Robustness Checks. Private protection rackets are famously xxxx in Russia and Ukraine, though they are less active in the other post-communist countries. In a survey of Russian shopkeepers by Xxxxxxxxxxx and Xxxx (1998), 33% reported that one of the roles of private protection organizations was to enforce agreements (though far more reported their role was to “protect” the shopkeepers from other criminals). According to anecdotes, though, the mafia plays a larger role with shops than with manufacturing firms of the sort we surveyed. Firms reporting disputes with trading partners were asked whether “an informal private agency specializing in such cases” aided in the resolution of the dispute. Only 5% of firms gave this response, though 48% of Russian firms and 26% of Ukrainian firms reporting disputes said they used such an agency. We create two variables that provide some control for the availability of private enforcement. When added to the basic regression reported in column 3, neither “other third party enforcement” (β=2.01, t=0.52) nor using “an informal private agency specializing in such cases” (β=-8.95, t=1.25) has a significant effect on credit. Their inclusion has almost no effect on the relational contracting variables or on the courts variable. The results shown on Table 3 are also robust to modifications in the sample criteria. All of the reported coefficients remain significant and of close to the same magnitude when relationships and/or firms started more than 10 years before the survey -- prior to the beginning of economic reforms -- are excluded from the sample. The results are similarly robust to limiting the sample to the three Eastern European countries. Including state-owned and export customers also has only modest effects. A regression with state enterprises and export customers is shown in Appendix B, along with separate regressions for each country. (Russia and Ukraine are combine because of the small sample sizes in these countries.) Courts are positively associated with credit in each of the country level regressions, though the effect is significant only in Slovakia.22 With the exception of social networks in Poland, the information network variables also have the right sign everywhere, and are significant most of the time.
Robustness Checks. The results reported on Table 6 are robust to the inclusion of other controls, and to alternative measures of the complexity of the inputs. Neither the availability of other private third party enforcement (β= 0.03, t=0.45) nor the use of private enforcement in resolving the firm’s most recent dispute (β= - 0.01, t=0.37) has a significant effect on the probability of rejecting and offer. The magnitude of the courts effect increases marginally to 7% when either of these are included in the regression. Using independent country and industry controls makes no difference to the results. As an alternative measure of the complexity of the input, we used the question regarding how long it would take the buyer to find alternative inputs if the seller failed to deliver. Longer replacement times imply more risk on the part of the buyer. The time to replace is correlated with both the production of inputs sold only to this buyer (ρ = .20) and with having no alternative supplier (ρ = .34). The time to replace is not significant when it is included with these other two variables, but is when it is used in place of these other variables. The measured effect of courts is not changed with the time to replace is added to the regression; the courts effect is somewhat larger (8%) when time to replace is used in place of the other two variables. Finally, we included measures of the frequency of ongoing visits by the supplier to the manufacturer and measures of the competitiveness of the buyer’s market, and the proportion of the incumbent suppliers bill which the buyer pays with delay. Ongoing visits from the supplier to the customer are significantly associated with higher rates of rejection of the new supplier. One measure of competitive markets (the manager’s estimate of demand elasticity) is significant and indicates that buyers in more competitive markets are more likely to reject the offer. Two measures of competitiveness—the number of competitors located nearby and indication that the buyer prices its goods with reference to competitor’s prices rather than through bargaining with the customer—are not significant. Credit received from the existing supplier also has no significant effect. None of these variables has any effect on the estimated magnitude of the effect of courts. Excluding relationships and firms started more than ten years before the survey has no effect on the magnitude of the courts effect. Neither does limiting the sample to the three Eastern European c...
Robustness Checks. This paper focuses on explaining the degree of access that interest groups obtain to public officials. However, access can be conceptualized as a two-step process in which first interest groups gain access, and, subsequently, they can have repeated access. This section evaluates the robustness of the results while accounting for those organizations without access. More specifically, Table 4.3 below presents the results of hurdle negative binomial models, a two- step method that first assesses the probability of obtaining the binary outcome, in this case obtaining access or not, and subsequently calculates the effects of the same explanatory factors on the level of access (see Xxxxxx et al. 2019 for a similar approach). This model is appropriate in the case of a sequential decision-making process. Even though the two stages of granting access once and deciding to grant access multiple times are estimated separately, the second stage should be interpreted as conditional on the first stage. Prioritizing Professionals? The first step of the model (binary logit), shows that organizational capacity increases the likelihood of gaining access. In contrast, member involvement and functioning as a transmission belt is not related to the probability of gaining access.31 That is, the same organizational factors that explain the level of access seem to explain the likelihood of gaining access. Additionally, the second step of the model (zero-truncated negative bi- nomial) confirm the results presented in Table 4.2. The only differences are found in the significance levels of some control variables. More specifically, the second step of hurdle models show that only organizational scale and resources are significantly related to the degree of access, yet this result is not consistent across all model specifications in Table
Robustness Checks. In order to further test for the validity of the main results of the study, several robustness checks are presented in this section. In Table 1.7, I take advantage of the longitudinal nature of the PSID and estimate model with individuals fixed effects. In the main part of the paper, I use the data as cross sections given that respondents typically do not remain eligible for the EITC for many consecutive years which leads to a drastic reduction of the sample size. Testing whether fixed effect models provide consistent results as the baseline models can further remove concerns about the fact that changes in the composition of the sample are potentially driving the findings of improved health.
