FOREIGN DIRECT INVESTMENT IN MEXICO, CRIME, AND ECONOMIC FORCES.
According to the 2016 World Investment Report, prepared by the United Nations Conference on Trade and Development (2017), Mexico was the 13th largest foreign direct investment (FDI) recipient in 2014 ($26 billion) and 2015 ($30 billion). The reason behind such outstanding FDI performance is Mexico's macroeconomic discipline and the recent enactment of several important economic reforms in the energy, telecom, and financial sectors. Indeed, Mexico's productivity has recovered since 2009, even in the face of external shocks and an uncertain world economy (Organisation for Economic Co-operation and Development 2017).
Nevertheless, despite Mexico's sound economic policies and ambitious structural reforms, drug-related crimes have escalated considerably across the country in the past few years. This has led some studies to measure the effect of drug-violence on economic variables at the country, state, and municipal levels. In the international arena, several studies have been proposed to analyze the effect of crime on economic growth. Most researchers conclude that there is a negative correlation between per capita output and crime. Cardenas and Rozo (2008) for Colombia; Estrada and Ndoma (2014) for Guatemala; and BenYishay and Pearlman (2014), Enamorado, Lopez-Calva, and Rodriguez-Castelan (2014), and Balmori de la Miyar (2016) for Mexico offer evidence for Latin America. Besides, the relationship might be nonlinear as discovered by Detotto and Otranto (2010) who identify asymmetric impacts of crime in recession and expansion periods in Italy. More recently, Goulas and Zervoyianni (2015) examined crime and per capita output growth in a panel of 26 countries and observed that potential gains exist from reducing crimes during periods when markets are pessimistic. On the other hand, when markets are optimistic and there is economic growth, crime is not found to be an important deterrent to economic growth.
Regarding the impact of crime on the Mexican economy, Albuquerque (2007) finds strong geographic concentration in homicides on the Mexican side of the U.S.-Mexico border. Rodriguez-Oreggia and Flores (January 2012) developed spatial econometric analysis at the Mexican municipal-level for the 2007-2008 period and their results showed that federal intervention is effective in reducing crime. Pan, Widner, and Enomoto (2012) observe that the growth rate in per capita real gross domestic product (GDP) in any given state is positively influenced by the growth rate of such variable in the neighboring state, but negatively correlated with the crime growth rate in all neighboring states. BenYishay and Pearl man (2014) acquired strong evidence to show that higher theft rates reduce the probability of Mexican microenterprises expanding their operations. Gonzalez-Andrade (2014) concludes that there is a small negative relationship between economic growth and crime rates. Balmori de la Miyar (2016) finds that there is a 0.5% decrease in GDP per capita in states with military operations. BenYishay and Pearlman (2013) study how homicides affect the hours worked by employees in different categories; they found that the hours worked drop between 1 % and 2% and the larger impact is on the self-employed. Cabral, Mollick, and Saucedo (2016) analyzed Mexico's labor productivity (GDP per worker) across its 32 subnational states from 2003 to 2013 and found that crime had negative effects on the Mexican labor productivity during the "war on drugs" period, which started in 2006. (1)
Almost all of these studies on crime in Mexico look at the effects on real GDP per capita or labor productivity. This study, however, examines the impact of crime on FDI inflows at the state level. To do so, we turn to the literature on the determinants of FDI, which is vast and conflicting. The problems range from estimation methods to the choice of control variables. For a recent example of the FDI location decision in Chinese cities, there is the study by Blanc-Brude et al. (2014) who employed spatial panel methods and a vast array of explanatory factors, such as GDP per capita, agglomeration measures, human capital, distance to the coast, trade openness, wages, and government revenues as a percentage of GDP and as a percentage of the rural population, among others. In another recent example of the study of an economy suspected to be affected by crime, Daniele and Marani (2011), using panel data for 103 Italian provinces for the period 2002-2006, and controlling by population size and GDP per capita, found that organized crime has a negative effect on FDI inflows.
Part of the mixed effects obtained from the determinants of FDI may be owing to the reverse causation that exists between FDI inflows and the real economy. For example, there is no clear general information in the literature regarding the effect of FDI on wages. Some argue that FDI does not have a general positive effect on wages but instead creates a higher positive impact on skilled wages than on nonskilled wages, and thus, generates a difference between these two types of wages. For the Mexican economy, in particular, Feenstra and Hanson (1997) documented rising wage differentials in maquiladoras; Chiquiar (2008) examined the changes in regional wage differentials in the 1990s; and Kato-Vidal (2013) searched for crowding-out effects of foreign investment on wages. In a study of more than 100 developing and developed countries, Figini (2011) discovered that the effect of FDI on wages differs between countries depending on each country's level of development. While FDI increases wage inequalities in developing countries, in developed countries wage inequality decreases.
Another example of the bidirectional effects can be seen in the exchange rate. A weaker Mexican peso may make foreign acquisitions more attractive to foreigners. Several studies analyze the effect of the exchange rate on FDI inflows in host countries. Most of the literature agrees that the value of the local currency, expectations about the value of the local currency, and exchange rate volatility are important factors for modeling FDI inflows. For example,
Kiyota and Urata (2004), in a Japan-based study, and Udomkerdmongkol, Morrissey, and Gorg (2009), in a study for 16 emerging market countries, observed that an expensive local currency, the expectation of depreciation in the local currency, as well as volatility in the exchange rate, are factors which have a negative impact on FDI inflows. Further, Martinez and Jareno (2014), in a Latin America-based study, noted that the exchange rate is an important and significant factor in explaining the FDI evolution in this region. Russ (2007) also suggests that a multinational firm's response to increases in exchange rate volatility differs depending on whether the volatility arises from shocks in the firm's native or host country. Other authors, however, found no significant link between exchange rate and FDI flows. They include Dewenter (1995), who examined data on foreign acquisitions of U.S. targets during 1975-1989 and, more recently, Xaypanya, Rangkakulnuwat, and Paweenawat (2015) who looked at ASEAN countries. Obviously, the cost of capital is important for investment decisions and the interest rate may signal to foreign companies the willingness to control inflation and reduce macroeconomic uncertainty.
In addition to wages, exchange rates, and interest rates, other factors may also play a role, which could make "omitted variables" the reason for inconclusive literature. Majocchi and Presutti (2009), in an Italy-based study, discovered that clusters and agglomeration of foreign firms are important factors to attract multinational firms, which have a high entrepreneurial culture. Kolstad and Villanger (2004), by studying 75 countries, observed that improvements in political rights and civil liberties increase FDI inflows, while reductions in corruption in the host country do not have any effect on FDI inflows. Both Thomas and Grosse (2001), in a Mexico-based study, and Agarwal and Feils (2007), in a Canada-based study, found that political risk, the level of economic development, and the attitude of the host country are important factors to explain FDI inflows. Kinda (2010) used firm-level data across 77 developing countries to show that physical infrastructure problems, financing constraints, and institutional problems discourage FDI inflows. Further, Asiedu and Lien (2011) employed dynamic panels for 112 developing countries from 1982 to 2007 and found that democracy promotes FDI arrivals under certain conditions, notably, if the value of the share of minerals and oil in total exports is low. Luo, Luo, and Liu (2008) also examined multinational corporations and observed that previous levels of FDI arrivals, human capital, and geographic distances and cultural gaps are important factors influencing FDI inflows to a country. Agarwal and Feils (2007) and Xaypanya, Rangkakulnuwat, and Paweenawat (2015) found that infrastructure facilities and level of openness are important factors to explain FDI inflows to a host country. In a Mexico-based study conducted by Alarcon-Osuna (2016), during 2007-2012, the results showed that tertiary and postgraduate enrollments have positive effects on FDI inflows, which, in turn, highlights the role of education. Meanwhile, using aggregate data, Vazquez-Galan and Oladipo (2009) and Cuadros, Orts, and Alguacil (2004) observed that FDI Granger causes exports but not the other way around, while Pacheco-Lopez (2005) reports evidence of bidirectionality.
Previous works for Mexico on crime and FDI at the state level include two papers by Ashby and Ramos (2013), who in one study estimated the impact of panels of FDI inflows into Mexican states and sectors of activity on economic factors and homicides, and then in another study, Ramos and Ashby (2013) linked the entry of FDI into Mexican states as follows: the negative impact of crime on FDI inflows decreases as the level of organized crime in foreign entrants' home country increases. Garriga and Phillips (2015) found evidence to show that measures of criminal violence and organized crime are not directly associated with FDI inflows, while factors theorized to attract FDI (democracy, economic growth, and average schooling years) are only associated with FDI entries when homicide rates are low.
