Control Variables. This section provides the empirical results of the determinants of the loan rate. We analyze the determinants of the loan rate by regressing the loan interest rate on our distance, relationship, competition, and control variables, which include loan contract characteristics, loan purpose, firm characteristics, and interest rates. We use the ordinary least squares estimation technique. To benchmark our empirical model, we first analyze and discuss a specification containing only the relationship and control variables. Afterwards, we add our competition and distance variables of interest,45 discuss and interpret the competition and distance results, and perform supplementary robustness tests. First, we regress the loan interest rate (in basis points) on the relationship characteristics and control variables.46 Most control coefficients remain virtually unaltered throughout the exercises in this paper. We therefore tabulate the estimated coefficients only once, in Table 6. The loan contract characteristics include whether the loan is collateralized, its repayment duration, and the loan revisability options. The coefficient of Collateral indicates that when a loan is collateralized, the loan rate decreases by approximately 51 basis points. This result is in line with the sorting-by-private-information paradigm, which predicts that safer borrowers pledge more collateral (e.g., Xxxxxx and Xxxxx (1990) and Xxxxxxx and Xxxxxx (1987)). However, it contrasts with results by Xxxxx and Xxxxxxx (1998) and Xxxxxxxx and Xxxxx (1998), who report a positive (though economically small) effect of collateralization on loan rates. The coefficient of ln(1+Repayment Duration of Loan) is significantly negative at a 1% level: an increase in duration from say five to six years reduces the loan rate by 14 basis points. However, Xxxxxx (1991) finds that an increase in duration from five to six years increases bond yield spreads by around 11 basis points. But the 72 corporate bonds in his sample have maturities longer than seven years, while 88% of our 15,044 sample bank loans have maturities shorter than seven years.47 To replicate his empirical model, we replace ln(1+Repayment Duration of Loan) by a linear and quadratic term in Repayment Duration, and restrict the coefficient on the Government Security variable to be equal to one. Sampling only loans with maturities longer than seven years, we also find that an increase in duration increases bond yield spreads, although the effect i...
Control Variables. Prioritizing Professionals? The paper controls for well-established variables that tap into groups’ characteristics and that relate to their capacity to provide different access goods to public officials. The first control is group type. It is included as a dichotomous variable indicating whether groups are business (e.g., European Dairy Association or the International Union of Combined Road-Rail Transport Companies) or non-business (e.g., European Consumer Organisa- tion or the European Federation of Employees in Public Services). Business groups are expected to be better represented in administrative venues such as the Commission (e.g., Fraussen et al., 2015; Xxxxxx-Xxxx et al., 2017; Xxxxxx et al., 2019). Moreover, recent 87 research has demonstrated that business organizations face more difficulties than citizen groups when establishing policy positions on specific policy issues, which implies that they have a more active involvement of their members (Xx Xxxxxxxx et al., 2019). Related, the correlation matrix in Table A1 in Appendix to Chapter IV shows that business organiza- tions are more likely to approximate the transmission belt ideal, and thus it is important to control for this in the multivariate models. The second control distinguishes whether organizations are mobilized at a national or supranational level (Bunea, 2014). Aligned with previous studies, the Commission is expected to favor the interaction with groups representing encompassing interests that go beyond their national preferences (Bouwen, 2004; Bunea, 2014). Thirdly, the scope of activity of the group – measured with the number of policy domains or sectors in which the group is involved – is included as a control. Here the distinction is between generalists and niche players, and the formers are expected to have more access to the Commis- sion because they are active in more policy domains. Fourthly, membership diversity is included as a count variable to assess the effect of having a diverse set of members on degree of access. The membership options are: private citizens, firms, local and regional governments, national associations, and European associations. Organizational age and resources are also included as controls. In line with previous studies, organizational age is expected to have a positive effect on the level of access to public officials since older groups may have more expertise to engage in lobbying and a wider circle of contacts among public officials (Dür & Xxxxx...
