ECON 266: Causality in Econometrics

Consider an econometric model designed to estimate the causal effect of a binary treatment, $T$, on an outcome variable, $Y$. The model includes a pre-treatment covariate, $X$, known to influence both the probability of receiving treatment and the outcome. However, the process determining treatment assignment is also influenced by an unobserved variable, $U$, rendering $T$ endogenous. Which of the following strategies would provide the MOST rigorous identification of the causal effect of $T$ on $Y$, assuming that $U$ can be consistently estimated?

In the context of econometric modeling, consider a scenario in which a researcher seeks to estimate the causal impact of firm-level investment in research and development (R&D) on subsequent firm performance, measured by total factor productivity (TFP). The researcher postulates that firms with higher managerial quality are more likely to invest in R&D and exhibit higher TFP, leading to a potential omitted variable bias. Furthermore, the researcher suspects the presence of feedback effects, whereby past TFP influences current R&D investment decisions. Which of the following econometric techniques is MOST appropriate for consistently estimating the causal effect of R&D on TFP, addressing both omitted variable bias and simultaneity?

A researcher is investigating the impact of a new educational program ($E$) on student test scores ($T$). Students are assigned to the program based on their prior academic performance ($A$). However, there is concern that unobserved student characteristics ($U$) may influence both program assignment and test scores, creating an endogeneity problem. To address this, the researcher proposes using the distance from each student's home to the nearest program center ($D$) as an instrumental variable (IV). Which of the following conditions is NECESSARY for $D$ to be a VALID instrument for $E$?

In the context of causal inference, consider a study examining the effect of a job training program on post-training earnings. Program participation ($T$) is voluntary, and individuals with higher motivation ($M$) are more likely to enroll. Earnings ($Y$) are also influenced by motivation and pre-training skills ($S$). The econometrician aims to estimate the average treatment effect on the treated (ATT). Which of the following estimation strategies would MOST effectively address the potential selection bias arising from the fact that participation is not random?

A researcher intends to study the effect of class size ($C$) on student academic performance ($A$). The student assignment into different classes isn't randomized; instead, it relies on student and school characteristics, including parental income ($I$), teacher experience ($X$), and student prior performance ($P$). The researcher suspects that some unobservable factors are lurking, consequently causing endogeneity. Among the choices below, which method will yield the MOST consistent estimation of the causal effect of the class size on academic performance, as long as the provided assumptions are satisfied?

Let's consider a regression model $Y = X\beta + \epsilon$, where $Y$ denotes the outcome variable, $X$ the predictor or independent variable, $\beta$ the coefficient, and $\epsilon$ comprises the error term. Now, presume $\epsilon$ depends on $X$, which therefore renders $X$ endogenous. Also assume that we possess a valid instrumental variable (IV) $Z$ that complies with relevance and exogeneity criteria. In two-stage least squares (2SLS), how can we MOST accurately summarize the causal effect of $X$ on $Y$?

When estimating a regression model $Y = \beta_0 + \beta_1X + \epsilon$, the econometrician is concerned about the presence of perfect multicollinearity. Under what condition does this problem ARISE?

In causal inference, endogeneity is a challenging issue that may jeopardize our ability to get unbiased estimations of causal effects. Which of the following statement regarding the causes of endogeneity is LEAST accurate?

Consider the challenge in econometrics of establishing causality versus correlation. Which of the subsequent statements is the MOST pertinent in differentiating causation from correlation?

Suppose that the error term in an econometric model isn't independent of the explanatory variables. This scenario is most likely to result in which of the following problems?

Within the framework of econometric analysis, what is the critical implication of exogenous variation in an independent variable for establishing causality?

Consider a randomized controlled trial (RCT) designed to evaluate the effect of a new drug on blood pressure. Participants are randomly assigned to either the treatment group (receiving the drug) or the control group (receiving a placebo). However, some participants in the treatment group do not adhere to the prescribed dosage, while some participants in the control group inadvertently take the drug. This situation is an example of...

A researcher wants to understand the effects of education on income based on the following. $Income_i = \beta_0 + \beta_1Education_i + \epsilon_i$ After an initial exploration, it is determined that the error term is correlated with education due to ability bias. If one were to use the number of books in the household as a child as an instrumental variable to address this endogeneity issue, which assumption is MOST critical for the instrumental variable to be valid?

A researcher wants to assess the impacts of a government-sponsored job program on the employment rates of participants. Random assignment was implemented in the design; however, some of those selected to participate did not actually join the program. Under such circumstances, which estimator is more suitable for assessing the treatment effect?

In econometrics, the error term encapsulates numerous factors that affect the dependent variable. What is the MOST important assumption about the error term in classical linear regression models for consistent and unbiased estimation?

When exploring a potential causal relationship between two variables ($X$ and $Y$), a researcher is faced with the problem of endogeneity. Assume that a valid instrument $Z$ has been located. How does Two-Stage Least Squares (2SLS) address endogeneity to estimate the causal effect?

An econometrician wants to estimate the causal effect of police presence on crime rates in different cities. However, cities with high crime rates tend to increase police presence, leading to reverse causality. Which of the following techniques would MOST effectively address this endogeneity issue?

In the context of experimental design, what constitutes the PRIMARY advantage of randomization when assigning subjects to treatment and control groups?

In an econometric model, an independent variable is considered exogenous under which conditions?

When using observational data in econometrics, what is a common challenge that researchers face when attempting to infer causal relationships?

In the study of causal inference, what is the purpose of employing instrumental variables (IV) in econometric analysis?

If a regression model suffers the issue of perfect multicollinearity, it breaks a core OLS assumption. What would be the MOST direct result of this violation?

What is the MOST important reason for employing a randomized controlled trial (RCT) in causal inference?

Suppose a researcher is studying the impact of a new agricultural technique on crop yield, where $Y$ is crop yield and $T$ is the treatment. Now, a severe drought happens to hit the region, reducing yields regardless of whether the technique was used. In this case, how would this affect the estimated coefficient, assuming the drought was not accounted for correctly?

Consider the expression $Y_i = \beta_o + \beta_1Schooling_i + \epsilon_i$. Considering this model, what does the error term capture?

In the provided image, what is an example of a question econometrics can answer?

What is the term for when correlation can be confused for causation?

With every econometric analysis, what is an analyst trying to avoid?

If a researcher generates exogenous variation, what can they be more confident in?