Corruption inhibits private investment and economic growth (Shleifer & Vishny, 1993; Mauro, 1995), reduces the legitimacy of government (Anderson & Tverdova, 2003) and affects political and social stability (Gupta & Abed, 2002). The evidence shows that corruption represents a challenge not only for transition economies but also for developed countries (Bellos & Subasat, 2012). The World Economic Forum, for instance, estimates that the cost of corruption represents more than 5% of global GDP (approximately US$ 2.6 trillion) and increases the cost of doing business by up to 10% on average (OECD, 2014).

Corruption and its effects, especially on FDI, are a central issue in IB literature (Duanmu, 2011). However, there is no agreement concerning the effect of corruption on FDI (Cuervo-Cazurra, 2008). There are theoretical arguments suggesting a negative effect of corruption on FDI but there are also arguments suggesting a positive effect (Duanmu, 2011). Although there is evidence supporting both views of corruption, the empirical evidence suggests a prevailing negative impact on FDI (Wei, 2000a; Wei, 2000b; Habib & Zurawicki, 2002; Lambsdorff, 2003; Voyer & Beamish, 2004; Cuervo-Cazurra, 2006; Duanmu, 2011).

The relationship between corruption and ownership strategies of MNEs has received less attention in the literature (Duanmu, 2011; Di Guardo, Marrocu & Paci, 2016). One of the earliest studies on this topic was conducted by Smarzynska and Wei (2000). In their cross-sectional analysis of FDI in Eastern Bloc countries, high levels of corruption in the host country are posited to lead to a preference of shared control, such as JVs over WOS (Smarzynska & Wei, 2000). The authors argue that a lower level of ownership helps to reduce commitment and risk associated with the investment and may also be a sensible reaction to the uncertainty experienced in these transition countries (Smarzynska & Wei, 2000). Based on institutional theory, Uhlenbruck, Rodriguez, Doh and Eden (2006) suggest that MNEs respond to pervasive and arbitrary corruption in host countries by selecting specific modes of entry. Analyzing data on telecommunications development projects in emerging economies, the authors find that the more pervasive the level of corruption in the host country, the greater the possibility of MNEs choosing WOS over JV. Finally, Di Guardo, Marrocu and Paci (2016) found that the relationship between the level of corruption of the host country and the firm’s probability of choosing full control mode in cross-border mergers and acquisitions is U-shaped. Moreover, the authors argue that industry relatedness between acquirer and target and the level of connectivity between home and host countries moderate the relationship between the level of corruption of the host country and the level of control (Di Guardo, Marrocu & Paci, 2016).

Corruption distance is defined as the difference in pervasive and arbitrary corruption of the public sector between home and host countries (Eden & Miller, 2004). The corruption distance offers another perspective to analyze the effect of corruption on MNEs’ decisions. It allows to consider simultaneously the characteristics of home and host countries (Eden & Miller, 2004). Host country corruption has a direct impact on FDI but also an indirect effect: formal institutions in the host country may interact with institutions in the home country, which may themselves interact with informal institutions (Holmes et al., 2013), affecting the way foreign investors act (Cuervo-Cazurra, 2008). Thus, the level of uncertainty and the costs associated with corruption may vary depending on the host country, on the home country of the investors and on the differences in the corruption levels between the home and the host countries (Cuervo-Cazurra, 2006). Using the corruption distance concept, Tekin-Koru (2006) and Duanmu (2011) provided two of the few studies on the impact of corruption distance on entry modes. Both studies suggest that the higher the corruption distance between the home country and the host country, the greater the preference of MNEs for WOS over JV (Tekin-Koru, 2006; Duanmu, 2011).

When the levels of corruption of the host country are too high, corruption is considered institutionalized (Gabbioneta et al., 2013). In this case, public and governmental authorities establish regulations which facilitate corruption and represent a source of income for them (Cheung, 1996). The institutional environment stimulates forms of direct corruption, by facilitating its occurrence and providing opportunities (Gabbioneta et al., 2013; Neu, Rahaman & Martinez , 2013). MNEs from countries which have a large corruption distance may need to warrant country-specific knowledge to operate in a dissimilar country – from a corruption perspective (Di Guardo, Marrocu & Paci, 2016). However, given the institutionalized corruption, the cost of maintaining a local partner and choosing a shared form of control outweighs the benefits making firms prefer WOS over JV (Di Guardo, Marrocu & Paci, 2016).

