- A researcher develops two different regression models to predict the level of political polarization in a country. The results are summarized in the table. Based on this information, which conclusion is the most methodologically sound?
Answer: Model A is likely preferable because its higher adjusted R-squared suggests a better balance of explanatory power and parsimony.
Adjusted R-squared is used to compare the fit of models with different numbers of predictors. It penalizes the inclusion of extra variables that do not improve the model's fit sufficiently. While Model B has a slightly higher R-squared, its adjusted R-squared is lower than Model A's. This indicates that the additional 5 predictors in Model B are not adding enough explanatory power to justify their inclusion. Therefore, the more parsimonious Model A is likely preferable. (A) ignores the penalty for adding more predictors and incorrectly focuses only on R-squared. (C) is a vague claim; adjusted R-squared provides a clear criterion for comparison. (D) is incorrect because adjusted R-squared is specifically designed for this type of comparison.
- A regression model examines support for increased environmental regulation (measured 0-100). The table below shows results for two variables:
Education (in years) and Party_ID (coded 1 for Democrat, 0 for Republican). Both coefficients are statistically significant. Based on these results, which statement accurately compares the effects of the two variables? Answer: The effect of being a Democrat on support is equivalent to the effect of approximately 8 additional years of education, holding other factors constant.
This requires comparing the magnitudes of the coefficients relative to their scales. The coefficient for Party_ID is 12.0, meaning a shift from Republican to Democrat is associated with a 12-point increase in support. The coefficient for Education is 1.5, meaning each additional year is associated with a 1.5-point increase. To find the equivalent effect, we divide the party coefficient by the education coefficient: 12.0 / 1.5 = 8. This means the effect of being a Democrat (a one-unit change) is equivalent to 8 years of education. (A) is a flawed comparison because the units are different. (C) is factually incorrect based on the coefficients. (D) is incorrect because a direct comparison in terms of variable units is possible and meaningful, even if a comparison of overall 'importance' might require standardization.
- A cross-sectional study of 50 nations finds a strong, statistically significant, positive correlation between the number of active civil society organizations and the country's level of press freedom. The model has a high R-squared. What is the most cautious and appropriate conclusion to draw from this finding?
Answer: Countries with more active civil society organizations tend to have greater press freedom, but this correlation does not in itself establish a causal relationship.
Regression and correlation show association, not causation. This is especially true in cross-sectional studies where temporal order cannot be established. The relationship could be causal in either direction (A or B), or it could be spurious, with a third factor (like level of democracy) causing both. Therefore, the most appropriate conclusion is a descriptive one that acknowledges the association without making a causal claim. (A) and (B) assume a specific causal direction. (C) incorrectly uses the word 'proves' and insists on a causal link.
- A researcher runs a simple regression and finds a strong, positive, and statistically significant relationship between the number of public libraries per capita in a city and the city's rate of violent crime. A critic argues this result is likely due to omitted variable bias. Which of the following potential omitted variables would most plausibly explain this counterintuitive finding?
Answer: The city's population density, because it is positively correlated with the need for public services like libraries and also tends to be positively correlated with crime rates.
Omitted variable bias occurs when a variable that is correlated with both the independent and dependent variables is left out of the model. To explain a positive coefficient between libraries and crime, the omitted variable must be positively correlated with both. Population density is a plausible candidate: denser areas require more services (like libraries) and also tend to have higher crime rates. (A) is incorrect because an omitted variable must be related to both X and Y. (B) is incorrect because the variable is unlikely to be related to the dependent variable (crime). (C) describes a variable that would likely induce a negative bias, not explain the positive finding.
- A study models the annual budget (in millions of dollars) of a federal agency. One independent variable is a dummy for whether the agency is a 'regulatory' agency (1=yes, 0=no). The regression results are in the table below. Based on these results, what can be concluded about the budgets of regulatory versus non-regulatory agencies?
Answer: On average, regulatory agencies have budgets that are $50.3 million higher than non-regulatory agencies, holding other factors constant.
The coefficient of a dummy variable represents the average difference in the dependent variable between the category coded as 1 and the reference category (coded as 0), controlling for other variables in the model. Here, the coefficient of 50.3 means regulatory agencies (coded 1) have budgets that are, on average, $50.3 million higher than non-regulatory agencies (coded 0). (A) misinterprets the coefficient as the total budget rather than the difference. (C) makes a calculation error; the predicted budget for a non-regulatory agency (when Regulatory=0) is the intercept, $125.1M. The predicted budget for a regulatory agency is the intercept plus the coefficient: $125.1M + $50.3M = $175.4M. While the calculation is correct, (B) is the more direct and accurate interpretation of the coefficient itself. (D) misinterprets the coefficient as a percentage change.
