

At most 1

288.6

0.0000

144.2

0.0000

At most 2

171.0

0.0000

85.61

0.0000

At most 3

105.9

0.0000

73.35

0.0000

At most 4

59.58

0.0005

51.89

0.0039

At most 5

47.20

0.0130

47.20

0.0130

Trend assumption: Linear deterministic trend

*Probabilities are computed using asymptotic chi‐square distribution.

The corresponding probabilities of the Trace Test and the Max-Eigen Test show that there are at most 4 and at most 5 co-integration equations (CEs). Hence, it may be argued that there exist stable long run relationships among the variables and a PVECM is employ to better capture and predict results about causality. The ECT (from Johansen procedure) is used together with current and past differenced values of FDI determinants to get speed of adjustment.

5.3 PANEL VECTOR ERROR CORRECTION MODEL [PVECM]

PVECM is applied to characterize both long run equilibrium relationships and short run dynamic adjustment processes between FDI and of FDI determinants, the independent variables. Before proceeding to estimate the short run relationships, the long run properties are first examined.

LONG RUN ESTIMATES

An estimate of this co-integrating vector, normalised on fdi, is shown in Table 5(a) below. The coefficients attached to the different explanatory variables not all are significant with the required theoretical sign. Like the market size and infrastructure are insignificant and the estimates of human capital, openness and exchange rate are significant. It is observed that not all ingredients are important to attract FDI in Sub-Saharan African countries in the long run.

Table 4: Estimated Cointegrating Vector and Error-Correction Equations

Estimated Co-integrating Vector

Variable[as in equation]



t-stat.

Fdi

1

GDPpc

-0.275635

-0.59917

HC

1.555958

2.94914 ***

OPNS

1.076819

1.87369 *

TEL

-0.314859

-0.82767

XTRA

0.211577

1.89762 *

Adapted From Eviews 7

From here onwards, *, **, and *** indicate significance at 10%, 5% and 1% respectively

The co-integration equation shown in Table 5(a) describes the long run equilibrium relationship between FDI on one hand and its determinants on the other for the sample of 14 Sub Saharan African countries. When this expression, which is also vector error- correction term (ECT), is statistically significant, it implies that FDI is weakly endogenous with respect to the long run parameters. In general, all the estimated parameter coefficients carry the expected signs. With exception of market size and infrastructure, the estimated coefficients of all specified determinants of FDI are statistically at the 10% (OPNS and XTRA) level and 1% (HC) level. The results of the long-run FDI equation can be summarized as follows:

Firstly, the estimated coefficient of GDP per capita carry the unexpected negative sign and statistically insignificant. The result indicated GDP per capita does not have a positively significant association with FDI inflows. However, the study findings by Alsan et al., 2006 and Anyanwu 2011 have also showed that gross domestic product per capita is negatively related to FDI inflow. The lack of positive significance of GDP per capita could be due to a balancing of the market size effect with the cost of production effect, which should work in the opposite directions. Moreover, Sub-Saharan Africa has historically received the smallest amount of FDI globally. The region accounted for only 5.1% of total world FDI inflows in 2009, compared to 26.0% for Asia and 11.9% for Central and South America (UNCTAD 2012). When compared to GDP levels, this figure is less surprising; since the region accounts for just 2.44% of World GDP, it is reasonable to expect that it should have a smaller share of global FDI than other regions. Since SSA received the smallest amount of FDI globally. It can be said that, most of the SSA countries in the sample taken for the analysis have small economy. Leading to an insignificant market size.

