Did the 2008 global financial crisis influence the host country corruption and inward foreign direct investments relationship? An empirical examination

Abstract

We use annual data over the 2002–2016 period with Panel Smooth Transition Regression model to examine how the 2008 global financial crisis affected the relationship between FDI inflows into developed, developing, and transition host countries and host country corruption levels. We also partitioned the data for each type of host into corruption regimes (low, moderate, and high) and examined the results separately for each regime. Our results indicate that the crisis altered the FDI inflows/corruption relationships and generally depend on the development level of the host country and corruption regimes. In addition, our evidence suggests a strong correlation between selected host country macroeconomic variables like financial depth, gross capital formation, and FDI outflows and sample variables. In turn, the crisis also affects these variables. Factor analysis results using all sample variables and separately for donor/aid recipient countries indicate the emergence of a few key factors (host country economic environment, host country international involvement, and host country economic growth potential) that are also affected by the crisis. Our analysis strongly suggests that the FDI/corruption is a complex one and corruption prone host countries must consider improvements to corruption, financial depth and domestic gross capital simultaneously to attract FDI inflows.

KEYWORDS:

• Inward Foreign Direct Investment
• Corruptions Perceptions Index
• global financial crisis
• nonlinear modelling
• developed
• developing
• and transition countries
• donor/aid countries

JEL Classifications:

• F21
• F23
• G15
• P45

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 The ‘grabbing hand theory’ suggests that doing business in highly corrupt countries tend to incur higher costs and higher risks (Shleifer and Vishny 1993; Bliss and Di Tella 1997; Wei 2000; Habib and Zurawicki 2002; Aidt 2003, and hence propose a negative relationship between inward FDI and host country corruption levels.

2 The ‘helping hand theory’ suggests a good relationship with host governments will allow firms to undertake investments even in corrupt countries. They can receive greater and possibly exclusive access to resources, operation processes acceleration, cost reduction, and competitive advantages in such countries (Lui 1985; Beck and Maher 1986; Saha 2001; Bjorvatn and Søreide 2005; Cuervo-Cazurra 2008; Khanna and Palepu 2011; Kolstad and Wiig 2013).

3 We thank an anonymous referee for directing our attention to this important paper. Consequently, we add development aid to the list of independent variables to control for any possible intervening effect of the development assistance on the stated relationship. Please also see footnote 9. Specific variables used in the study are outlined later.

4 We use 2009 since we expect the crisis to take time to affect investment decisions. Results are similar if a 2008 cut-off is used. The results of these alternate runs are available on request from the authors.

5 They also rebounded quicker than developed countries after the crisis.

6 We proxy the host country’s international competitiveness levels by the Trade/GDP ratio extracted from the World Bank website (http://www.worldbank.org/).

7 All other GDP and FDI data are extracted from the UNTCAD website (http://unctad.org/en/pages/home.aspx).

8 We have used the classification scheme provided by UNCTAD: http://unctadstat.unctad.org/EN/Classifications/DimCountries_DevelopmentStatus_Hierarchy.pdf. In this paper they add the following ‘ … There is no established convention for the designation of ‘developed’ and ‘developing’ countries or areas in the United Nations system. This section includes countries and territories classified according to three categories of development: developing economies, transition economies and developed economies … .’ (page 1, from http://unctadstat.unctad.org/EN/Classifications/DimCountries_DevelopmentStatus_Hierarchy.pdf).

9 Based on the literature review, and on the suggestions of two referees, a number of additional variables have been added to the study. For instance, one can question whether the results reported are robust to alternate available specifications (the ICRG Corruption Index) of the corruption proxy used in this paper. Referees have also suggested adding the one year lagged foreign direct investment net inflow as a percentage of GDP, FDI outflows from the host country, financial depth of the host country, level of domestic investment in the host country, development aid received by developing and transition countries, and official development assistance for developed countries. Unfortunately, adding more variables leaves us vulnerable to the availability of data on these additional variables to the extent that the validity of the overall statistical analysis can be brought into question. Hence, we restrict our analysis to the original set of variables. However, we understand the value of ensuring the robustness of results after inclusion of these additional variables. Hence, in an appendix, we report results of additional runs using data on sample variables and the results after incorporation of additional variables mentioned earlier, subject to data availability. Although we ran a total of 12 additional tables using these variables, only statistically reliable results using all of the additional variables are reported. Finally, since many of the control variables exhibit strong multi-collinearity, we conducted a factor analysis and used factor scores to examine the major contentions of this paper. These results are also reported in the Appendix and discussed in the main body of the paper. The results of any additional runs not reported are available on request from the authors.