Our paper adopts the determinants of the FDI empirical framework used before for Mexican states by Mollick, Ramos-Duran, and Silva-Ochoa (2006), Ashby and Ramos (2013), Ramos and Ashby (2013), Garriga and Phillips (2015), and Alarcon-Osuna (2016), but is not limited to homicides as the main measure of crime. We also explore the variability of quarterly data, together with a rich dataset of criminal statistics. Our first important finding is that a predefined lag-length (e.g., k = 1 or k = 4) for our model does not capture the effects of crime on FDI. Once a flexible lag-length approach is introduced, we observe that homicides and thefts have negative and statistically significant effects on FDI in Mexican states, while other crimes show no effects. (2) For a detailed examination of the magnitude and prevalence of the effects of crime on FDI inflows, we split the sample across the 16 most violent and least violent states (by ranking states in descending order for each kind of crime). The results suggest higher effects in more violent states; that is, coefficients for both homicides and theft become larger in absolute value among the most violent states. We interpret our findings and discuss alternative empirical models so they can serve as robustness checks.
The structure of the paper is as follows: Section II describes the dataset and includes maps of Mexico at state-level to illustrate the FDI patterns and homicide rates across the states. Section III describes the empirical model and Section IV details the results. The last section of the paper offers concluding remarks.
II. THE DATA
The dependent variable used in this paper is total FDI receipts by Mexican states. FDI data are available on a quarterly basis from the Ministry of the Economy in current U.S. dollars. Those figures are transformed into real dollars using the U.S. consumer price index (CPI) (2010= 100), which comes from the U.S. Federal Reserve.
Table 1 presents some descriptive statistics of the variables in our empirical model. The most important thing to note is that FDI averages US$201 million per quarter, but its volatility is significant, with foreign capital flows ranging from -1,290 to 3,800 million. The reason for these negative flows is that there may occasionally be capital repatriations by foreign multinationals. As our model is estimated in logs, we do not take into consideration negative FDI outflows. Once we omit nonpositive flows, the variability is reduced, the minimum FDI flow observed is US$252,330, and we end up with 1,304 observations (down from the 1,408 observations shown in Table 1). Figure 1 shows the evolution of the FDI log on a state-by-state basis. It indicates a fairly steady path of FDI inflows, even toward the end of the first half of the 2008-2009 financial crisis. Complementing the description of FDI flows, Figure 2 shows that the foreign capital inflows have been geographically concentrated mostly in the northern and some central states of Mexico.
In Table 2, we provide average state indicators for the main variables in our model over the sample period 2005 Ql -2015 Q4. It is clear that Mexico City, Mexico's capital, receives significantly more FDI inflows than the rest of the country (US$1,310 million on average per quarter) and its numbers are well above the national average. (4) Nuevo Leon and Estado de Mexico follow Mexico City, receiving US$621 million and US$612 million on average per quarter, respectively. Our estimates indicate that for the period analyzed, Mexico City is the most important recipient of foreign capital with about 20.3% of the total FDI inflows received by the country, followed by Nuevo Leon, Estado de Mexico, Chihuahua, and Jalisco. Together, these five states have captured 51.4% of the average FDI received by Mexico from 2005 to 2015. At the same time, the five states with the lowest FDI values are Yucatan, Tlaxcala, Campeche, Colima, and Chiapas; together, they captured just 2.4% of the average FDI flows received over the sample period. This dispersion of FDI receipts observed across the states in the sample may have spillovers beyond these local economies in terms of job creation, acquisition of new techniques, and so on.
Regarding the crime variable, the data series employed is collected from the National Public Security Executive Office (Secretariado Ejecutivo del Sistema Nacional de Seguridad Publica 2016). Those official crime statistics are published on a monthly basis and are converted into quarterly observations by aggregating monthly data into quarters. The four crime variables included in the analysis are homicides, thefts, property crimes, and other crimes, which includes kidnapping, extortions, and fraud, among others. Homicides, thefts, and property crimes account for roughly 60% of the total crimes reported by the source.
Looking at the homicide rates in Table 1, we find that the quarterly average is 7.3 per 100,000 people, roughly 29 homicides per year. However, the variability across states is significant as shown in Figure 2. For example, the peaks in homicide series in Chihuahua, Durango, and Sinaloa are visible in the middle of the period and are probably linked to the intensification of the "war on drugs" in Mexico. Figure 3 shows the mean state-level homicide rate for the period analyzed in this paper. Results show that in terms of homicides, the states with the highest quarterly rates are Sinaloa (16.44 homicides per 100,000 people), Chihuahua (15.55), and Guerrero (14.55). On the other hand, the states with the lowest murder rates are Yucatan (2.47), Baja California Sur (3.78), and Campeche (3.84). These numbers imply that homicides are highly concentrated in some geographic regions of the country. For example, the number of homicides for every 100,000 people in Sinaloa is seven times higher than the number of homicides committed in Yucatan. Figure 4 also shows that the states in the center of the country are mostly ranked among those with the lowest level of homicides.
Looking at crimes other than homicides in Table 2, the highest crime rates are found in Baja California (486 per 100,000 people), for property crimes in Baja California Sur (139.63), and for other crimes in Yucatan (415), while the lowest rates for thefts, property crimes, and other crimes are seen in Campeche (17.7,6.5, and 24.6, respectively). Overall, homicides are less frequent (quarterly mean of 7.3), followed by property crimes (59). Thefts and other crimes are more common (means of 142 and 155, respectively). However, the ratio of mean to standard deviation is similar, ranging from 1.46 for thefts to 1.62 for property crimes.
For the control variables, we use wages, taken from the Mexican Social Security Institute (Instituto Mexicano del Seguro Social) for employees in the formal sector (workers with social security benefits), as used recently by Mollick and Cabral (2015) for Mexican regions in post-North
American Free Trade Agreement (NAFTA) years. Daily wage data are available on a monthly basis; quarterly figures are obtained by calculating the average wage over the corresponding 3-month periods. We then convert the nominal quarterly average wages into real U.S. dollars by dividing them by the nominal peso-dollar exchange rate and then deflating them using the U.S. CPI. As shown in Table 2, the states that pay the highest daily real wages are Mexico City (US$24.8), Campeche (US$21.2), and Nuevo Leon (US$20.8), while those that pay, on average, the lowest wages are Yucatan (US$13.4), Sinaloa (US$13.5), and Chiapas (US$13.5).
We also include the real exchange rate (RER) and the interest rate as determinants of FDI inflows; both variables are taken from Mexico's Central Bank (Banco de Mexico 2016). Both indicators are country-level rates, and for that reason are constant across states. The RER is calculated as a basket of currencies weighted according to the importance of Mexico's trade with other countries and is an indicator of Mexico's competitiveness--an increase in the RER indicates a real depreciation of the Mexican peso. Daily RER observations are averaged to obtain quarterly observations. The interest rate used (the equilibrium interbank interest rate from the Spanish acronym THE, "tasa de interes interbancaria de equilibrio") can be considered as the Mexican counterpart of the United States' federal funds rate. It is available on a daily basis and the quarterly data are also obtained by averaging the daily rates. Finally, we include in the model electricity sales at the state level, which may serve as a proxy for the infrastructure level in the state and even of economic activity. Data are available on mega-watts per hour from the Ministry of Energy on a monthly basis and are accumulated to generate quarterly series.
In Table 3, we report the correlation matrix for the main variables of our model. Real FDI inflows correlate positively with real wages (0.45) and electricity sales (0.51). The correlation between
FDI inflows and crime series varies from negative and small for homicides (-0.10) to positive for thefts (0.22).
The correlation between FDI and interest rates is weakly positive (0.09) and that between FDI and the RER is negative (-0.10). Overall, we do not find high correlation rates that could signal multicollinearity problems in the empirical model of FDI inflows detailed below. (5) Not every crime is negatively correlated with FDI flows. This could suggest some quite complex dynamics and heterogeneous effects of crime in our econometric model. We address this problem by adjusting long-term lags to our FDI determinants. In addition, we also observe that not all the crimes are positively correlated among themselves. Property crime is positively correlated with theft crime (0.61) and with other crimes (0.66), but the correlation between homicides and thefts is weaker (0.096), as is the correlation between homicides and property crimes (--0.16) and between homicides and other crimes (-0.10). This could be attributed to the fact that, as observed in Table 2, not all crimes necessarily rank equally across states.