Control Variables. Beyond my key independent variables I include several control variables to guard against spurious results. To control for La Porta et al.’s legal heritage theory, I include a dichotomously coded variable coded 1 if a country uses common law (Common Law) and 0 otherwise. Based on their findings that countries in the British legal tradition have higher corporate governance standards, I expect the coefficient on Common Law to be positive. Moreover, common law and majoritarian legal institutions have a common antecedent in British colonial rule. As such, it is important to ensure that my results concerning proportionality are not the results of a spurious correlation. I also control for partisanship because it is possible the countries with proportional electoral rules are more prone to left governments, which would create a spurious correlation if it was in fact the partisan identity of the government that impacts corporate governance policy outcomes, as suggested by Xxx (2003). To do so, I include a variable for right government, taken from DPI, which is coded 0 for left governments, 1 for centrist governments and 2 for right governments. While this variable has its well known limitations, it is the only data source I am aware of. Its inclusion turns out to be innocuous to the regression estimates of other variables. I control for the log of GDP to guard against the possibility that larger economies have better corporate governance policies, perhaps because of their larger realized or unrealized potential for stock market development, and this could covary with several of my key independent variables. In models using panel data, I also include a trend term.21
Control Variables. There may be other variables confounding the effects of political shock on language regime change. To test for this, I include cultural factors (Arab culture and British colonialism), structural constraints (number of languages and non-dominant group size), and the degree of power-concentration. I also control for temporal and spatial autocorrelation.
Control Variables. In order to examine the relationship between one’s decision to stay unmarried and animation consumption, other factors that affect the marriage decision have to be included in an econometric analysis. The JGSS 2008 provides various social and economic factors that might affect an individual’s marriage decision. In this study in particular, I control for gender, education attainment, geographical residence, living with children and employment status using dummy variables. I also control an individual’s age, personal overall income, number of siblings, and degree of interaction with other people outside family. Table 1 presents the summary statistics of all the variables in the 2008 JGSS being used in this study. Selected control variables used are discussed further below.
Control Variables. In examining the relationship between the independent and dependent variables of the six hypotheses, it is worthwhile to control for potentially confounding factors. Therefore, age, sex, race, income, and education will be included as control variables in the analysis (see Table 2.3). The operationalization of these variables and phrasing of survey questions varies slightly by dataset, but they are nearly identical. Inclusion of such factors is worthwhile given research has found that demographic characteristics are significant predictors of social distancing measures (Nikolov, 2020). Specifically, being Caucasian and having a higher household income are associated with lower adoption of disease mitigating behaviors. By contrast, being older and identifying as a woman are predictive of more mask-use. It follows that these factors should be included in the analysis; however, controlling for education is more complicated. Given I am interested in healthcare workers for evaluating several hypotheses, it would not make sense to include education as a control in these models given healthcare worker status is contingent on educational attainment. Part of the analysis is specifically focused on whether having higher levels of scientific knowledge, or scientific education, affects the extent to which partisanship is predictive of compliance with public health recommendations. Therefore, education should be examined as a control where appropriate (i.e., in evaluating hypotheses 2a-c) and not included in assessing hypotheses 1a-c. Results Qualtrics Questionnaire
Control Variables. In order to eliminate any alternate causes of either homicide or electoral turnover, I will use a number of control variables that have been identified in the literature, given that the data are available. One of the constraints of a project of this depth is time, and given the limited time and data availability to create my own datasets, some of my control variables are measured at the municipal level while others are the state level. At the municipal level I will control for the political party in power, while at the state level I will control for the state economy, individual economic security, resource deprivation, and the presence of drug war activity.
Control Variables. In addition to the similarities across the explanatory variables that occur in some of the cases, Ukraine and Belarus have several common traits. These help eliminate possible alternative explanations for the findings.
Control Variables. The data on the countries’ corporate tax rate, CORPTAX, comes from the BEA survey. I calculated each country’s corporate tax rate from the survey’s Income Statement of Affiliates, Country by Account. I divided the aggregate amount of income taxes paid by foreign affiliates in each country by the aggregate net income of foreign affiliates in each country. I calculated the aggregate net income by subtracting the affiliates’ costs and expenses, excluding income tax, from the foreign affiliates’ total income. As you can see in Tables 2 and 3, some countries had negative corporate income tax rates because the aggregate cost and expenses exceeded the aggregate total income.4 4 A negative tax rate is, in effect, a subsidy. The government gives corporations tax breaks that result in a positive return, rather than a tax liability. The data for the variables GDPCAP and TRADE come from the World Development Indicators available online from the World Bank, and the data for the variable EDU is from UNESCO and is also available online. Data for the variable WTO comes from the official WTO membership list which also gives the effective date of membership. The variable EF, which measures economic freedom, comes from the organization Free the World, and the variable IPR_STRENGTH an intellectual property rights index created by Xxxx and Ginarte (1997).
