*3.4. Controls*

We follow the literature in our choice of benchmark controls, which consist of GDP per capita growth, urban population as a percentage of total population, labor force participation rates, trade openness and polity as a measure of political institutions. Studies like Huang and Temple (2005) and Svaleryd and Vlachos (2002), for instance, show that trade openness is a significant determinant of financial development. Based on both demand and supply side arguments, studies like Jung (1986), Goldsmith (1969), Gurley and Shaw (1967) and Patrick (1966) have stressed a causal relationship from economic growth to financial development. For this reason, we include GDP per capita growth as a measure of economic development within our benchmark controls. As a further measure of economic development (based on the demand side argument), we control for urban population as a percentage of total population. Lastly, since multiple studies have shown that democratic institutions are an important determinant for financial development, we consider the Polity 2 variable, which runs from −10 to +10 with higher values implying more democratic institutions (Begovi´c et al. 2017; Bhattacharyya 2013; Yang 2011; Huang 2010; Clague et al. 1996).

#### **4. Empirical Methodology and Benchmark Results**

#### *4.1. Empirical Specification and Methodology*

Based on our hypothesis, we test the following regression specification:

$$\text{FD}\_{it} = \beta\_0 + \beta\_1 FTR\_i + controls + \mathcal{Q}\_t + \epsilon\_{it}$$

FD*it* represents the specific financial development measure considered for country *i* in time *t*. Our main independent variable of interest is FTR, which represents the future time reference dummy. We remind our readers that FTR dummy takes 1 for weak FTR countries, and 0 for strong FTR countries. According to our hypothesis, we expect *β*<sup>1</sup> to be positive and significant. This would imply that financial development is higher in weak FTR countries (i.e., those with languages that do not require a dedicated form when referring to future events) relative to strong FTR countries. *β*<sup>1</sup> being negative and significant would mean the opposite.

Our benchmark measure of financial development is private credit to the real sector by deposit money banks as a percentage of GDP. As an alternate benchmark measure of financial development, we consider private credit to the real sector by deposit money banks and other financial institutions as a percentage of GDP. Our robustness analysis considers alternate measures of financial development.

Since linguistic features and financial development are likely the product of deeper, unobserved factors, we follow the literature in constructing our battery of controls, including observables that can help explain differences in financial development across countries. These, in turn, are unlikely to make the effect of FTR on financial development exogenous and uncorrelated with the error term. Controlling for observables that can help explain

differences in financial development across countries is the first step to mitigate bias arising out of omitted variable bias.<sup>8</sup> As explained, our benchmark controls are GDP per capita growth, urban population as a percentage of total population, labor force participation rates, trade openness and polity as a measure of political institutions. These variables have been shown to matter for financial development. In the robustness section, we discuss further controlling for additional variables as well as mitigating the effect of unobserved heterogeneity with respect to the effects of FTR on financial development.

Our benchmark analysis consists of ordinary least squares (OLS) regressions. ∅*t*, our time fixed effects, help us take into account time shocks. For example, global shocks, like the 2009 recession, that likely impact financial development should be captured in time fixed effects. In addition to our OLS specifications, we consider quantile regressions to make sure our results are not driven by the presence of outliers. We talk about endogeneity and how our findings should be interpreted in subsequent sections.

#### *4.2. Benchmark Results*

In Table 1, we present our first set of benchmark results with OLS regressions. The dependent variable considered is private credit to GDP. In column (1), we run a bivariate regression without any controls to assess the variation in financial development that is attributed to weak and strong FTR languages. Based on column (1), we find that, relative to strong FTR countries, weak FTR countries have 41 percent more private credit (as a percentage of GDP). Providing an example, this suggests that when compared against strong FTR countries (such as India), weak FTR countries (such as Indonesia) should have much more private credit. However, the coefficient of the FTR dummy in column (1) is capturing effects of other variables that also affect financial development. We add controls in subsequent columns. In column (2), we add labor force participation rate, and in column (3), we control for urban population as a percentage of total population. In column (4), GDP per capita growth is included, and in column (5), we control for trade as a percentage of GDP. Finally in column (6), we add the polity score, which is a measure of how relatively democratic a nation is in terms of its governance.