III. EMPIRICAL MODEL
Our empirical specification controls for traditional domestic and foreign determinants of FDI plus crime in its various forms. Defined in logs, the empirical equation employed is given by:
(1) [mathematical expression not reproducible]
where FDI stands for the total FDI inflows received by each of the 32 subnational Mexican states; W refers to real wages; CRIME is any of the violence measures discussed in the previous section (homicides, thefts, property or other crimes). In addition, X is a set of control variables, which includes domestic and foreign factors that may influence FDI decisions. Across the set of domestic factors, we consider real wages, the interest rate, and electricity sales, which is our proxy for infrastructure (as well as another measure of economic activity). The RER and a dummy that control for the effects of the financial crisis can be taken as foreign factors affecting FDI flows to Mexico. We therefore combine domestic and external factors, as used in the literature on FDI determinants in Mexico. (6)
Since our data have quarterly frequency and because FDI inflows respond to our different controls with delay of unknown order, we introduce in Equation (1) several lags of our explanatory variables. To define the appropriate number of lags for W, CRIME, and each of our control variables in X, we use a flexible approach starting with a maximum of six lags (i.e., a year and a half)- (7) We then test the statistical significance of the coefficient and shorten the lag by one period if we cannot reject the null hypothesis that the effect of the longest lagged coefficient is zero, at least at the 5% significance level. We continue shortening the lag-length until the trailing lag coefficient is statistically significant or leave just one lag even if the independent variable is not significant. This procedure implies that the minimum lag-length is one quarter for each (right-hand side [RHS]) series in Equation (1) and the maximum is six quarters.
FDI decisions usually involve time for capital flows to be redirected to new productive facilities. For this reason, it is reasonable to expect some delay between economic and social developments and the decision to send new capital flows to plants. Previous works have adopted this perspective in various contexts. Fielding (2004) examined the dynamics of the causal links between the intensity of the Israeli-Palestinian conflict and Israeli investors' decisions about where to locate their physical capital assets.
Besides economic factors, he included indicators associated with the level of violence and political instability. The results indicate that current conflict intensity and conflict intensity in previous periods lead to substantial capital flight. Ramos and Ashby (2017) in a Mexico-based study for the 2001-2015 period, proposed the existence of a geographic halo effect, whereby, in the face of signals of high levels of violent criminal activity in specific places within a country during the previous periods, foreign investors draw overly general impressions about the country, which negatively impacts investment. Their results show that the highest number of state homicides is associated with lower FDI across states.
In general, we expect the full long-term effect of wages on FDI to be positive [mathematical expression not reproducible], because of its role as proxy of labor productivity, although real wages may also reflect labor costs and their ultimate effect on FDI is ambiguous. For crime, in its various forms, we expect the full long-term effect to be invariably negative [mathematical expression not reproducible] because of the effect of violence and social unrest on business climate. A similar logic is expected to explain the rest of the control variables in our model: interest rate and electricity sales are both expected to have positive cumulative effects on total FDI. Sales of electricity, our proxy for infrastructure, signal a more suitable and attractive environment for FDI flows across Mexican states. We interpret the interest rate as a proxy for domestic capital productivity. Owing to the expected complementarity between domestic and foreign capital, an increase in the interest rate is assumed to have a positive effect on FDI. Finally, in theory, an increase of the RER would imply a real depreciation of the peso against other currencies, which makes exports more competitive; thus, FDI is more likely to flow into Mexico. In this sense, when the RER moves up (a real depreciation of the peso against a basket of more than 100 currencies), the peso becomes weaker, which makes the USD figure of the dependent variable (FDI inflows in USD) lower. To the extent that the USD is a major currency in the basket to calculate the RER, a real peso depreciation will lead to lower FDI inflows in USD, thereby implying a negative coefficient for RER.
Given the potential endogeneity of some RHS regressors included in Equation (1), we also revisit the determinants of FDI and the influence of crime using a dynamic specification. Following the estimation of Equation (1), we define a more general equation in which we control for the persistence or inertia of FDI flows, and the potential endogeneity implicit in some of our RHS determinants. The empirical model is thus:
(2) [mathematical expression not reproducible]
A problem that emerges while estimating Equation (2) using ordinary least squares is that the lag of the dependent variable is correlated with the error term. The generalized method of moments (GMM) procedure derived by Blundell and Bond (1998) allows us to correct this bias by instrumenting the lagged dependent variable and any other endogenous regressor using lags and lagged differences of the suspected endogenous variable. GMM estimators are said to be consistent in the absence of second-order autocorrelation in the residual and if the instruments employed are valid according to the Hansen test of overidentified restrictions. As we need to restrict the number of instruments to avoid overidentification problems rather than using lags of the RHS regressors (except for FDI), we only employ the contemporaneous effects of crime, as well as of the domestic and foreign factors implicit in the vector of controls.
We start by employing a fixed lag-length approach for the estimation of Equation (1), with k - 1 and k = 4. (8) In Table 4, we observe that when k = 1 is used, there are no statistically significant effects in any of the columns, both for the baseline in column (1) and for the next four columns with different types of crime. In all cases, the [R.sup.2] is 0.02 (within) or 0.01 (overall). If correct, this would suggest that none of the series used is able to help explain FDI inflows across Mexican states, despite previous studies reporting the contrary. When the fourth lag is included in the regression (k = 4), the coefficients associated with wages ([[beta].sub.2]) and crime ([[beta].sub.3]) remain statistically insignificant. Interest rates have a positive coefficient (around 1.5) only at one lag, electricity has a positive coefficient (around 0.90) only at two lags, and the RERs have negative coefficients in three out of the five columns (around -3.3) only at one lag. All these are consistent with expected signs and the overall fit of the model improves with the [R.sup.2] moving up to 0.07 (within) or 0.06 (overall). It is, thus, reasonable to conclude that some additional dynamics are necessary. It is less clear, however, that 1 year (k = 4) captures the necessary dynamics. We conjecture that the arbitrary lag-selection in Table 4 is not a good choice since each factor is assumed to have the same dynamic effect on FDI inflows, a very questionable assumption. For the nature of the socioeconomic variables and given FDI as the dependent variable, there are indeed reasons to suspect that some flexible lag-length structure is needed. To allow for sufficient dynamics and also to avoid overparametrizing the fixed-effects model, we adopt a data-dependent procedure, following Campbell and Perron (1991), who chose the optimal number of lags in the augmented Dickey-Fuller unit root tests. This flexible autoregressive structure was explained earlier in Section III and leads to Table 5.
Table 5 presents the estimations of Equation (1) using fixed-effects methods for each crime subcategory (homicides, thefts, and property and other crimes) based on our approach of identifying the lag-length of each regressor. In column (1), we first estimate a benchmark model without crime, while columns (2) to (5) take into consideration the effects of the different crime categories on FDI flows. The number of lags is consistent across our five models and for each variable except crime. First, we observe that only homicides and theft crimes, perhaps considered high-impact crimes, have negative and significant effects on FDI; the first form of crime is statistically significant five quarters after a crime is committed and the latter as long as six quarters after. Interestingly, the effect of thefts on FDI inflows is slightly higher (coefficient of -0.33) than on homicides (coefficient of -0.28). Property and other crimes are found to have no statistically significant effects on FDI inflows.
The effect of wages on FDI is positive and statistically significant at the 5% significance level, with statistical significance at three (>0), five (<0), and six (>0) lags. For column (1), the net effect is positive using the 5% significance level. Interest rates also present the expected positive sign meaning that this variable could act as a proxy for capital productivity, and thus, there is a complementarity effect between domestic and foreign capital. The interest rate coefficient ranges from 0.26 to 0.36, with the net effect being 0.34, as seen from column (1). Electricity, our proxy for infrastructure, is not statistically significant at any relevant level. For the RER, its cumulative effect ranges from -4.02 in the baseline column (1) to -4.32 in column (3). Finally, although the financial crisis dummy shows the expected negative sign, it is not statistically significant, which suggests FDI inflows are very persistent and not subject to transitional factors.
Overall, we find positive long-run effects of wages and interest rate on FDI. The dummy variable that captures the financial crisis is not statistically significant, suggesting that this event may not have disturbed the allocation of foreign investment arriving in Mexico. With respect to our various types of crime series, we observe negative statistically significant effects of crime on FDI for only two subcategories: homicides and theft crimes. On average, a 1% increase in the homicide rate causes a reduction of 0.28% in FDI flows after five quarters. Meanwhile, a 1% increase in theft crime causes a long-term contraction of 0.33% in FDI after six quarters.