Secondly, the estimated coefficient of human capital is positively signed as expected and statistically significant at 1% level. In other words a one percentage point improvement in the school enrollment in the population would induce FDI flows to rise by approximately 1.56% annually. Basically, a good level of human capital enhances the investment climate and thus attracts FDI. This occurs directly through the improved skill level of the labour force, and indirectly through socio-political stability and health conditions (World Bank, 2003; UNESCO and OECD, 2003)

Thirdly, openness has been regarded as one of the most influential enzymes in the process of FDI accumulation. Most of the SSA nations in my sample receive vertical FDI which requires the recipient economy to have liberal trade regimes and open markets (Lim, 2001). The estimated coefficient of openness is statistically significant at the ten percent level, which is consistent with theory. In other words, a one percentage point increase in the economy’ openness would induce approximately 1.08 percentage point of FDI. Kravis and Lipsey(1982), Culem (1988), Chakrabarti(2001), Asiedu(2002), Ang(2008), among others reported similar results previously. Moreover, the statistical significance of trade opennesson FDI corroborates with the results of Naude and Krugell (2007); Demirhan and Masa (2008); and Seetanah and Rojid (2009) who acknowledge openness as being one of the major determinants of FDI flows.

Fourthly, the estimated coefficient of infrastructure is negatively signed and statistically insignificant. This result is in line Marr (1997). Most findings show that the infrastructure development leads to more FDI inflows. However, the finding is supported by previous study (Aseidu 2002). The study by Aseidu (2002) showed that the infrastructure development doesn’t have effect on FDI inflow to Sub-Saharan Africa compared to other non-Sub-Saharan Africa.

Finally, the estimated coefficient of real exchange rate is positively signed and statistically significant at the 10% level, suggesting that real appreciation of exchange rate causes FDI flows to surge into the sample of 14 SSA countries. The finding is consistent with those by Edwards (1990), Hasan (2008), among others, who found positive correlation between exchange rate and FDI flows. Hasan (2008) argued that a weak currency is likely to increase the FDI flows to a country over time. In contrast, in another recent study on Malaysia, Ang (2008) found negative correlation between exchange rate and FDI. These conflicting findings obviously reflect how elusive the effect of exchange rate on FDI flows.

SHORT RUN ESTIMATES

Table 4(b): Short Run Estimates (PVECM)

Dependent Variable

Δfdiit

ΔGDPpcit

ΔHCit

ΔOPNSit

ΔTELit

ΔXTRAit

Independent Variables

ECit 1

-0.048161***

0.022063***

-0.005940

-0.009012*

0.001126

-0.02748***

fdiit 1

0.10633*

0.086127***

-0.023579

0.094374***

0.030065

-0.078587***

GDPpcit 1

0.034805

0.177225***

0.0015033

-0.078336

0.063054

0.177900***

HCit 1

0.215743***

-0.005935

-0.056104

0.008893

0.010410

0.045162

OPNSit 1

-0.119138

0.116884*

0.054935

-0.113715*

0.100913*

-0.118903*

TELit 1

0.126816

0.008339

0.060402

-0.038184

0.349372***

-0.033650

XTRAit 1

0.056233

-0.028509

-0.107272

-0.037560

-0.039665

0.441479***

Constant

0.042132**

0.020322*

0.038296***

0.010883

0.034189***

0.038177***

Adapted From Eviews 7

Lag length criteria indicates the use of 1 lag.

The previous, long run, results are generally validated even in the short run with some changes in table 6(b) in the following the error correction model. Thus, only human capital is significant in explaining the short-run variation in FDI while such is not the case with market size, openness, infrastructure and exchange rate. Indeed a 1 percentage-point increase in the growth rate of the human capital leads to a 0.215743 percentage-point increase in the growth rate of FDI ratio after one year. Such interpretation can be extended to the other explanatory variables. These are estimates of the direct effect of each explanatory variable on FDI in the short-run. The market size (GDPpcit 1) is insignificant and positive. This result is in line with Astatike and Assefa (2005). The infrastructure and exchange rate is also positive and insignificant.