Robustness Checks. The previous analysis showed that the implementation of the NMW led to relative health improvements for low-paid workers. This section provides several robustness checks for the main findings. First, I estimate IV models by using an indicator for the policy change as an instrument for reported monthly income. Table 2.10 presents the treatment effects obtained from this specification, whereas the estimates correspond to changes following an increase of income by $1000. It is noticeable that the results are fairly consistent with the previous DD results. The findings provide additional evidence supporting the claim that income improves health and financial well-being of treated workers. In order to test for additional robustness of the main findings, I construct a second control (control group 2) which consists of workers earning a fixed salary who were financially unaffected by the policy change.24 This allows testing for whether the observed health declines were solely experienced by workers who are paid hourly wages. The estimates in Table 2.11 confirm the presence of significant health improvements when comparing outcomes of treated individuals with the new control group. The fact that the magnitude of the ordered logit coefficient and the marginal effects remain analogous to the baseline results indicates that the estimates are robust to the selection of the control group and provide additional evidence for the presence of a downward trend in health in the UK shortly after the reform. Similar finding are found when looking at changes in health conditions between the two groups. Next, I conduct a falsification test by comparing changes in health between the two control groups. Since neither group was financially affected by the reform, no health differences are expected to be found. The results in Table 2.12 confirm this expectation. Financially unaffected workers who are paid hourly are 0.04 and 0.01 percentage points more likely to report excellent health and very good health, respectively. Furthermore, I find no differential impacts on the likelihood of reporting several health conditions as a result of the policy. These findings strengthen the claim that the observed health improvements shown in section (6) are a result of increases in wages that followed the introduction of the NMW.
Robustness Checks. In this section we report the results of several robustness tests of our findings to alternative variable definitions and sample restrictions. In doing so, we build on the single-split specification (II) to 12Regression coefficients in probit models cannot be interpreted as simple slopes as in ordinary linear regressions, but have to be interpreted in terms of Z-scores (i.e. as changes in Z-score for one unit increase in the explanatory variable). consider all available observations and guarantee a sufficient number of observations for the different sample restrictions. First, Table 7 reports the results obtained from modifications of specification (II) aimed at ensuring a vertical-type connection between a firm’s imported inputs and its core export product. In particular, columns (1) and (2) report the results when we restrict the sample to import transactions that are classified as intermediates or capital goods according to the Broad Economic Categories classification; columns (3) and (4) show the results when we use our alternative dependent variable (d integr IFEXihjt), which conditions the classification of transactions as intra-firm also on the existence of a firm’s affiliate in the source country declaring intra-firm export activities. The results in Table 7 confirms those in Table 4: better IPR quality diminishes the propensity to integrate in relatively downstream stages for complements, while the impact for substitutes is not statistically significant. Moreover, the differences between complements and substitutes, in line with theoretical predictions, become more pronounced both with respect to inputs’ upstreamness and relative knowledge intensity along the value chain. Specifically, the impact of Upstr remains significantly negative for complements, while it becomes significantly positive for substitutes in column (2); the interaction between lnIPR and Upstr becomes significantly negative in column (2); and the impact of d knint downstr turns insignificant for complements in column (3), while remaining highly significant and positive for substitutes. Second, Table 8 presents the results obtained using two alternative indicators of sequential com- plements/substitutes described in Section 4.2. In particular, columns (1)-(4) use the indicator d complrho×alpha(ind.) based on the core product’s demand elasticity rho (as a proxy for ρ) and the industry average of the Xxxxxxxxxx index (as a proxy for (inverse) α); columns (5-8) use in- stead the dumm...
Robustness Checks. As reported in Tables 5 and 6, I re-estimated the models by including market orientation and cross-functional integration as control variables to ensure that CBM effect was still significant after controlling for those two related conditions. H1a and b and the results for the interactions are consistent with the results of the analysis without these control variables. Moreover, the incremental R2 explained by adding CBM to models that included market orientation and cross functional integration, respectively, is statistically significant at p<0.05 in the sales growth and employee engagement models indicating that a CBM measure that focuses on internal and community constituencies explains variance over and above these two established variables9. To examine persistence and proxy for omitted variables, I included the lag dependent variable in the sales growth model, in addition to the firm and country level controls. The pattern of results remained largely stable to this addition. In order to verify that the choice of the estimation approach did not bias the results, I also re-estimated the four recursive equations utilizing Roodman’s (2009) conditional mixed process 9 The effect of the CBM measure that includes market constituents however becomes weaker when market orientation is added to the model and in some cases becomes insignificant.