V. SAMPLE PARTITIONS
Following this initial set of results, we split our sample into two groups containing the 16 most and least violent states in Mexico, according to the two subcategories that were significant in Table 5: homicides and theft crimes. The aim of this sample partition is to conduct a detailed examination on the magnitude and prevalence of the effects of crime on FDI inflows. We follow the same data-dependent procedure to establish the suitable number of lags in our RHS variables, provided that the dynamics of the subsamples are not the same as for the full sample.
We notice two important results from these partitions reported in Table 6. First, in homicides we see that the 16 most violent states are the ones that are significantly affected in their FDI receipts. This time, the effect is felt only three quarters after crimes are committed and in a stronger way than across the full sample, with a coefficient of -0.43 (relative to -0.28, when we estimate the model for the full sample of 32 states where the effects of homicides are felt five quarters after crimes were committed). No effects on FDI are observed across the 16 states with the lowest homicide rates. The second important finding in this partition is that theft is relevant across both subsamples, but differently. The 16 states with the highest theft rates saw a negative impact on FDI only three quarters after a crime was committed, much faster than across states with lower theft rates in which, as with the full sample, the impact on FDI is observed only after six quarters. In addition, the impact on the former group is also more than threefold that of the latter (coefficients of -0.97 vs. -0.30, respectively).
A problem with this sample partition is that the coefficients for interest rate and the RER are now quite unstable: interest rates become negative and coefficients of RER change signs. This might have to do with the reduction in the sample size, but also perhaps with the implicit endogeneity of some of our RHS variables. To deal with the latter problem, we address endogeneity concerns and the persistence of our dependent variable by employing a dynamic specification under the system generalized method of moments (SGMM) estimation techniques.
VI. DYNAMIC SPECIFICATION
Table 7 presents the estimations of Equation (2), first for the benchmark model with no crime effects in column (1) and then for each of the crime categories in columns (2) to (5). For this specification, we do not fit lags of our control variables, but rather we use as instruments the lags and lagged differences of FDI, real wages, the RER, and interest rate. To avoid overidentification of the model, we employ the collapse procedure suggested by Roodman (2009) to minimize the lagged instruments and achieve a specification with fewer instruments (14) than cross-sectional units (32). (9)
The results show that there is significant persistence of FDI flows with coefficients ranging from 0.53 (homicides) to 0.68 (other crimes), except in the model featuring property crimes, when the standard error of the lagged dependent variable is larger. Nonetheless, crime effects--including homicides and thefts--are muted, along with the effects of most of the other determinants of FDI. While the Hansen test and the Arellano-Bond test of second-order serial correlation suggest that our models are correctly specified, the ability of the model to capture FDI dynamics appears quite limited at this stage. Column (5) for other types of crimes presents some statistically significant effects, including the crisis dummy as in column (1), although the coefficient of other crimes (estimated negative at -2.29) has a p value of .055.
VII. CONCLUDING REMARKS
This paper re-examines the determinants of FDI inflows in Mexico using not only standard factors previously studied in the literature, but also different types of crime (homicides, thefts, and property and other types of crime), which may have affected the decisions by foreign companies to invest in Mexico. The research design is built on a very simple premise put forward by Detotto and Otranto (2010): "Criminal activity acts like a tax on the entire economy: it discourages domestic and FDIs, it reduces firms' competitiveness, and reallocates resources creating uncertainty and inefficiency." Mexico provides an excellent case study to reconsider this research question given the rise in crimes associated with the "war on drugs" and also given recent concerns related to changes of trade patterns with the United States, which could make the stability of FDI inflows even more important than in the past. (10)
In addition to the vast literature on the determinants of FDI across states or cities, this study builds upon previous works by Ashby and Ramos (2013), Ramos and Ashby (2013), and Garriga and Phillips (2015) for Mexico, but is not limited to homicides as a measure of crime and also explores the variability of quarterly data, which, to the best of our knowledge, has not been done before. Starting first with a fixed lag-length approach to model FDI determinants, we are unable to capture any significant effect of crime on FDI. Under a flexible lag-length approach that captures the changing dynamics of crime and other factors over time, we observe that homicides and theft crimes have negative statistically significant effects on FDI into Mexican states, while other crimes show no effects. When we split the sample across the 16 most violent and least violent states, to observe the magnitude and prevalence of crime, the results suggest higher effects in more violent states with coefficients for both homicides and theft becoming larger in absolute value among the most violent states. We also find that dynamic panels indicate a significant degree of persistence in FDI inflows and in most cases, the lagged dependent variable is responsible for all of the statistical significance, ranging from .53 (homicides) to .68 (other crimes). Further work should help foster a better understanding of what makes FDI inflows move into Mexico given the uncertainty and economic and social forces.
Agarwal, J., and D. Feils. "Political Risk and the Internationalization of Firms: An Empirical Study of Canadian-Based Export and FDI Firms." Canadian Journal of Administrative Sciences, 24(3), 2007, 165-81.
Alarcon-Osuna, M. A. "Human Capital Formation and Foreign Direct Investment: Is It a Nonlinear Relationship?" Acta Universitaria, 26(4), 2016, 66-78.
Albuquerque, P. H. "Shared Legacies, Disparate Outcomes: Why American South Border Cities Turned the Tables on Crime and Their Mexican Sisters Did Not." Crime, Law and Social Change, 47(2). 2007, 69-88.
Ashby, N. J., and M. A. Ramos. "Foreign Direct Investment and Industry Response to Organized Crime: The Mexican Case." European Journal of Political Economy, 30, 2013,80-91.
Asiedu, E., and D. Lien. "Democracy, Foreign Direct Investment and Natural Resources." Journal of International Economics, 84, 2011, 99-111.
Balmori de la Miyar, J. R. "The Economic Consequences of the Mexican Drug War." Peace Economics, Peace Science and Public Policy, 22(3), 2016, 213-46.
Banco de Mexico. "Statistics and Other Indicators." 2016. Accessed April 2016. http://www.banxico.org.mx/estadisticas/index.html.
BenYishay, A. and S. Pearlman. "Homicide and Work: The Impact of Mexico's Drug War on Labor Force Participation." Australian School of Business Working Paper, 2013. Accessed July 1,2018. http://dx.doi.org/10.2139/ssrn.2302437.
--. "Crime and Microenterprise Growth: Evidence from Mexico." World Development, 56(1), 2014, 139-52.
Blanc-Brude, F., G. Cookson, J. Piesse. and R. Strange. "The FDI Location Decision: Distance and the Effects of Spatial Dependence." International Business Review, 23,2014,797-810.
Blanco, L. R. "The Impact of Crime on Trust in Institutions in Mexico." European Journal of Political Economy, 32, 2013,38-55.
Blundell, R., and S. Bond. "Initial Conditions and Moment Restrictions in Dynamic Panel Data Models." Journal of Econometrics, 87(1), 1998, 115-43.
Cabral, R., A. V. Mollick, and E. Saucedo. "Violence in Mexico and Its Effects on Labor Productivity." Annals of Regional Science, 56(2), 2016, 317-39.
Campbell, J. Y., and P. Perron. "Pitfalls and Opportunities: What Macroeconomists Should Know about Unit Roots." NBER Macroeconomics Annual, 6, 1991, 141-201.
Cardenas, M.. and S. Rozo. "Does Crime Lower Growth? Evidence from Colombia." The International Bank for Reconstruction and Development/The World Bank Working Paper No. 30, 2008.
Chanegriha, M., C. Stewart, and C. Tsoukis. "Identifying the Robust Economic, Geographical and Political Determinants of FDI: An Extreme Bounds Analysis." Empirical Economics, 52, 2017, 759-76.
Chiquiar, D. "Globalization, Regional Wage Differentials and the Stolper-Samuelson Theorem: Evidence from Mexico." Journal of International Economics, 74(1), 2008, 70-93.
Cuadros, A., V. Orts, and M. Alguacil. "Openness and Growth: Re-examining Foreign Direct Investment, Trade and Output Linkages in Latin America." Journal of Development Studies, 40(4), 2004. 167-92.
Daniele, V., and U. Marani. "Organized Crime, the Quality of Local Institutions and FDI in Italy: A Panel Data Analysis." European Journal of Political Economy, 27(1), 2011, 132-42.
Detotto, C, and E. Otranto. "Does Crime Affect Economic Growth?" Kyklos, 63(3), 2010, 330-45.
Dewenter, K. L. "Do Exchange Rate Changes Drive Foreign Direct Investment?" Journal of Business, 68(3), 1995, 405-33.
Enamorado, T, L. Lopez-Calva, and C. Rodn'guez-Castelan. "Crime and Growth Convergence: Evidence from Mexico." Economics Letters, 125, 2014, 9-13.