Interestingly, the empirical result for the short estimates reveals that the variable openness is not significant. This implies that foreign investors did not place much importance to the economic openness of the host country while deciding about the location of their projects in SSA. This is contradictory to some of the theories as well as to some empirical studies (Garibaldi et al 2001; Compos, et al. 2003; Cheng and Kwan 2000; Asiedu 2002; and Onyeiwu and Shrestha 2004) which show that openness of country does influence the FDI inflows. The reason why in our study coefficient of openness turned out to be insignificant may be explained in terms of the nature of FDI inflows into SSA. The developing countries attract mostly market seeking investments and when investments are market-seeking, trade restrictions (and therefore less openness) can have a positive impact on FDI (Jordaan 2004). The reason stems from the tariff jumping hypothesis, which argues that MNCs that seek to serve local markets may decide to set up subsidiaries in the host country if it is difficult to import their products to the country. In contrast, MNCs engaged in export-oriented investments may prefer to invest in a more open economy since increased imperfections that accompany trade protection generally imply higher transaction costs associated with exporting. Therefore, it may be concluded that FDI inflows to developing countries are primarily of market-seeking type or tariff- jumping type and hence least affected by trade restrictions.

Besides, short term negative relationship is found to run between trade openness and fdi. Moreover, the coefficients of the past values of fdiit 1 on ΔGDPpit and ΔOPNSt are statistically significant and positive at 1% respectively. The EC for openness and exchange rate are negative and significant at 10% and 1% respectively, meaning that these variables achieve convergence in the LR. Thus, a deviation from the LR equilibrium relation in one period is corrected in the next period by 0.90% for openness and 2.75% for exchange rate. It is important to note, for instance, that from the 4th row, hc depends on fdi, implying causality as well, thus depicting an endogeneous relationship. These are quite interesting results. As such, we report in table 7 the granger causality between FDI and the independents variables. There exist a uni-directional causality and a bi-directional relationship between the variables

In essence, a good level of human capital enhances the investment climate and thus attracts FDI. This occurs directly through the improved skill level of the labour force, and indirectly through socio-political stability and health conditions (World Bank, 2003; UNESCO and OECD, 2003). In turn, FDI helps in enhancing the level of human capital as MNCs bring with them advanced skills, information and technology and also provide training to their workers.

Furthermore, a VAR modeling also allows us to investigate the possibility of other indirect effects. For instance, a 1 percentage-point increase in the growth rate of market size leads to a 0.116884 percentage-point increase in the growth rate of openness after one year. Likewise, a 1 percentage-point increase in the growth rate of openness leads to a 0.094374 percentage-point increase in the growth rate of FDI after one year. The two pieces of information above, taken together, imply that a 1 percentage-point increase in the growth rate of market size leads to a 0.116884 X 0.094374 percentage-point increase in the growth rate of FDI after two years. This is an estimate of the indirect effect of market size on FDI in the short-run via the openness channel. As such, we can observe a number of possible indirect effects of selected explanatory variables on FDI .

Like, a 1 percentage-point increase in the growth rate of exchange rate leads to a 0.01189 percentage-point decrease in the growth rate of openness after one year. Likewise, a 1 percentage-point increase in the growth rate of openness leads to a 0.094374 percentage-point increase in the growth rate of FDI after one year. Therefore, a 1 percentage- point increase in the growth rate of exchange rate leads to a -0.01189 X 0.094374 percentage-point decrease in the growth rate of FDI after two years.

5.4 PAIRWISE GRANGER CAUSALITY TEST

Table 5

Sample: 1986- 2010

Lags: 1

 Null Hypothesis:

Obs

F-Statistic

Prob. 