10 Data presents results as variable (post crisis) – variable (pre-crisis).

11 Figures 1a (FDI) and 1b (ADJCPI) presents plots of the data; values are presented for pre and post crisis periods.

12 The sign and significance of this ratio is examined later once other control variables are included.

13 Unfortunately, a problem with this methodology is that it allows only for a small number of classes.

14 For more on the benefits of the PSTR over PTR and other linear models, see (Wang, Padmanabhan, and Huang, op. cit.).

15 The linearity test is also referred to as the homogeneity test.

16 Each regression result of a regime in equation 3 are presented in Table 8. Where the first part of equation 3 ${\text{β}}_1^{\mathrm{\prime }}(ADJCP{I}_{it}+\text{GDPP}{\text{C}}_{\text{it}}+\text{GDP}{\mathrm{\%}}_{\text{it}}+TRAD{E}_{\text{it}})$ is for the first regime. The second part of equation 3 $\sum _{\text{j}=1}^{\text{r}}{\text{β}}_{\text{j}}^{\mathrm{\prime }}{\text{g}}_{\text{j}}(ADJCP{I}_{\text{it}}^{\text{j}};{\text{r}}_{\text{j}},{\text{c}}_{\text{j}})(ADJCP{I}_{it}^j+\text{GDPP}{\text{C}}_{\text{it}}+\text{GDP}{\mathrm{\%}}_{\text{it}}+TRAD{E}_{\text{it}})+{\text{ε}}_{\text{it}}$ is for the second regime and third regime. The total impact of each independent variable on FDI is the sum of the coefficient of each regime as shown in equation ${\text{β}}_1^{\mathrm{\prime }}(ADJCP{I}_{it}+\text{GDPP}{\text{C}}_{\text{it}}+\text{GDP}{\mathrm{\%}}_{\text{it}}+TRAD{E}_{\text{it}})+\sum _{\text{j}=1}^{\text{r}}{\text{β}}_{\text{j}}^{\mathrm{\prime }}{\text{g}}_{\text{j}}(ADJCP{I}_{\text{it}}^{\text{j}};{\text{r}}_{\text{j}},{\text{c}}_{\text{j}})(ADJCP{I}_{it}^j+\text{GDPP}{\text{C}}_{\text{it}}+\text{GDP}{\mathrm{\%}}_{\text{it}}+TRAD{E}_{\text{it}})+{\text{ε}}_{\text{it}}$.

17 The parameter constancy (no remaining nonlinearity) test is a test of misspecification in the PSTR model (Equation (2)) for any value of r (where r is the slope of the transition function), and examines whether there is any remaining nonlinearity for one, two, or three transition functions.

18 Given the sensitivity of the results to the inclusion of other controls (discussed later) and the fact that the estimation methodology does not deal with endogeneity, we make no inferences related to causality, and caution must be exercised when interpreting all results. We thank and anonymous referee for directing our attention to this important point.

19 There were no regime 3 countries post crisis in the full sample.

20 Based on our results, although the overall level of FDI increased after the 2008 global financial crisis evidence suggests that firms are less (more) willing to invest in high (low) corruption countries post 2008. The marginal impact results show that for the full sample, during the pre-crisis period, one-unit increase in GDPPC increased FDI inflows by 1.98 units (1.047 units) in low (high) corruption countries. Different from the results of pre-crisis period full sample, during the post-crisis period, one-unit increase of GDPPC will negatively impact FDI both in low and high corruption countries. These changes indicate that firms’ FDI inflows target countries with lower GDP per capita after the 2008 global financial crisis for all subgroups of countries in the sample. These flows may be induced in part by lower production costs, including labor cost, raw material cost, etc., in these countries after a crisis. (Min 2016; Werner 2016; Palley 2018). Clearly, production costs became an important factor after the global financial crisis. Next, we explain marginal effect results associated with GDP %. Results suggest that for the pre-crisis period, one-unit increase in GDP% will change FDI by 1.263 (- 0.239) units in low (high) corruption countries. Post crisis, a one-unit increase in GDP% will impact FDI positively (negatively) in low (high) corruption countries. These results indicate an improvement in the attractiveness of the market post crisis in both low and high corruption developed countries. Market potential seems to be important in such countries post crisis. In contrast, market potential becomes a more (less) attractive factor in low (high) corruption developing countries post crisis, and signals a disinclination to invest in high corruption countries even with increases in GDP% (Maza and Villaverde 2015; Ozturk, Joiner, and Cavusgil 2015). Finally, TRADE seems to be negatively (positively) associated with FDI in low (high) corrupt countries before the global financial crisis and these relationships are unaffected by the crisis for the overall sample. In contrast, TRADE negatively impacts FDI for high corruption countries post crisis and indicate that even developed countries can experience a FDI decrease accompanied by increased TRADE in high corruption countries. In contrast, TRADE positively impacts FDI into highly corrupt developing and transition countries post crisis. These findings are consistent with the literature that trade liberation in developing/transition countries have attracted increased levels of FDI post crisis. (Dreher, Mikosch, and Voigt 2015; Osnago, Rocha, and Ruta 2016).

21 To link our findings with extant literature, it is noted that only the pre-crisis results are consistent with those presented by Cuervo-Cazurra (2008).

22 The literature reference to justify inclusion of these variables and the data sources for these variables are presented in the Appendix.

23 We thank anonymous reviewers for directing us to this important area of research.

24 The results of these alternate runs are available on request from the authors. They appear to be generally qualitatively similar to results reported for the ADJCPI variables.

25 Granted that the results presented in the Appendix includes a greater number of control variables, this discussion is based primarily on results reported with the smaller set of control variables. We believe that it is important to consider how the relationship affected transition countries and the extended data does not contain any transition countries as a separate set because of data availability.

26 One possible explanation for the marked decrease in corruption post crisis is that host countries take active measures (sometimes successfully) to combat corruption after a crisis. The crisis could be used to ‘clean house’. Our extended results suggest that a complex set of macroeconomic variables like financial depth and gross capital formation are also affected by corruption. These aspects are presented in the Appendix and discussed later in the paper.

27 However, the sample sizes using the alternate CPI becomes too small for any meaningful conclusions using the PSTR model (which requires a large size sample for stable and robust results. Hence we do not reproduce the results generated with ADJCPI using the ICRG corruption index.

28 They contend that for countries with high expropriation risk, official aid can support FDI inflows into such countries. We find that this contention is supported even for moderate corruption hosts.