Estrada, M. A. R., and I. Ndoma. "How Crime Affects Economic Performance: The Case of Guatemala." Journal of Policy Modeling, 36(5), 2014, 867-82.
Feenstra, R. C, and G. H. Hanson. "Foreign Direct Investment and Relative Wages: Evidence from Mexico's Maquiladoras." Journal of International Economics, 42(3), 1997,371-93.
Fielding, D. "How Does Violent Conflict Affect Investment Location Decisions? Evidence from Israel during the Intifada." Journal of Peace Research, 41(4), 2004, 465-84.
Figini, P. "Does Foreign Direct Investment Affect Wage Inequality? An Empirical Investigation." The World Economy, 34(9), 2011, 1455-75.
Garriga. A. C, and B. Phillips. "Criminal Violence and the Subnational Distribution of FDI." CIDE Documentos de Trabajo,No. 279, 2015.
Gonzalez-Andrade, S. "Criminalidad y Crecimiento Economico Regional en Mexico." Frontera Norte, 26(51), 2014, 75-111.
Goulas, E., and A. Zervoyianni. "Economic Growth and Crime: Is There an Asymmetric Relationship?" Economic Modelling, 49, 2015, 286-95.
Kato-Vidal, E. L. "Foreign Investment and Wages: A Crowding-Out Effect in Mexico." Latin American Journal of Economics, 50(2), 2013, 209-31.
Kinda, T. "Investment Climate and FDI in Developing Countries: Firm Level Evidence." World Development, 38(4), 2010,498-513.
Kiyota, K., and S. Urata. "Exchange Rate, Exchange Rate Volatility and Foreign Direct Investment." The World Economy, 27(10), 2004, 1501 -36.
Kolstad, I., and E. Villanger. How Does Social Development Affect FDI and Domestic Investment? Bergen, Norway: Chr. Michelsen Institute, 2004.
Liu, Y, T. Fullerton Jr., and N. J. Ashby. "Assessing the Impacts of Labor Market and Deterrence Variables on Crime Rates in Mexico." Contemporary Economic Policy, 31(4), 2013, 669-90.
Luo, L. J., Y Z. Luo, and C. Liu. "Multilevel Determinants of Subsequent FDI in China." Thunderbird International Business Review, 50(2), 2008, 105-20.
Majocchi, A., and M. Presutti. "Industrial Clusters, Entrepreneurial Culture and the Social Environment: The Effects on FDI Distribution." International Business Review, 18(1), 2009, 76-88.
Marti'nez, E., and F. Jareno. "Foreign Direct Investment by Spain in Latin America: Brazil, Argentina and Mexico." Applied Econometrics and International Development, 14(2), 2014, 129-44.
Mollick, A. V, and R. Cabral. "Assessing Returns to Education and Labor Shocks in Mexican Regions after NAFTA." Contemporary Economic Policy, 33(1), 2015, 190-206.
Mollick, A. V, R. Ramos-Duran, and E. Silva-Ochoa. "Infrastructure and FDI Inflows into Mexico: A Panel Data Approach." Global Economy Journal, 6(1), 2006.
Organisation for Economic Co-operation and Development (OECD). "OCED Economic Surveys: Mexico." 2017. Accessed January 2017. http://www.oecd.org/eco/surveys/Mexico-2017-OECD-economic-survey-overview.pdf.
Pacheco-Lopez, P. "Foreign Direct Investment, Exports and Imports in Mexico." The World Economy, 28(8), 2005, 1157-72.
Pan, M., B. Widner, and C. Enomoto. "Growth and Crime in Contiguous States of Mexico." Urban & Regional Development Studies. 24(1/2), 2012, 51-64.
Ramos, M. A., and N. J. Ashby. "Heterogeneous Firm Response to Organized Crime: Evidence from FDI in Mexico." Journal of International Management, 19, 2013, 176-94.
--. "The Halo Effect: Violent Crime and Foreign Direct Investment." Multinational Business Review, 25(4), 2017,287-306.
Rodriguez-Oreggia, E., and M. Flores. "Structural Factors and the 'War on Drugs' Effects on the Upsurge in Homicides in Mexico." Harvard University, Center for International Development, CID Working Paper No. 229, January 2012.
Roodman, D. "A Note on the Theme of Too Many Instruments." Oxford Bulletin of Economics and Statistics, 71(1), 2009, 135-58.
Russ. K. N. "The Endogeneity of the Exchange Rate as a Determinant of FDI: A Model of Entry and Multinational Firms." Journal of International Economics, 71(2), 2007, 344-72.
Secretariado Ejecutivo del Sistema Nacional de Seguridad Publica. "SEGOB, Incidencia Delictiva del Fuero Comun y Federal." 2016. Accessed May 2016. http://www.secretariadoejecutivosnsp.gob.mx/en/SecretariadoEjecutivo/Incidencia_Delictiva.
Thomas, D. E., and R. Grosse. "Country-of-Origin Determinants of Foreign Direct Investment in an Emerging Market: The Case of Mexico." Journal of International Management, 7(1), 2001, 59-79.
Udomkerdmongkol, M., O. Morrissey, and H. Gorg. "Exchange Rates and Outward Foreign Direct Investment: U.S. FDI in Emerging Economies." Review of Development Economics, 13(4), 2009, 754-64.
United Nations Conference on Trade and Development. "World Investment Report 2016: Investor Nationality: Policy Challenges." 2017. Accessed July 1, 2018. http://unctad.org/en/PublicationsLibrary/wir2016_en.pdf.
Vazquez-Galan, B. I., and O. S. Oladipo. "Have Liberalisation and NAFTA Had a Positive Impact on Mexico's Output Growth?" Journal of Applied Economics, 12(1), 2009, 159-80.
Whelan, R. "Gloom Descends on Free-Trade Capital." Wall Street Journal, January 27, 2017.
Xaypanya, P., P. Rangkakulnuwat, and S. W. Paweenawat. "The Determinants of Foreign Direct Investment in Asia: The First Differencing Panel Data Analysis." International Journal of Social Economics, 42(3), 2015, 239-50.
RENE CABRAL, ANDRE VARELLA MOLLICK and EDUARDO SAUCEDO (*)
(*) An earlier version of this paper was presented at the 92nd Annual Conference of the Western Economic Association International (WEAI) in San Diego, California, at the VII Congreso de Investigacion Financiera FIMEF in Mexico City. and at an internal seminar in Banco de Mexico, also in Mexico City. The authors wish to thank Maria Esther Caballero for her comments during the WEAI Conference and also Brian Cadena and Leonardo Torre for their very detailed comments that helped improve this paper. We also acknowledge two anonymous referees of this journal for helpful comments that improved exposition and readability. We thank Eva Gonzalez for her assistance in preparing the political maps of Mexican regions. The usual disclaimer applies. This paper was partially written while the first author was collaborating with Banco de Mexico. [Correction added on 20 July 2018. after first online publication: Acknowledgment of the first author's collaboration with Banco de Mexico has been added.]
Cabral: Associate Professor of Economics, Department of Economics and Finance, EGADE Business School Tecnologico de Monterrey, Garza Garcia, CP 66269, Mexico. Phone +52-81-8625-6148, Fax +52-81-8625-6095, E-mail email@example.com
Mollick: Professor of Economics, Department of Economics and Finance, University of Texas Rio Grande Valley, Edinburg, TX 78539-2999. Phone 956-665-7136, Fax 956-665-5020, E-mail firstname.lastname@example.org
Saucedo: Assistant Professor of Economics, Department of Economics and Finance, EGADE Business School Tecnologico de Monterrey, Garza Garcfa, CP 66269. Mexico. Phone +52-81-8625-6172, Fax +52-81-8625-6095. E-mail email@example.com
CPI: Consumer Price Index
FDI: Foreign Direct Investment
GDP: Gross Domestic Product
GMM: Generalized Method of Moments
NAFTA: North American Free Trade Agreement
RER: Real Exchange Rate
RHS: Right-Hand Side
SGMM: System Generalized Method of Moments
THE: Tasa de Interes Interbancaria de Equilibrio
(1.) Conceptually different from these studies that attempt to identify the effects of crime on the Mexican economy, is the one by Liu, Fullerton, and Ashby (2013), which explains the effect of crime rates of Mexican states on economic factors, such as wages and unemployment rates. Furthermore, Blanco (2013) used surveys to estimate satisfaction with democracy and trust in institutions as functions of independent variables, including illegal trade activity.