- A study on the effects of foreign aid on democratic development reports a coefficient for 'Aid as % of GDP'. Instead of a p-value, it provides a 95% confidence interval for this coefficient, which is [0.05, 0.45]. Based on this confidence interval, what can be concluded?
Answer: The effect of foreign aid is statistically significant at the 0.05 level because the interval does not contain zero.
A confidence interval provides a range of plausible values for the true population parameter. Statistical significance at the α = 0.05 level is established if the 95% confidence interval does not contain the null hypothesis value (which is typically 0). Since the interval [0.05, 0.45] is entirely positive and does not include 0, we can reject the null hypothesis and conclude the effect is statistically significant. (A) is a common misinterpretation; in frequentist statistics, the interval is random, not the parameter. The correct phrasing is about confidence in the procedure, not probability about the parameter. (B) is incorrect; the width of the interval relates to precision, but significance is determined by whether it contains zero. (D) misinterprets the confidence level as a measure of predictive accuracy.
- A study models citizens' trust in the national legislature on a scale of 0 to 10. The independent variables include
Age (in years) and Education (in years). The sample includes respondents aged 18 to 92. The regression output is presented in the table. What is the correct interpretation of the intercept? Answer: It is the predicted level of legislative trust for a newborn with zero years of education, which is a substantively meaningless extrapolation.
The intercept is the predicted value of the dependent variable when all independent variables are zero. In this case, it is the predicted trust for a person of age 0 with 0 years of education. Since the data only includes adults (age 18+), this is an out-of-sample extrapolation that has no practical or substantive meaning. (A) confuses the intercept with the mean of the dependent variable. (B) confuses the intercept with the minimum value of the dependent variable. (D) describes the coefficients of the other variables, not the intercept.
- A political scientist models the effect of campaign spending on electoral outcomes. The dependent variable is a candidate's vote share (from 0 to 100), and the key independent variable is campaign spending, measured in units of $100,000. The regression output is shown below. Based on the table, which of the following is the most accurate interpretation of the coefficient for campaign spending?
Answer: A $100,000 increase in campaign spending is associated with a 0.5 percentage point increase in the candidate's vote share, on average.
The coefficient of 0.50 represents the change in the dependent variable (Vote Share in %) for a one-unit change in the independent variable. Since the independent variable is measured in units of $100,000, a one-unit change corresponds to a $100,000 increase, which is associated with a 0.5 percentage point increase in vote share. (B) is incorrect because it misunderstands the units of the independent variable. (C) is incorrect because regression analysis shows association, not causation. (D) misinterprets the coefficient as a direct share of total votes rather than a marginal change in the percentage share.
- A cross-national study examines the relationship between economic inequality (measured by the Gini coefficient) and political instability. A key finding for the Gini coefficient variable is a p-value of 0.03. Which of the following is the most accurate conclusion that can be drawn from this specific p-value?
Answer: If the null hypothesis of no relationship is true, there is a 3% probability of observing a relationship at least as strong as the one found in the sample.
The p-value is the probability of observing the current result, or a more extreme one, assuming the null hypothesis is true. A p-value of 0.03 means that there is a 3% chance of getting a result this strong or stronger by random chance if there were actually no relationship in the population. (A) is a common but incorrect interpretation of the p-value; it is not the probability that the null hypothesis is true. (C) makes a causal claim, which cannot be proven by the model, and misinterprets the meaning of the p-value as a confidence level. (D) confuses statistical significance with substantive significance; a small p-value does not necessarily mean the effect size is large or practically important.
- A researcher investigates factors influencing judicial confirmations in the U.S. Senate. The model predicts the number of 'nay' votes a judicial nominee receives. One key independent variable is
Nominee_Ideology, a scale from -1 (most liberal) to +1 (most conservative). A second is Divided_Government, a dummy variable coded 1 if the presidency and Senate are controlled by different parties, and 0 otherwise. Given the results in the table, what is the predicted number of 'nay' votes for a ideologically moderate nominee (score of 0) under a divided government? Answer: 33 votes
To find the predicted value, one must sum the intercept and the product of each coefficient and its corresponding variable value. For an ideologically moderate nominee (Nominee_Ideology = 0) under a divided government (Divided_Government = 1): Predicted Nay Votes = Intercept + (β₁ Nominee_Ideology) + (β₂ Divided_Government) = 18 + (10 0) + (15 1) = 18 + 0 + 15 = 33. (A) incorrectly uses only the Divided_Government coefficient. (B) incorrectly uses only the intercept. (C) is the sum of the Divided_Government coefficient and the Nominee_Ideology coefficient, ignoring the intercept.
