Questions: R-Squared and Model Fit

5 questions to test your understanding

Score: 0 / 5
Question 1 Multiple Choice

A researcher adds three variables to a regression and reports that R² increased from 0.45 to 0.52. A colleague concludes the model is now substantially better. What is wrong with this interpretation?

AR² should have decreased when variables are added
BR² mechanically cannot decrease when variables are added, so the increase tells us nothing about whether the new variables are informative — adjusted R² is needed
CAn increase of 0.07 in R² is always too small to be meaningful
DR² only measures fit for the training data, so this comparison is invalid
Question 2 Multiple Choice

A randomized experiment finds that a job training program increases wages by $200/week (p < 0.001), but R² = 0.03. A critic argues: 'The model explains almost nothing — this result can't be trusted.' What is the correct response?

AThe critic is right — a low R² indicates the estimate is biased
BThe critic is wrong — R² measures explained variance, not causal validity; wages vary for many reasons beyond the program, and the coefficient can be unbiased even if R² is low
CThe critic is right — a larger sample would raise R² and validate the result
DThe critic is wrong, but R² should be at least 0.10 to report results in economics
Question 3 True / False

Adding any regressor to a regression, even an irrelevant one, can never decrease R².

TTrue
FFalse
Question 4 True / False

A regression with R² = 0.92 provides stronger evidence for a valid causal estimate than one with R² = 0.08, most else equal.

TTrue
FFalse
Question 5 Short Answer

Why do econometricians pursuing causal identification often report low R² without apology, and what would actually need to be true for their coefficient estimates to be valid?

Think about your answer, then reveal below.