(2.) In this paper, we refer to one of the categories as thefts, which in the English language refer to nonviolent crimes, while robbery refers to a violent crime. However, the original dataset for thefts in Spanish includes both violent and nonviolent crimes.
(3.) The total number of FDI Hows with nonpositive values in the entire sample is 104. which is only 7.4% of the total number observations. This is the number of observations not included in the regressions estimated in this paper.
(4.) Mexico's FDI figures used to show bias from the fiscal registration of FDI flows. However, the figures employed here have recently been revised by the source to eliminate such bias.
(5.) We omit the correlation figures for education and GDP per capita, which were found to be highly correlated with real wages. This way, our real wage variable carries both supply considerations (labor productivity and human capital) and a demand side variable that affects consumption, and thus, GDP. There is also some mild correlation between wages and interest rates (0.61) and between wages and real exchange rates (-0.48). The correlation between the domestic interest rate and the real value of the Mexican peso is -0.42: a weaker peso is associated with higher interest rates, possibly because of the interest rate defense when the local currency depreciates suddenly.
(6.) A recent paper by Chanegriha, Stewart, and Tsoukis (2017), which used panel data for 168 countries from 1970 to 2006, considers 58 potential variables and finds about one third to be robust. They also identify three factors that are endogenously determined with FDI: current account balance to GDP, GDP growth, and per capita GDP. Although we collect real GDP per capita figures for our dataset, we use real wages as the main demand variable, which may also indicate labor productivity (human capital) and the cost of labor.
(7.) The reason for using lags is that criminal activities show some trend, seasonality, and cyclically over time. This means foreign investors assess crime patterns rather than just the prevalence of crime at the time of making investment decisions. Henceforth, crime figures today will be taken to affect not the current FDI but rather FDI projections in the future.
(8.) As mentioned earlier, all tables presented in this section are estimated in logs, which do not take into consideration negative FDI outflows. Nevertheless, tables in this section have also been estimated using the transformation proposed by Ashby and Ramos (2013), which includes negative and positive FDI values. Results obtained under this approach are consistent but, in general, are harder to interpret. Lastly, to keep negative FDI values in the analysis, tables in this section are also estimated in levels. Results in this case are not statistically significant and are difficult to interpret. Hence, tables in this section are reported only under the log specification. Alternative tables under the transformation proposed by Ashby and Ramos (2013) and tables with regression in levels are available upon request from the authors.
(9.) When we include energy expenditures as endogenous, the number of instruments becomes 17. The results are very similar and the diagnostic tests remain satisfactory.
(10.) A recent article in the Wall Street Journal entitled "Gloom descends on free-trade capital" examined the uncertainty affecting the industrial city of Monterrey in the northern state of Nuevo Leon: "A flood of NAFTA-inspired foreign investment has helped turn Monterrey into Mexico's Free Trade capital, lifting tens of thousands of workers into the middle class and making this city's mighty industrialist families even richer. These days, people at both ends of the economy are scrambling to figure out what to do now that the new U.S. president, Donald Trump, appears to be following through on vows to renegotiate NAFTA and build a wall between the two countries." (Whelan 2017).
TABLE 1 Descriptive Statistics Variables Obs. M SD Real FDI (millions) 1,408 201 359 Positive flows of real FDI (millions) 1,304 220 364 Real wage 1,408 16.43 3.64 Electricity sales (thousands) 1,408 1,500 1,126 Interest rate 1,408 5.87 2.04 Real exchange rate 1,408 93.40 6.21 Crime rates (per 100,000 people) Homicide 1,408 7.30 4.54 Thefts 1,408 142.48 97.46 Property 1,408 58.96 36.48 Other crimes 1,408 155.23 100.88 Descriptive Statistics Variables Min Max Real FDI (millions) -1,290 3,800 Positive flows of real FDI (millions) 0.25 3,800 Real wage 9.1 32.0 Electricity sales (thousands) 184 5,194 Interest rate 3.3 10.1 Real exchange rate 81.4 107.4 Crime rates (per 100,000 people) Homicide 0.6 34.0 Thefts 10.1 659.4 Property 0 220.5 Other crimes 2.9 560.5 Notes: FDI and wages are given in real dollars of 2010. Electricity rates are given in mega-watts per hour. Crime rates are per 100,000 people. TABLE 2 Average State Indicators Real FDI Real Wage State (USD millions) (USD) Homicide Theft Aguascalientes 77 15.9 4.66 175.57 Baja California 294 18.1 6.15 485.96 Baja California Sur 140 17.0 3.78 320.89 Campeche 30 21.2 3.84 17.72 Chiapas 28 13.5 7.75 37.68 Chihuahua 439 16.7 15.55 211.82 Coahuila 177 17.1 5.79 154.14 Colima 29 15.5 6.47 136.15 Ciudad de Mexico 1,310 24.8 4.25 245.69 Durango 57 13.4 11.20 130.91 Estado de Mexico 612 18.4 5.69 152.94 Guanajuato 244 15.1 6.69 132.73 Guerrero 70 15.1 14.55 69.90 Hidalgo 45 15.3 4.71 92.38 Jalisco 348 16.9 5.20 101.85 Michoacan 124 15.5 10.95 95.63 Morelos 62 17.9 12.69 242.12 Nayarit 45 14.3 8.32 68.45 Nuevo Leon 621 20.8 5.96 152.87 Oaxaca 80 14.6 7.39 66.41 Puebla 155 16.9 5.01 112.44 Queretaro 189 20.7 4.51 151.49 Quintana Roo 112 14.7 9.73 231.71 San Luis Potosi 153 16.1 4.33 91.72 Sinaloa 61 13.5 16.44 115.36 Sonora 212 15.2 8.77 111.04 Tabasco 47 15.9 5.45 158.05 Tamaulipas 208 17.1 8.10 166.74 Tlaxcala 32 15.0 8.16 66.58 Veracruz 196 15.5 4.63 68.17 Yucatan 37 13.4 2.47 86.41 Zacatecas 188 14.8 4.27 107.93 Total 201 16.4 7.30 142.48 Other Electricity Sales State Property Crimes (Thousands of MW/hou Aguascalientes 83.35 148.90 582 Baja California 106.29 290.37 2,330 Baja California Sur 139.63 275.89 457 Campeche 6.48 24.85 279 Chiapas 22.21 62.13 640 Chihuahua 73.52 155.09 2,507 