- An analyst is studying political participation in a small N study of 15 Western European countries. The regression model shows a positive coefficient for the effect of compulsory voting laws on voter turnout, but the p-value is 0.15. What is the most methodologically sound interpretation of this finding?
Answer: The study fails to reject the null hypothesis of no effect, which may be due to the small sample size providing insufficient statistical power.
When analyzing statistical results in small-N comparative studies, you need to distinguish between statistical significance, practical significance, and the limitations imposed by sample size. This question tests your understanding of hypothesis testing and statistical power in political science research.
Answer A correctly interprets this finding. With a p-value of 0.15, you fail to reject the null hypothesis at conventional significance levels (0.05 or 0.01), meaning you cannot conclude there's a statistically significant effect. However, the positive coefficient suggests a relationship might exist, and with only 15 countries, the study likely lacks sufficient statistical power to detect even meaningful effects. Small samples make it harder to achieve statistical significance even when real relationships exist.
Answer B makes a fundamental error by claiming the study "proves" no effect exists. Failing to reject the null hypothesis never proves the null is true—it simply means you lack sufficient evidence to conclude an effect exists. This is a classic Type II error interpretation.
Answer C confuses statistical significance with practical significance. The coefficient's size doesn't determine whether it's "meaningful"—statistical significance testing is about whether observed effects are likely due to chance, not about substantive importance.
Answer D is too extreme and ignores the broader context. While some fields use p < 0.10, dismissing results entirely because they don't meet arbitrary thresholds wastes valuable information, especially in small-N studies where power is inherently limited.
Remember: in comparative politics with small samples, always consider statistical power alongside significance tests when interpreting null findings.
- A researcher tests the hypothesis that a new online voter education program increases political knowledge. After a randomized controlled trial, a regression of political knowledge scores on a 'treatment' dummy variable (1=exposed to program, 0=control) yields a coefficient of 0.50 with a p-value of 0.40. What is the most appropriate conclusion from this analysis?
Answer: The analysis does not provide sufficient statistical evidence to conclude that the program is associated with a change in political knowledge.
A p-value of 0.40 indicates that the observed result is very likely to have occurred by chance if the true effect were zero. Therefore, we fail to reject the null hypothesis. The most precise conclusion is that there is insufficient evidence to support the research hypothesis. (A) is too strong; failing to find evidence of an effect is not the same as proving there is no effect. (B) is misleading because it presents the positive point estimate (0.50) as a finding, whereas the high p-value indicates this estimate cannot be distinguished from random noise. (D) is incorrect; randomization ensures internal validity but does not guarantee a significant result, especially if the treatment effect is small or the sample size is limited.
- A political scientist develops a model to explain variation in voter turnout across different precincts in a city. The model includes variables for average income, percentage of residents with a college degree, and average age. The model produces an R-squared value of 0.18. Which of the following is the correct interpretation of this R-squared value?
Answer: Eighteen percent of the variation in voter turnout across precincts is explained by the variation in the model's independent variables.
R-squared, or the coefficient of determination, measures the proportion of the variance in the dependent variable that is predictable from the independent variables. An R-squared of 0.18 means that 18% of the variance in voter turnout is accounted for by the model. (A) is incorrect because R-squared does not establish causality. (B) misinterprets R-squared as a measure of individual prediction accuracy. (D) is a subjective judgment; in social sciences, an R-squared of 0.18 can be quite meaningful, and there is no universal threshold for what constitutes a 'good' fit.
- In a regression model, the standard error of a coefficient is a critical piece of information used to assess statistical significance. Which of the following best describes the conceptual meaning of a coefficient's standard error?
Answer: It is a measure of the typical distance that the estimated coefficient is from the true, unobservable population coefficient.
When you encounter questions about regression analysis, focus on distinguishing between different measures of uncertainty and model performance. The standard error of a coefficient is fundamentally about sampling variability—how much your estimated coefficient would fluctuate if you repeated your study with different samples from the same population.