MNEs’ decide to have a higher level of ownership to reduce the costs and the risks (Di Guardo, Marrocu & Paci, 2016). From the investing MNE’s perspective, particularly if the firm is from a transparent country, WOS provides a higher level of control over the way the business is conducted (Di Guardo, Marrocu & Paci, 2016) and ensures the respect of the values of the parent-firm (Gaur & Lu, 2007). The extra costs the investing firm may incur in addressing corrupt environments can be justified by considering the implications of protecting and sustaining its global image (Di Guardo, Marrocu & Paci, 2016). A higher control allows firms to reduce the costs associated with operating alone (Madhok, 1997) and makes possible for foreign investors to access and acquire organizationally embedded knowledge (Barkema & Vermeulen, 1998). Taking into account these arguments, we advance:

MNEs’ decide to have a higher level of ownership to reduce the costs and the risks (Di Guardo, Marrocu & Paci, 2016). From the investing MNE’s perspective, particularly if the firm is from a transparent country, WOS provides a higher level of control over the way the business is conducted (Di Guardo, Marrocu & Paci, 2016) and ensures the respect of the values of the parent-firm (Gaur & Lu, 2007). The extra costs the investing firm may incur in addressing corrupt environments can be justified by considering the implications of protecting and sustaining its global image (Di Guardo, Marrocu & Paci, 2016). A higher control allows firms to reduce the costs associated with operating alone (Madhok, 1997) and makes possible for foreign investors to access and acquire organizationally embedded knowledge (Barkema & Vermeulen, 1998). Taking into account these arguments, we advance:

The level of corruption in the host country has different effects on foreign investors, depending on the level of corruption of the home country (Godinez & Liu, 2015). The effects of corruption on MNEs’ decisions vary according to the direction of corruption distance (Duanmu, 2011). When the levels of corruption in the host country are higher than the levels of corruption in the home country there is a negative corruption distance between them (Godinez & Liu, 2015). Conversely, when the destination is less corrupt than the home country there is a positive corruption distance between them (Godinez & Liu, 2015). MNEs perceive and respond to corruption differently, according to their perceptions regarding the corruption level of their home country (Godinez & Liu, 2015). Several studies suggest that MNEs from more transparent countries avoid investing in highly corrupt countries (Habib & Zurawicki, 2002). When the country of origin has low levels of corruption many firms may not have the knowledge and the skills necessary to deal with corruption (Pajunen, 2008). On the other hand, firms from highly corrupt environments may not be as sensitive to high corruption levels abroad (Godinez & Liu, 2015). A high level of corruption in the home country helps firms to develop the skills needed to deal with the phenomenon of corruption (Cuervo-Cazurra, 2006). In this case, corruption is not an obstacle to investment and may even by a source of competitive advantage for these firms, comparing to those from transparent countries (Suchman, 1995; Cuervo-Cazurra, 2006; Cuervo-Cazurra & Genc, 2008).

When investing in a more corrupt country MNEs may prefer to have a local partner (Godinez & Liu, 2015). Higher corruption represents higher uncertainty, higher risk and higher costs for MNEs (Di Guardo, Marrocu & Paci, 2016). To deal with these problems, MNEs may establish a partnership with a local firm with market knowledge, sharing the risks and the costs associated with the investment and the control over the business operations (Smarzynska & Wei, 2000; Uhlenbruck et al., 2006). The establishment of a partnership with a local firm may be the only way foreign investors have to access to the market (Smarzynska & Wei, 2000; Uhlenbruck et al., 2006) and achieve the necessary external legitimacy (Beamish, 1985). Therefore, we hypothesize that when the direction of corruption distance is negative, the effect of corruption distance on ownership strategy of MNEs may diminishes.

When the levels of corruption of the host country are low or significantly lower than those of the home country, the benefits of full ownership balance its costs (Di Guardo, Marrocu & Paci, 2016). In countries with relatively good governance, i.e. “with a transparent, impartial and effective legal system that protects property and individual rights; with stable, credible and honest public institutions; and with government policies that promote open and free markets.” (Hernández & Nieto, 2015, p. 124), there is a better chance of firms entering through FDI (Globerman & Shapiro, 2003). Foreign investors should be able to operate easier in environments in which the “rules of the game” are more clearly established (Cuervo-Cazurra & Genc, 2011). When the host countries show significant lower levels of corruption than the home country, firms tend to adapt easier to those markets since they are entering less uncertain environments (Hernández & Nieto, 2015). Thus, foreign investors will feel safer to invest alone without the need for a local partner (Di Guardo, Marrocu & Paci, 2016). The benefits of total ownership balance the costs of the investment (Di Guardo, Marrocu & Paci, 2016) and a higher level of control allows firms to reduce the costs of operating alone. Considering these arguments, we propose the following hypothesis:

Hypothesis 2a.Negative corruption distance between the home country (Portugal or Spain) and the host country moderates the effect of corruption distance, thus making smaller the ownership MNEs will seek in local subsidiaries.