Coahuila 80.21 144.75 2,379 Colima 42.24 181.95 393 Ciudad de Mexico 80.58 156.32 3,470 Durango 49.56 113.85 705 Estado de Mexico 37.67 208.64 4,133 Guanajuato 75.27 159.71 2,382 Guerrero 29.75 109.41 683 Hidalgo 50.67 160.52 837 Jalisco 58.73 110.94 2,870 Michoacan 29.91 56.68 1,776 Morelos 98.35 273.68 602 Nayarit 43.19 101.34 314 Nuevo Leon 32.31 58.47 3,997 Oaxaca 52.87 138.54 597 Puebla 56.22 114.84 1,790 Queretaro 48.91 98.92 1,024 Quintana Roo 125.01 217.72 901 San Luis Potosi 76.12 157.33 1,314 Sinaloa 19.75 102.50 1,386 Sonora 49.87 147.20 2,447 Tabasco 57.72 388.92 740 Tamaulipas 32.14 157.46 2,113 Tlaxcala 20.39 68.28 445 Veracruz 47.77 91.90 2,570 Yucatan 112.71 415.03 742 Zacatecas 47.17 75.10 580 Total 58.96 155.23 1,500 TABLE 3 Correlation Matrix Real Real Variables FDI Wage Homicides Thefts Property Real FDI 1 Real wage 0.4478 1 Homicide -0.0982 -0.2722 1 Theft 0.2223 0.199 0.0961 1 Properly 0.0667 0.1408 -0.1578 0.6113 1 Other crimes -0.0343 -0.0352 -0.1044 0.4686 0.663 Interest rate 0.0889 0.6119 -0.1492 -0.0469 0.1572 Real exchange rate -0.1046 -0.4771 0.1693 0.0759 -0.0602 Electricity sales 0.5127 0.2621 -0.0326 0.2092 -0.0348 Other Interest Real Electricity Variables Crimes Rate Exchange Rate Sales Real FDI Real wage Homicide Theft Properly Other crimes 1 Interest rate 0.0261 1 Real exchange rate -0.0221 -0.4204 1 Electricity sales -0.0606 -0.0928 0.0329 1 TABLE 4 Fixed-Effects Panel Data (FEM) by Different Crimes with Fixed Lag-Length Approach (1 and 4 Lags) Base Model (No Crime) Homicides 1 Lag 4 Lags 1 Lag 4 Lags L1. Crime 0.07 -0.02 (0.09) (0.15) L2. Crime 0.11 (0.2) L3. Crime -0.23 (0.16) L4. Crime 0.18 (0.13) L1. Wage 0.42 -1.36 0.39 -1.49 (0.45) (1.15) (0.46) (1.2) L2. Wage 0.05 0.3(1.7) (1.65) L3. Wage 1.8 1.48 (1.74) (1.8) L4. Wage 0.24 0.46 (1.57) (1.61) L1. Interest Rate -0.06 1.52 (**) -0.05 1.55 (**) (0.18) (0.66) (0.19) (0.66) L2. Interest Rate -2.04 -2.08 (1.13) (1.13) L3. Interest Rate 0.03 0.01 (0.98) (0.97) L4. Interest Rate 0.68 0.71 (0.51) (0.51) L1. Electricity -0.28 -0.24 -0.29 -0.23 (0.23) (0.32) (0.23) (0.32) L2. Electricity 0.90 (**) 0.89 (**) (0.37) (0.37) L3. Electricity 0.15 0.17 (0.34) (0.34) L4. Electricity -0.16 -0.18 (0.28) (0.28) L1. Real Ex. Rate -0.79 -3.33 (**) -0.88 -3.49 (**) (0.5) (1.55) (0.51) (1.61) L2. Real Ex. Rate 0.95 1.2 (1.87) (1.85) L3. Real Ex. Rate 0.91 0.61 (1.86) (1.91) L4. Real Ex. Rate 0.75 0.93 (1.69) (1.74) Dummy crisis -0.13 -0.12 -0.13 -0.11 (0.08) (0.11) (0.08) (0.11) Sample size 1,272 1,181 1,272 1,181 [R.sup.2] within 0.02 0.07 0.02 0.07 [R.sup.2] overall 0.01 0.06 0.01 0.06 Thefts Property 1 Lag 4 Lags 1 Lag 4 Lags L1. Crime 0.13 0.03 -0.02 -0.18 (0.13) (0.21) (0.1) (0.12) L2. Crime 0.18 0.04 (0.29) (0.15) L3. Crime -0.34 -0.03 (0.36) (0.17) L4. Crime 0.19 0.21 (0.3) (0.18) L1. Wage 0.25 -1.24 0.41 -1.31 (0.49) (1.2) (0.46) (1.17) L2. Wage -0.24 -0.01 (1.76) (1.64) L3. Wage 2.02 1.82 (1.82) (1.68) L4. Wage 0.05 0.05 (1.52) (1.53) L1. Interest Rate 0 1.52 (**) -0.04 1.45 (**) (0.21) (0.66) (0.19) (0.67) L2. Interest Rate -1.98 -1.89 (1.11) (1.11) L3. Interest Rate -0.02 0.02 (0.97) (0.98) L4. Interest Rate 0.73 0.67 (0.53) (0.5) L1. Electricity -0.28 -0.26 -0.28 -0.25 (0.22) (0.32) (0.23) (0.32) L2. Electricity 0.94 (**) 0.90 (**) (0.38) (0.39) L3. Electricity 0.13 0.12 (0.35) (0.34) L4. Electricity -0.12 -0.15 (0.29) (0.28) L1. Real Ex. Rate -1.04 -3.22 -0.79 -3.31 (**) (0.58) (1.6) (0.52) (1.58) L2. Real Ex. Rate 0.64 0.9 (1.98) (1.86) L3. Real Ex. Rate 1.1 0.96 (1.91) (1.82) L4. Real Ex. Rate 0.57 0.66 (1.63) (1.71) Dummy crisis -0.13 -0.12 -0.13 -0.14 (0.08) (0.11) (0.08) (0.11) Sample size 1,272 1,181 1,269 1,178 [R.sup.2] within 0.02 0.07 0.02 0.07 [R.sup.2] overall 0.02 0.06 0.01 0.06 Other Crimes 1 Lag 4 Lags L1. Crime -0.02 -0.35 (0.11) (0.21) L2. Crime 0.07 (0.15) L3. Crime 0.1 (0.23) L4. Crime 0.1 (0.15) L1. Wage 0.42 -1.17 (0.45) (1.18) L2. Wage 0.06 (1.64) L3. Wage 1.52 (1.75) L4. Wage 0.19 (1.58) L1. Interest Rate -0.06 1.51 (**) (0.19) (0.66) L2. Interest Rate -2 (1.13) L3. Interest Rate 0.17 (0.97) L4. Interest Rate 0.58 (0.5) L1. Electricity -0.28 -0.21 (0.22) (0.32) L2. Electricity 0.87 (**) (0.4) L3. Electricity 0.13 (0.33) L4. Electricity -0.11 (0.28) L1. Real Ex. Rate -0.8 -3.17 (0.49) (1.6) L2. Real Ex. Rate 0.88 (1.86) L3. Real Ex. Rate 0.72 (1.86) L4. Real Ex. Rate 0.73 (1.73) Dummy crisis -0.13 -0.14 (0.08) (0.11) Sample size 1,272 1,181 [R.sup.2] within 0.02 0.08 [R.sup.2] overall 0.01 0.06 Notes: Newey-West robust standard errors to heteroscedasticity and serial correlation are reported in parentheses. The symbols (**) and (***) refer to levels of significance of 5% and 1 %, respectively. TABLE 5 Fixed-Effects Panel Data (FEM) by Different Crimes with Flexible Lag-Length Approach Base Model (No Crime) Homicides Thefts Variables (1) (2) (3) L1. Crime 0.05 0.11 (0.17) (0.26) L2. Crime 0.08 0.16 (0.20) (0.31) L3. Crime -0.28 -0.44 (0.18) (0.39) L4. Crime 0.30 0.25 (0.15) (0.29) L5. Crime -0.28 (**) 0.17 (0.12) (0.18) L6. Crime -0.33 (**) (0.13) L1. Wage -0.75 -0.97 -0.95 (1.01) (1.08) (1.11) L2. Wage -2.99 -2.68 -3.14 (1.57) (1.71) (1.68) L3. Wage 4.93 (**) 4.64 (**) 5.26 (**) (1.93) (1.97) (1.98) L4. Wage .525 0.99 0.45 (1.61) (1.58) (1.59) L5. Wage -6.36 (***) -6.69 (***) -6.21 (***) (1.97) (2.01) (2.01) L6. Wage 5.95 (***) 6.08 (***) 5.96 (***) (1.61) (1.68) (1.60) L1. Interest Rate 1.27 (.77) 1.17 1.18 (0.77) (0.76) L2. Interest Rate -1.07 -0.97 -1.00 (1.18) (1.16) (1.16) L3. Interest Rate .42(1.07) 0.38 0.44 (1.07) (1.05) L4. Interest Rate -1.08 -1.02 -1.06 (1.27) (1.27) (1.27) L5. Interest Rate 2.91 (**) 2.91 (**) 2.97 (**) (1.18) (1.16) (1.16) L6. Interest Rate -2.57 (***) -2.65 (***) -2 71 (***) (.60) (0.58) (0.58) L1. Electricity -.060 -0.05 -0.08 (.27) (0.26) (0.28) L1. Real Ex. Rate -3.38 (**) -3.65 (**) -3.73 (**) (1.64) (1.67) (1.78) L2. Real Ex. Rate -1.88 -1.56 -2.01 (1.86) (1.94) (2.01) L3. Real Ex. Rate 3.80 (**) 3.52 4.13 (**) (1.82) (1.90) (1.90) L4. Real Ex. Rate 1.80 2.33 1.92 (2.02) (2.04) (1.97) L5. Real Ex. Rate -6.34 (**) -6.67 (**) -6.16 (**) (2.57) (2.57) (2.57) L6. Real Ex. Rate 5.70 (***) 5 87 (***) 5.57 (***) (1.78) (1.84) (1.79) Dummy crisis -0.23 -0.24 -0.24 (.121) (0.12) (0.12) Sample