Control Variables. In most of the estimations GDPCAP was strongly statistically significant and showed an elasticity of between 1.4 and 1.8. However, in the fixed effects estimations TABLE 4 – RANDOM EFFECTS ESTIMATES Corrected for Autocorrelation Not Corrected for Autocorrelation y = ln(RD) Basic Model 20 year Interactions Pharmaceutical Interactions Basic Model 20 year Interactions Pharmaceutica l Interactions Natural log of GDP per capita 1.410*** (0.1772) 1.447*** (0.1853) 1.316*** (0.1847) 1.378*** (0.1810) 1.391*** (0.1878) 1.143*** (0.1861) % total enrollment in secondary school -0.000005 (0.0004) -0.000005 (0.0005) -0.00001 (0.0005) -0.0007 (0.0006) -0.0007 (0.0006) -0.0007 (0.0006) Economic Freedom 0.2097** (0.0838) 0.20808 (0.0844) 0.2131** (0.0844) 0.373*** (0.0757) 0.3439*** (0.0760) 0.3598*** (0.0748) Membership in WTO 0.2563** (0.1186) 0.2652** (0.1197) 0.2700** (0.1194) 0.448*** (0.1019) 0.4474*** (0.1027) 0.4524*** (0.1010) Corporate tax rate -0.1096 (0.0874) -0.1085 (0.0882) -0.1137 (0.0888) -0.1161 (0.1073) -0.1238 (0.1076) -0.1361 (0.1068) % trade volume -0.0006 (0.0030) -0.0010 (0.0030) -0.0015 (0.0029) 0.0027 (0.0027) 0.0026 (0.0027) 0.0020 (0.0026) Xxxx and Ginarte IPR Index 0.7722 (0.1174) 0.0736 (0.1178) 0.7580 (0.1177) -0.0788 (0.1047) -0.0877 (0.1056) -0.0871 (0.1039) 20 year patent duration 0.4217*** (0.1426) 0.4497** (0.1802) 0.4687** (0.1500) 0.2671** (0.1250) 0.1316 (0.2097) 0.4054*** (0.1303) Pharmaceutical patent protection -0.0634 (0.1638) -0.0787 (0.1678) -0.2314 (0.2245) -0.0551 (0.1394) -0.0452 (0.1439) 0.5028 (0.2391) TWENTY * mid income 0.0534 (0.2651) 0.2683 (0.2143) TWENTY * high income -0.2109 (0.2972) -0.0825 (0.2458) PHARMA * mid income 0.3197 (0.2998) 0.6731*** (0.2552) PHARMA * high income 0.32056 (0.3323) 0.9743*** (0.2873) R2 within 0.4016 0.4028 0.4088 0.4268 0.4306 0.4402 R2 between 0.5382 0.5367 0.5517 0.4943 0.4912 0.5184 R2 overall 0.5299 0.5267 0.5425 0.7896 0.4857 0.5175 Number of obs. 492 492 492 492 492 492 Number of groups 49 49 29 49 49 49 25 TABLE 5 – FIXED EFFECTS ESTIMATES Corrected for Autocorrelation Not Corrected for Autocorrelation y = ln(RD) Basic Model 20 year Interactions Pharmaceutica l Interactions Basic Model 20 year Interactions Pharmaceutical Interactions Natural log of GDP per capita 0.2510*** (0.0959) 0.2502*** (0.0970) 0.2287*** (0.0961) 1.763*** (0.3929) 1.746*** (0.3927) 1.696*** (0.3893) % total enrollment in secondary school -0.0001 (0.0004) -0.0001 (0.0004) -0.0001 (0.0004) -0.0007 (0.0006) -0.0...