**Table 1. Private Credit and FTR.** OLS regressions with time fixed effects: The dependent variable is private credit by deposit money banks as a percentage of GDP. FTR is future time reference dummy with 1 indicating weak FTR countries and strong FTR countries. The controls for labor force participation rate (LFPR), urban population as a percentage of total population, GDP per capita growth, trade as percentage of GDP and polity. Robust standard errors in parentheses \*\*\* *p* < 0.01, \*\* *p* < 0.05.



**Table 1.** *Cont.*

Once we include all of the aforementioned controls, the effect of FTR on private credit drops to 23 percentage points. This implies that weak FTR countries have 23 percentage points more private credit than strong FTR countries. Our controls are predominantly significant and are of expected sign and significance.9

To account for potential outliers driving our results, we replicate the specifications from Table 1 in Table 2 by using quantile regressions. The results are very similar. The magnitude of FTR dummy is marginally higher compared to Table 1.

**Table 2. Private Credit and FTR.** Quantile regressions with time fixed effects: The dependent variable is private credit by deposit money banks as a percentage of GDP. FTR is future time reference dummy with 1 indicating weak FTR countries and strong FTR countries. The controls for labor force participation rate (LFPR), urban population as a percentage of total population, GDP per capita growth, trade as percentage of GDP and polity. Robust standard errors in parentheses \*\*\* *p* < 0.01, \*\* *p* < 0.05.


For example, in column (1), when we do not include any control variables, the magnitude of difference between weak FTR and strong FTR countries in terms of private credit is 46 percentage points. Once we control for all the variables, it drops to 26 percentage points. The sign and significance of the control variables remain similar to our previous table.

In Table 3, we consider an alternate measure of financial development—private credit to the real sector by deposit money banks and other financial institutions as a percentage of GDP. As mentioned earlier, this is a broader measure of financial depth. We consider both OLS and quantile regressions including all controls. Column (1) in Table 3 presents OLS regression, while column (2) presents quantile regressions. Here, we find that the impact of the FTR dummy for the OLS regression is stronger than in the case of quantile regression. While weak FTR countries have 21 percentage points more private credit (including financial institutions) relative to strong FTR countries in the case of OLS regression, the effect drops to 15 percentage points in the case of quantile regression. Labor force participation rate (LFPR) and urban population are positive and significant in both regressions, while GDP per capita growth is negative and significant in both cases. Trade is not significant in Table 3 specifications.

**Table 3. Private Credit (including Financial Institutions) and FTR.** OLS and Quantile regressions with time fixed effects: The dependent variable is private credit by deposit money banks and other financial institutions as a percentage of GDP. FTR is future time reference dummy with 1 indicating weak FTR countries and strong FTR countries. The controls for labor force participation rate (LFPR), urban population as a percentage of total population, GDP per capita growth, trade as percentage of GDP and polity. Robust standard errors in parentheses \*\*\* *p* < 0.01, \*\* *p* < 0.05, \* *p* < 0.1.


#### **5. Robustness Analysis**

For our robustness analysis, we conduct an array of tests to make sure our results are not sensitive additional controls, alternate fixed effects, or other measures of our linguistic variable (FTR). We start our robustness analysis by controlling for continent fixed effects within our benchmark specifications, the importance of which is emphasized by Ang (2019). Additionally, we consider regional dummies based on the country income classification by the World Bank, as is commonly used in the literature. As Mavisakalyan et al. (2018) point out, because linguistic features can be spatially correlated, this implies that linguistic features can be concentrated in certain areas. Thus, the effect of FTR can be biased due to geographic and climatic factors that are correlated. We present the results in Table 4 with both measures of financial development considered in our benchmark analysis.

Our main conclusions remain unchanged. The impact for private credit as evident from specification (1) is around 12 percentage points, which again implies that weak FTR countries have more private credit compared to strong FTR countries. As anticipated, controlling for continent fixed effects does reduce the magnitude of FTR dummy relative to previous specifications.

We continue our robustness analysis by controlling for additional variables to further mitigate omitted variable bias. Following the extensive literature on the determinants of financial development, we include different measures of human capital in Table 4 to bolster our benchmark set of controls. As Ibrahim and Sare (2018) found, human capital has a robust influence on financial development, thus creating greater demand for financial intermediation and services that constitute the process of financial development.10 In Table 4 column (1), we consider a measure of human capital—net primary enrollment. As an alternate measure of human capital, secondary (net) enrollment in considered in column (2).