Hypothesis 2b. Positive corruption distance between the home country (Portugal or Spain) and the host country heightens the effect of corruption distance, thus making greater the ownership MNEs will seek in local subsidiaries.

Ownership. The dependent variable in our empirical models is the ownership level of Spanish and Portuguese MNEs in subsidiaries located abroad. Ownership can take any value over 0% and less than or equal to 100%. For this study, we excluded all observations where the ownership in local subsidiaries was lower than 5%. Thus, our dependent variable may take any value between 5 and 100%. Data on this variable were collected from Sabi database.

Corruption distance. The main explanatory variable of interest is corruption distance. Corruption distance is the difference in the levels of corruption between the home and host countries (Eden & Miller, 2004). The level of corruption is the degree of corruption of the public sector perceived by analysts, businesspeople and experts in a specific country. Corruption distance is a continuous variable and it corresponds to the absolute difference between the inverse of the value of the Corruption Perceptions Index (CPI) for the home country (Portugal or Spain) and the inverse of the value of the CPI for the host country. The CPI is an aggregate indicator that combines data on corruption from 13 independent global institutions. Although some researchers have emphasized the need to use more experience-based measures of corruption, the CPI still offers a reliable measure of perceived corruption in different countries, which significantly affects investors’ decisions (Mauro, 1995; Di Guardo, Marrocu & Paci, 2016). The CPI ranges from 0 to 100. A value of 0 means that the country is “highly corrupt” and 100 means the country is “very clean”. For the sake of clarity, we use the inverse of the original index so that low values indicate low levels of corruption and high values indicate high levels of corruption (Di Guardo, Marrocu & Paci, 2016).

Distance direction. This is a binary variable that takes the value 1 when the corruption distance is negative (i.e., the host country is more corrupt than the home country) and the value 0 when the corruption distance is positive (i.e., the host country is less corrupt than the home country). This variable allows us to identify the observations via the relative position of the host country compared to the home country in terms of corruption distance. The study constructs another independent variable – Corruption distance ´ Distance direction – by means of the interaction of these variables.

All estimated models include control variables that may influence the relationship between corruption distance and MNEs ownership strategies. These variables control for characteristics of the home country, characteristics of the host country and characteristics of the firms. The full sources and definitions of the variables are reported in Appendix I. Home country GDP. We used the natural logarithm of home country GDP to control for the size of the country. We collected the data from World Development Indicators, a World Bank report covering a wide number of issues.

Geographic distance. Geographic distance increases transportation and telecommunication costs (Cuervo-Cazurra, 2008; Ferreira et al., 2017) that would discourage trade and favor FDI (Cuervo- Cazurra, 2008). Therefore it is a factor that is likely to influence MNEs’ investment decisions. We used the natural logarithm of the distance in km between Lisbon or Madrid and the capital of the host country.

Host country GDP. We used the natural logarithm of host country GDP to control for the host country economic size (Cuervo-Cazurra, 2008). Data were collected from World Development Indicators.

Host country risk. We also controlled for the risk of the host country, since it may influence MNEs ownership strategy (Di Guardo, Marrocu & Paci, 2016). The indicator used is a composite index, computed by the Political Risk Service (PRS) group, which includes several dimensions, such as government stability, socioeconomic conditions, investment profile, internal and external conflicts, military in politics, religious tensions, law and order, ethnic tensions, democratic accountability and bureaucracy quality. This index ranges from 0 to 100 but once again we use the inverse of the original index so that a low score indicates lower risk and a high score indicates a high risk.

Host country unemployment. We used the unemployment rate of the host country to indicate the attractiveness of the country since investors are aware that employees will be more loyal when they lack other opportunities of employment (Godinez & Liu, 2015). Data were collected from World Development Indicators.

Investing firm size. Operating revenue (thousands of euros) was used as a proxy for MNE size. It has been argued that larger firms have greater resource availability to acquire a larger ownership in other firms (Chari & Chang, 2009). Data were collected from SABI.

Subsidiary size. The natural logarithm of the number of employees of the subsidiary was used as an indicator of the size of the subsidiary. Data were collected from SABI.