null hypothesis

 GDPPC does not Granger Cause FDI

 336 #

 2.44638

0.1187

Accept

 FDI does not Granger Cause GDPPC

 31.0400

5.E-08**

Reject

 HC does not Granger Cause FDI

 336

 2.03696

0.1545

Accept

 FDI does not Granger Cause HC

 10.8623

0.0011**

Reject

 OPNS does not Granger Cause FDI

 336

 1.18316

0.2775

Accept

 FDI does not Granger Cause OPNS

 3.09732

0.0793*

Reject

 TEL does not Granger Cause FDI

 336

 0.08853

0.7662

Accept

 FDI does not Granger Cause TEL

 5.84274

0.0162**

Reject

 XTRA does not Granger Cause FDI

 336

 0.32920

0.5665

Accept

 FDI does not Granger Cause XTRA

 129.029

2.E-25**

Reject

 HC does not Granger Cause GDPPC

 336

 4.95978

0.0266**

Reject

 GDPPC does not Granger Cause HC

 8.47626

0.0038**

Reject

 OPNS does not Granger Cause GDPPC

 336

 44.0062

1.E-10**

Reject

 GDPPC does not Granger Cause OPNS

 0.54917

0.4592

Accept

 TEL does not Granger Cause GDPPC

 336

 8.35470

0.0041**

Reject

 GDPPC does not Granger Cause TEL

 1.68426

0.1953

Accept

 XTRA does not Granger Cause GDPPC

 336

 0.20298

0.6526

Accept

 GDPPC does not Granger Cause XTRA

 21.2305

6.E-06**

Reject

 OPNS does not Granger Cause HC

 336

 5.75192

0.0170**

Reject

 HC does not Granger Cause OPNS

 0.02004

0.8875

Accept

 TEL does not Granger Cause HC

 336

 15.3293

0.0001**

Reject

 HC does not Granger Cause TEL

 0.06996

0.7916

Accept

 XTRA does not Granger Cause HC

 336

 0.37950

0.5383

Accept

 HC does not Granger Cause XTRA

 34.3175

1.E-08**

Reject

 TEL does not Granger Cause OPNS

 336

 0.63051

0.4277

Accept

 OPNS does not Granger Cause TEL

 12.3937

0.0005**

Reject

 XTRA does not Granger Cause OPNS

 336

 4.59447

0.0328**

Reject

 OPNS does not Granger Cause XTRA

 55.3010

9.E-13**

Reject

 XTRA does not Granger Cause TEL

 336

 0.53390

0.4655

Accept

 TEL does not Granger Cause XTRA

 36.0588

5.E-09**

Reject

Source: Author’s own estimate

# Observations after lag.

*(**) Indicates significant causal relationship at 10% (5%) significance level.

Recall that although co-integration between two variables does not specify the direction of a causal relation, if any, between the variables. Economic theory guarantees that there is always Granger Causality in at least one direction. Researchers verify the direction of Granger Causality between GDPpc, OPNS, HC, XTRA and TEL. In the study, F-Statistics and probability are used to measure causality between the variables. F-statistics and probability values constructed under the null hypothesis of non causality shows that there is a causal relationship between those variables.

The results of pairwise granger causality between FDI and the five independent variables are illustrated in Table 7 above. Causality between human capital and market size; Openness and exchange rate are bi-directional. The bi-directional causality runs from exchange rate to openness and vice versa. This is in conformity with the expectation and with the realities in the SSA economy, that is, just as exchange rate appreciation could result in an increase in imports leading to an improvement to openness. , a rise in the trade openness could also leads to an appreciation in the level of the exchange rate

Significant probability values denote rejection of the null hypothesis. This study accept the null hypothesis if the probability value is more than 10% otherwise reject the null hypothesis if the probability value is less than 10%. It is found that, there exist unidirectional causality between fdi and market size; fdi and human capital; fdi and openness to trade; fdi and infrastructure; fdi and exchange rate; at 5 and 10 % significant level. It means that fdi follow its mature counterparts in the short-run that there exists a lead-lag relationship between fdi and the independent variables.

The causality test also tested between two independent variables. There is unidirectional causality running between market size and openness, infrastructure and market size, exchange rate and market size, openness and human capital, infrastructure and human capital, exchange rate and human capital, infrastructure and openness, exchange rate and infrastructure.