The standard error captures the typical distance between your estimated coefficient and the true population parameter you're trying to estimate. Since you can never observe the true population coefficient directly, the standard error quantifies the uncertainty inherent in your estimate. This makes option A correct—it's essentially measuring how "off" your estimate might be from reality.
Option B describes the residual sum of squares or mean squared error, which measures how well your model fits the observed data points, not the precision of individual coefficients. Option C confuses the standard error with a p-value; the standard error helps calculate p-values, but it's not itself a probability. Option D describes a standardized coefficient (beta coefficient), which shows the effect size in standard deviation units rather than measuring estimation uncertainty.
The standard error serves as the denominator when calculating t-statistics for hypothesis testing: t=standard errorcoefficient estimate. A smaller standard error means more precise estimation and larger t-statistics.
Study tip: Remember that standard errors always relate to sampling uncertainty—they tell you how much estimates would vary across different samples. This distinguishes them from model fit measures or effect size measures. - A sociologist of politics models the relationship between a citizen's age and their level of political engagement. To capture a potentially non-linear relationship, the model includes both
Age and Age-squared as predictors. The results are shown in the table below. What do these coefficients suggest about the relationship between age and political engagement? Answer: Political engagement increases with age up to a certain point, after which it begins to decline.
A positive coefficient on Age and a negative coefficient on Age-squared indicates a concave, inverted U-shaped relationship. This means the dependent variable (political engagement) initially increases as age increases, but at a diminishing rate. Eventually, the relationship peaks and then begins to decline as the negative effect of the squared term overtakes the positive linear term. (A) ignores the significant squared term. (B) misunderstands how to interpret polynomial terms. (D) would be true if the Age coefficient were negative and the Age-squared coefficient were positive.
- A political economist models the number of anti-government protests per year in various countries. To account for the wide range in economic development, the model uses the natural log of GDP per capita,
ln(GDP), as an independent variable. The dependent variable is the count of protests. The coefficient for ln(GDP) is -25.0 and is statistically significant. How should this coefficient be interpreted? Answer: A 1% increase in a country's GDP per capita is associated with a decrease of approximately 0.25 protests per year.
This is a level-log model, where the dependent variable is in its original units (level) and the independent variable is logged. The interpretation of the coefficient β is that a 1% change in the independent variable is associated with a (β/100) unit change in the dependent variable. Here, β = -25.0. So, a 1% increase in GDP per capita is associated with a (-25.0 / 100) = -0.25 change in the number of protests. (A) incorrectly interprets the logged variable as if it were linear. (C) is the interpretation for a log-log model. (D) misinterprets the sign of the coefficient.
- A researcher studies how political ideology and gender relate to opinions on free trade. The model includes
Conservatism (a scale from 1=low to 10=high), Female (1=female, 0=male), and an interaction term Conservatism * Female. The dependent variable is support for free trade (1-10 scale). Based on the significant coefficients in the table, how does the relationship between conservatism and support for free trade differ for men and women? Answer: For men, each one-point increase in conservatism is associated with a 0.4-point increase in support; for women, the association is weaker (0.1-point increase).
Interpreting an interaction term requires combining coefficients. For men (Female=0), the effect of conservatism is just its main coefficient: +0.4. For women (Female=1), the effect of conservatism is the sum of its main coefficient and the interaction term's coefficient: 0.4 + (-0.3) = +0.1. Therefore, the positive relationship between conservatism and support is weaker for women than for men. (B) misinterprets the interaction term. (C) incorrectly calculates the effect for women. (D) misinterprets the main effect of Female (-2.5) as being constant, ignoring that its total effect depends on the level of conservatism due to the interaction term.
- A researcher presents a regression table modeling support for a universal basic income (UBI) policy on a 0-10 scale. Based on the full table of results, which statement provides the most accurate and complete summary of the findings?
Answer: Being younger and identifying as liberal are associated with higher support for UBI, while income level is not a statistically significant predictor, and these factors jointly explain 32% of the variance in support.
A complete summary must address the significant and non-significant findings and the overall model fit. The table shows Age has a significant negative coefficient, Liberal_ID has a significant positive coefficient, and Income is not significant (p=0.25). The R-squared is 0.32, meaning the model explains 32% of the variance. Statement (C) correctly synthesizes all these points. (A) is incorrect because Income is not a significant predictor. (B) is an oversimplification that ignores the significant effect of Age. (D) makes a causal claim ('proves that economic self-interest drives...') which is not warranted by the regression results, especially since Income was not significant.