We used an OLS (Ordinary Least Squares) regression model in this study to analyze the effects of corruption distance on international ownership strategies of Spanish and Portuguese firms. The OLS approach is appropriate given the use of continuous variables in the study. Moreover, we conducted a regression diagnostic using the variance inflation factors (VIF) to identify any potential problems associated with multicollinearity. The test did not suggest any serious problem of multicollinearity between variables since none of the VIF values exceeded 8.475, which is lower than the threshold of 10 used by most scholars (O’Brien, 2007).

The dependent variable is measured at the end of 2015, while the explanatory variables are measured one year earlier to account for the time lag that occurs between the decision regarding the ownership strategy and the actual implementation of the strategy.

The baseline model is specified as follows:

where Own_Start sub is the dependent variable which takes any value between 5 and 100%; B0 is the constant; B1, B2, B3, B4, B5 and B6 are the coefficients of interest; is the absolute corruption distance between the oth country of origin and the dth destination country; Distance direction is a binary variable that takes the value 1 if corruption distance is negative and 0 if corruption distance is positive; (Corruption distance * Distance direction) presents the interaction between the absolute corruption distance between the oth country of origin and the dth destination country and distance direction; Culture distance is the cultural distance between the oth country of origin and the dth destination country;

represents the moderation of the relationship between the absolute corruption distance between the oth country of origin and the dth destination country and the cultural distance between the oth country of origin and the dth destination country; represents the control variables and is the error term.

It should be noted that in those observations in which the host country displays a higher level of corruption than the home country, i.e., there is a negative corruption distance between both countries (Distance direction equals 1), the model is determined by:

And in the opposite case, for those observations in which the host country displays a lower level of corruption than the home country, i.e., there is a positive corruption distance between both countries, (Distance direction equals 0), the model is determined by the following equation:

In terms of interpreting the results, then, the coefficient of Corruption distance (B1 ) is used to test H1 and H2b and the coefficients of Corruption distance and Corruption distance ´ Distance direction are used to test H2a (requiring us to add B1 and B3 together). The coefficient of Distance direction (B2) represents the difference in the constant term of the observations of Distance direction with values equal to 0 and 1. Hypothesis 1 and 2b are supported if B1 is positive and statistically significant. Hypothesis 2a is supported if (B1 + B2) is negative and statistically significant.

In addition to the tests presented above, we tested the 3 hypotheses for Spanish and Portuguese subsidiaries separately. For the Spanish subsidiaries, the results match the results presented above, supporting Hypotheses 1, 2a and 2b. For the Portuguese subsidiaries, none of the Hypotheses is supported. Since the home country characteristics (concerning corruption) are similar, we further investigated the sub-samples for firm-level characteristics (Table 3). We concluded the three firm- level variables (Ownership, Investing firm size, and Subsidiary size) are significantly different in the Spanish and Portuguese sub-samples (for p<0.05). This suggests other factors may influence the ownership strategy of Portuguese firms. We further elaborate on the discussion section.

Table 3
Results of t-test for independent samples

The empirical findings are subject to some limitations that emerge from the nature of the data presented. First, Spanish subsidiaries represent 80% of our sample and Portuguese ones only 20%. A more evenly distributed sample could produce more accurate results. Second, we did not account for the industry to which the parent-firm and subsidiary belong. Some industries in various countries limit foreign equity participations. In that case, the ownership strategy is determined by legal restrictions and not by any of the variables used in this study. Finally, this is a cross- sectional study that makes it impossible to explore the effects of ownership determinants over time. Future research can explore the impact of corruption distance on ownership strategies over time, in a longitudinal study.

In the future, it would be interesting to study the effect of corruption distance on the ownership strategy of firms with subsidiaries located in Portugal and Spain. Since corruption distance seems not to have any significant impact on ownership strategies of Portuguese MNEs, it would be also interesting for future work to analyze other institutional factors that may influence these decisions. The recognized impact of the institutional environment makes it necessary to explore the effect of institutional dimensions, such as the role of the government or the impact of legal issues, on the international ownership strategies of Portuguese firms. This effect is still little known.

The effect of corruption on MNEs’ decisions has been receiving scholarly attention for a long time. Nevertheless, the full extent of corruption on MNEs’ is still underexplored. We contribute to shed light on this issue, specifically analyzing the differences in corruption levels between home and host countries. We conclude that greater differences lead firms to seek a greater degree of control. Furthermore, we investigate the effect of asymmetric differences in corruption levels. Our findings reveal that when operating in a country which is more corrupt than the home country (i.e. negative corruption distance), the MNEs will seek a lower equity position. In face of increased uncertainty caused by a more corrupt environment, MNEs rely more on local partners with more knowledge of the institutional setting.