**Table 4. Private Credit and FTR—Including additional controls.** OLS regressions with time fixed effects: The dependent variable is private credit by deposit money banks as a percentage of GDP. FTR is future time reference dummy with 1 indicating weak FTR countries and strong FTR countries. The benchmark controls for labor force participation rate (LFPR), urban population as a percentage of total population, GDP per capita growth, trade as percentage of GDP and polity. The additional controls are school enrollment (primary), school enrollment (secondary), foreign direct investment inflows, constraints on the chief executive, durability (democracy) and inflation. Robust standard errors in parentheses \*\*\* *p* < 0.01, \*\* *p* < 0.05.


Likewise, political institutions have been shown to be an important determinant of financial development in the literature. As Pagano and Volpin (2001) point out, selfinterested policy makers can intervene in financial markets for promotion of group interests. Rajan and Zingales (2003) also emphasize the role that interest groups can play in financial development. As Huang (2010) argues, the presence of a stronger elite group favors the interests of elites and restricts democratic participation. Greater shift of power towards elite groups potentially makes the system more autocratic and results in greater obstacles for financial development. In this context, Girma and Shortland (2008) have shown that both democracy and regime change matters for financial development.

With these in mind, we check the sensitivity of our findings with alternate measures of political institutions (other than polity, which we used in our benchmark regressions). The first variable we use is constraints on the chief executive. Based on the data and definition provided by Marshall et al. (2019), the variable, "refers to the extent of institutionalized constraints on the decision-making powers of chief executives, whether individuals or collectivities". It ranges from 1 to 7, with higher numbers denoting greater constraints as measured by the ability in which "accountability groups" may impose limitations. For example, in Western democracies, these typically take the form of legislatures. Column (3) of Table 4 considers the constraint measure instead of polity.

As an alternate measure, we consider the durability of the political system. Based on the definition set forth by Marshall et al. (2019), it is measured as the number of years "since the last substantive change in authority characteristics (defined as a 3-point change in the POLITY score)". We consider this measure in column (4). Finally, in column (5) of the table, we consider inflation as an additional control variable.

For all the specifications in Table 4, we consider private credit by deposit money banks (excluding other financial institutions) as the dependent variable. As we can see from the table, the coefficient of FTR is positive and significant for all specifications. In terms of magnitude, for weak FTR countries, weak FTR countries have between 11 and 15 percentage points more private credit. These findings, given our additional controls, suggest that greater constraints and a more durable political system enhance financial development. Likewise, the effects of both education measures are positive and significant as well.

Next, we next consider alternate measures of financial development as dependent variables. We present these results in Table 5. The first alternate measure we consider is liquid liabilities as a percentage of GDP. The measure, used by King and Levine (1993), is the broadest indicator of financial intermediation as it encompasses currency as well as interest bearing liabilities of banks and other financial intermediaries. One other alternate measure considered is bank credit as a percentage of bank deposits. As Beck et al. (2009) state, this measure indicates the ratio of claims on the private sector to deposits in money banks. It is a measure of the efficiency of the financial system as it assesses the extent to which "banks intermediate society's savings into private sector credits". The other two measures considered are equity market indicators, including stock market capitalization and stock market total value, both as a percentage of GDP. The size of the equity markets relative to the size of the economy are captured by these two indicators. The first measure, stock market capitalization to GDP, assessing activity of the stock market equals total shares traded on the stock market as a percentage of GDP. The second measure, stock market total value to GDP, also measures activity of the stock market but in terms of trading volume as a share of national output.

In Table 5, we find that the FTR dummy is positive and significant across all measures. The effect of the FTR dummy is the strongest in the case of the liquid liabilities measure, while the magnitude is the least in the case of stock market capitalization measure. In the case of liquid liabilities, weak FTR countries have 34 percentage points more liquid liabilities relative to strong FTR countries. In the case of bank deposits, the magnitude is about 23 percentage points.