size 1,120 1,120 1,120 [R.sup.2] within .097 0.11 0.10 [R.sup.2] overall .133 0.08 0.08 Property Other Crimes (4) Variables (5) -0.05 L1. Crime (0.09) -0.23 (0.16) L2. Crime L3. Crime L4. Crime L5. Crime L6. Crime -0.69 L1. Wage (1.03) -0.72 -3.00 (1.03) L2. Wage (1.57) -2.98 4 92 (**) (1.56) L3. Wage (1.92) 4.76 (**) 0.55 (1.92) L4. Wage (1.61) 0.62 -6.38 (***) (1.60) L5. Wage (1.98) -6.39 (***) 5.92 (***) (1.99) L6. Wage (1.61) 5.77 (***) 1.25 (1.62) L1. Interest Rate (0.77) 1.24 -1.05 (0.77) L2. Interest Rate (1.18) -1.02 0.44 (1.18) L3. Interest Rate (1.07) 0.57 -1.09 (1.06) L4. Interest Rate (1.28) -1.17 2.90 (**) (1.28) L5. Interest Rate (1.18) 2.94 (**) -2.56 (***) (1.18) L6. Interest Rate (0.59) -2.58 (***) -0.06 (0.59) L1. Electricity (0.27) -0.03 -3.31 (0.28) L1. Real Ex. Rate (1.67) -3.46 (**) -1.89 (1.68) L2. Real Ex. Rate (1.86) -1.88 3 79 (**) (1.83) L3. Real Ex. Rate (1.82) 3.69 1.81 (1.82) L4. Real Ex. Rate (2.03) 1.97 -6.35 (**) (2.02) L5. Real Ex. Rate (2.58) -6.40 (**) 5.67 (***) (2.59) L6. Real Ex. Rate (1.79) 5.43 (***) -0.23 (1.82) Dummy crisis (0.12) -0.24 1,120 (0.12) Sample size 0.10 1,120 [R.sup.2] within 0.08 0.10 [R.sup.2] overall 0.08 Notes: Newey-West robust standard errors to heteroscedasticity and serial correlation are reported in parentheses. The symbols (**) and (***) refer to levels of significance of 5% and 1%, respectively. TABLE 6 Fixed-Effects Panel Data (FEM) Ranked by Crime Levels with Flexible Lag-Length Approach Homicide Theft Variables Top 16 Bottom 16 Top 16 L1. Crime 0.28 -0.08 0.41 (0.27) (0.23) (0.18) L2. Crime 0.09 0.47(0.31) (0.34) L3. Crime -0.43 (**) -0.97 (**) (0.17) (0.35) L4. Crime L5. Crime L6. Crime L1. Wages -2.08 .409 (.49) -0.05 (1.17) (1.25) L2. Wages -2.88 -4.07 (**) (2.42) (1.90) L3. Wages 5.88 (**) 6.20 (**) (2.68) (2.54) L4. Wages -0.52 0.10(1.80) (1.73) L5. Wages -8.75 (***) -7.82 (**) (2.71) (3.10) L6. Wages 7.54 (***) (2.37) (2.36) L1. Interest Rate 1.60 2.18 (***) 0.48(1.01) (1.33) (0.51) L2. Interest Rate -1.39 -1.60 0.02(1.59) (1.81) (1.18) L3. Interest Rate 1.64 -1.51 1.24(1.89) (1.58) (1.58) L4. Interest Rate -1.17 1.01 -2.94 (1.55) (2.09) (1.88) L5. Interest Rate 3.01 (**) 1.87 3.65(1.73) (1.25) (2.11) L6. Interest Rale -2.67 (***) -2.32 (**) -2.57 (***) (0.82) (1.01) (0.78) L1. Electricity 0.20 -0.79 0.07 (0.37) (0.32) (0.58) L2. Electricity L3. Electricity L4. Electricity L5. Electricity L6. Electricity L1. Real Ex Rate -6.29 (***) -0.55 -1.23 (1.67) (0.60) (1.53) L2. Real Ex Rate -0.17 -4.06 (2.61) (2.19) L3. Real Ex Rate 4.33 5.25 (**) (2.60) (2.42) L4. Real Ex Rate 1.19 1.73(2.67) (1.55) L5. Real Ex Rate -9.16 (**) -7.74 (3.12) (4.21) L6. Real Ex Rate 8.65 (***) 6.50 (**) (2.45) (2.77) Dummy crisis -0.36 -0.18 -0.22 (0.24) (0.09) (0.16) Sample size 564 556 570 [R.sup.2] within 0.16 0.07 0.14 [R.sup.2] overall 0.12 0.05 0.11 Variables Bottom 16 L1. Crime -0.09 (0.32) L2. Crime 0.07 (0.47 L3. Crime -0.33 (0.60) L4. Crime 0.61 (0.33) L5. Crime 0.18 (0.23) L6. Crime -0.30 (**) (0.12) L1. Wages -1.08 (0.84) L2. Wages L3. Wages L4. Wages L5. Wages L6. Wages L1. Interest Rate 2.01 (**) (0.69) L2. Interest Rate -1.71 (1.11) L3. Interest Rate -0.89 (1.24) L4. Interest Rate 1.28(1.58) L5. Interest Rate 2.14(1.61) L6. Interest Rale -2.29 (**) (0.94) L1. Electricity -0.32 (1.00) L2. Electricity 0.40(0.67) L3. Electricity -0.66 (0.43) L4. Electricity 0.50(0.37) L5. Electricity -1.11 (0.92) L6. Electricity 1.89 (**) (0.76) L1. Real Ex Rate -3.32 (**) (1.18) L2. Real Ex Rate L3. Real Ex Rate L4. Real Ex Rate L5. Real Ex Rate L6. Real Ex Rate Dummy crisis -0.35 (**) (0.16) Sample size 550 [R.sup.2] within 0.13 [R.sup.2] overall 0.09 Notes: Newey-West robust standard errors to heteroscedasticity and serial correlation are reported in parentheses. The symbols (**) and (***) refer to levels of significance of 5% and 1%, respectively. TABLE 7 SGMM Results for FDI in Mexican States: Dependent Variable Is Total Log FDI in USD Full Sample No Crime Homicide Theft Coefficients on (1) (2) (3) Lagged FDI 0.61 (**) 0.53 (**) 0.55 (**) (0.25) (0.28) (0.24) Crime -0.26 0.34 (0.84) (0.37) Real wage -1.38 -1.18 -1.65 (0.97) (1.01) (1.05) THE 1.04 1.07(0.73) 1.02 (0.67) (0.63) Energy expenditures -1.11 -1.09 -1.13 (1.04) (1.04) (0.80) RER 2.37 3.36 (6.27) 0.41 (2.25) (2.75) 2008-2009 dummy crisis -0.58 (**) -0.69 -0.45 (0.27) (0.55) (0.27) Lags of end. vars./ [2,3] [2,3] [2,3] collapse Observations (W) 1,187 1,187 1,187 States 32 32 32 No. of instruments 14 14 14 AB(2) 0.60 0.34 0.55 p value [-55] [.73] [-58] Hansen 9.05 8.71 5.63 p value [.25] [-19] [.47] Property Other Crimes Coefficients on (4) (5) Lagged FDI 0.40 (0.32) 0.68 (**) (0.34) Crime -1.71 -2.29 (1.05) (1.19) Real wage 0.45 (1.75) -2.14 (**) (0.98) THE 0.73 (0.87) 1.56 (0.84) Energy expenditures -2.66 -2.12 (**) (1.63) (0.84) RER 4.39 (4.60) 5.44 (4.24) 2008-2009 dummy crisis -0.80 -0.96 (**) (0.52) (0.40) Lags of end. vars./ [2,3] [2,3] collapse Observations (W) 1,184 1,187 States 32 32 No. of instruments 14 14 AB(2) -0.11 0.08 p value [.91] [.94] Hansen 6.05 3.90 p value [.42] [.69] Notes: SGMM basic specification has lagged log FDI once and all (RHS) are contemporaneously also in logs. A constant is included but is not reported in the table. Endogenous variables are FDI and labor and macroeconomic economic forces (real wages, interest rate, and RERs) with two lags and collapsed procedure to achieve number of instruments < cross-section units. Crisis dummy is assumed as purely exogenous. Robust standard errors (with Windmeijer's finite-sample correction) are in parentheses with two-step estimators. With orthogonal transform, instruments are replaced with their deviations from past means. AB(2) refers to the Arellano-Bond test of second order serial correlation. Regressors FDI, RER, Real Wage, and THE are considered as endogenous.
|Printer friendly Cite/link Email Feedback|
|Author:||Cabral, Rene; Mollick, Andre Varella; Saucedo, Eduardo|
|Publication:||Contemporary Economic Policy|
|Date:||Jan 1, 2019|
|Previous Article:||GROWTH IN THE SHADOWS: EFFECT OF THE SHADOW ECONOMY ON U.S. ECONOMIC GROWTH OVER MORE THAN A CENTURY.|
|Next Article:||DYNAMIC ENTRY WITH DEMAND AND SUPPLY SIDE SPILLOVERS.|