Robustness Checks. Earlier findings that link experience with risk taking could be driven by the fact that there is an inverse correlation between Tracking Error and market capitalization of a fund’s holdings. That is, all other things being equal, a large cap fund has lower Tracking Error than a small cap fund. Furthermore, given that junior managers manage smaller funds, on average, we would expect more Tracking Error for junior managers simply because they manage smaller funds. In order to alleviate these concerns, I divide all funds into two groups based on Morningstar’s categorization of fund size capitalization. Using Morn- ingstar Category grouping, I run the same regressions for risk taking by experience for large and mid/small cap funds separately. Results are reported in Table 11. In Table 11, I again find results consistent with my findings in Table 9 for all fund size groups. Funds managed by junior fund managers take significantly more risk when com- pared to their seasoned counterparts for the more recent period. The results for the earlier period are also consistent throughout the fund size group with earlier finding. With these results, I show that my findings of the relationship between risk taking and experience are robust to the negative correlation between Tracking Error and fund market capitalization. Next, I use other measures of risk taking to test whether my finding is robust to the choice of measuring risk in Table 12. The first alternative measure I use is Amihud and Goyenko (2013) R2 (AG Rsq). I regress the future twelve months of monthly fund excess return over the one month T-xxxx rate on Xxxx-Xxxxxx-Xxxxxxx 4 Factor return to get the R2. As I noted previously, since a high R2 implies lower risk taking/selectivity, I subtract the R2 measure from one to be in accordance with other measures that capture risk taking. The R2 measure has the benefit of not having to know or define the specific benchmark each mutual fund is using and thus is able to successfully detect funds that are truly active in picking stocks against funds that invest in multiple index funds and hide under the radar of other active management measures. The results of the first two specifications show the consistent result that junior fund managers take more risk in the more recent period compared to their seasoned counterparts when AG Rsq is used. The next two measures are based on the holdings of each mutual fund. In order to use holdings-based measures, it is necessary that I co...
Robustness Checks. Thus far, we have relied on bidder announcement returns to measure the effect of CEO home bias on the value of the firm. In this section, we discuss some potential econometric concerns with such an approach and provide a series of robustness checks which, by and large, confirm our main conclusions. Endogeneity is often a major concern in corporate finance studies. In our setting, causal interpretations of the coefficients of interest are only valid if, conditional on our other explanatory variables, the “CEO home bias’’ is randomly assigned. To illustrate this omitted variable problem, suppose that birth rates are higher in exactly the same target states that, for whatever reason, are associated with value destroying acquisitions. In this case, our results could be driven by this omitted variable. In order to address this problem, we try to control for the joint distribution of acquirers and targets using simulations.14 To illustrate this approach, consider first the subsample of cross-state acquisitions. For each cross-state acquisition in which the CEO birth state was the same as the target state, we randomly choose another acquisition with the same bidder and target state but with different CEO birth state. This produces a sample in which the likelihood of a CEO home bias is fifty percent. Next, we run a regression of bidder announcement returns on the CEO home bias dummy and the controls described in Table 3. To prevent our results from being driven by this particular choice of control acquisitions, we repeat this process 1,000 times and use the distribution of coefficients to draw our statistical inferences. Table 9 presents the results using both the states (Panel A) and distances (Panel B) as our measure of birth region proximity. For brevity, we only report the empirical distributions and empirical p-values for the Home Bias coefficients. Consistent with our previous results, we find a negative and significant impact of home bias, but only for distant mergers. For example, in Panel A, the home bias coefficients for in state mergers are not statistically significant, and the economic magnitude is roughly 1/7th of the cross state mergers. The results for distance-based home bias mergers in Panel B are similar. Another potential problem with the interpretation of the coefficients in Table 3 is that our approach relies on bidder announcement returns, whereas it is possible that the market incorrectly assesses the relative merits of home bias mergers. I...
Robustness Checks. Finally, I check the robustness of the findings. Because crises are nonrandom events, the basic selection problem may arise if the crisis actors differ in significant unmeasured ways from the actors who were not involved in crises. There may be important factors not included in the equation for testing crisis outcomes that drive states into crises and try harder to win. Therefore, to account for the nonrandom selection into crises, I use a Xxxxxxx selection model and test the equation for crisis outcomes using the independent variables used in the equation for crisis initiation. A linear combination used to examine the predicted probabilities of the challenger’s victory at different levels of audience costs provides support for the hypothesis that nuclear weapons have a coercive effect on crisis outcomes only when the challenger has high ACC and incurs high audience costs.