**Table 5. Alternate Measures of Financial Development and FTR.** OLS regressions with time fixed effects: The dependent variables are liquid liabilities (% of GDP), stock market capitalization (% of GDP), stock market total value traded (% of GDP) and bank deposits as percentage of bank credits in columns (1), (2), (3) and (4), respectively. FTR is future time reference dummy with 1 indicating weak FTR countries and strong FTR countries. The benchmark controls for labor force participation rate (LFPR), urban population as a percentage of total population, GDP per capita growth, trade as percentage of GDP and polity. Robust standard errors in parentheses \*\*\* *p* < 0.01, \*\* *p* < 0.05, \* *p* < 0.1.


For the final part of our robustness analysis, we investigate if strong or weak FTR matters differently for high or low levels of financial development. In other words, do countries speaking weak FTR languages benefit more if they have *lower* levels of financial development relative to countries speaking weak FTR countries but have *higher* levels of financial development? We run quantile regressions for the 25th, 50th and 75th percentiles of financial development based on our sample. In Table 6, we consider our benchmark measure of financial development—private credit. We also consider the other broader measure of private credit—private credit to the real sector by deposit money banks and other financial institutions as a percentage of GDP. In column (1), we consider the private credit measure, and in column (2), we consider the private credit (plus financial institutions) measure. We report the results for our main variable of interest, FTR dummy, for the 25th, 50th and 75th percentiles of financial development.<sup>11</sup>

We find that the FTR dummy has the strongest effect for both measures of financial development for the 25th percentile. For both measures, weak FTR language speaking countries with financial development in the 25th percentile have about 26–27% more private credit or private credit (plus financial institutions) relative to strong FTR language countries in the same financial development percentile. For countries in the highest percentile (75th percentile) of financial development, weak FTR language speaking countries also benefit more than strong FTR language speaking countries but by a lesser magnitude. Thus, across all specifications, we observe that weak FTR languages are associated with enhanced financial development relative to strong FTR countries.

**Table 6. Financial Development Percentiles and Private Credit.** OLS Regressions with time fixed effects: The dependent variables are liquid liabilities (% of GDP), stock market capitalization (% of GDP), stock market total value traded (% of GDP) and bank deposits as percentage of bank credits in columns (1), (2), (3) and (4), respectively. FTR is future time reference dummy with 1 indicating weak FTR countries and strong FTR countries. The benchmark controls for labor force participation rate (LFPR), urban population as a percentage of total population, GDP per capita growth, trade as percentage of GDP and polity. Robust standard errors in parentheses \*\*\* *p* < 0.01, \*\* *p* < 0.05.


#### **6. Conclusions**

Given the implications of financial development for economic growth and varied development outcomes (Ibrahim and Alagidede 2017b; Mishra and Narayan 2015; Masten et al. 2008; Rioja and Valev 2004; Calderon and Liu 2003), the factors that shape a country's economic development remains an important research question to consider. Our results add to the literature on the determinants of financial development by finding that linguistic structures of countries play an important role in affecting financial development. Specifically, our results show that countries speaking weak future time reference (FTR) language experience enhanced financial development relative to countries speaking strong future time reference (FTR) languages. In light of this, weak FTR languages discount the future relatively less and maintain the connection between present and the future. Due to this, individual speakers of these languages are more likely to work towards creating and bettering property rights institutions, investor protection, efficient corporate governance and information symmetry for all participants in financial markets. In terms of policy implications, it does not seem reasonable to build policies to change language structures. Linguistic structures are exogenous factors and prevalent in countries over the very long term. Yet, being aware of how such language structures can affect financial development can help policy makers to create an environment that can mitigate the adverse effect of strong FTR languages.

We want to point out that this is a preliminary analysis exploring the relationship between linguistic traits and financial development. We want to remind our readers again that our results and economic interpretations should be read as significant correlation between the variables and not as causation. Future studies can establish identification considering external instruments or matching techniques. We have stuck to the benchmark measure of FTR used in the literature, which is the dummy indicating strong or weak FTR countries based on major spoken language. Additional nuanced measures considering language family or verb ration is beyond the scope of our analysis. We hope this study leads to further research on the important topic of language structures and financial markets and institutions.

**Author Contributions:** N.D. obtained the data and conducted the empirical analysis. N.D. and G.W.C. discussed the results and wrote the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** N. Dutta gratefully acknowledges the financial support from Menard Family Initiative (MFI), University of Wisconsin-La Crosse.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data supporting reported results is available upon request.

**Conflicts of Interest:** The authors declare no conflict of interest.
