ECON2515 · Intermediate Applied Econometrics Ii
The Complete Exam Bible · S2 2026Intermediate Applied Econometrics II
ECON 2515 Intermediate Applied Econometrics is the University of Adelaide's second-year course on using real-world observational data to answer economic questions — does more education raise income, does a subsidy change behaviour — not just to spot patterns. You learn to estimate and interpret regression models in R, moving from the simple OLS estimator to multiple regression, then to inference (t-tests, confidence intervals, p-values and F-tests) and the specification problems that break it (omitted-variable bias, multicollinearity, heteroskedasticity). The recurring theme is that regression is not causation: a significant coefficient is only causal if the key exogeneity assumption holds. The closed-book final rewards interpretation and judgement over formula recall — you must read the R output, defend the test you chose, and diagnose the model, so practising worked interpretations matters far more than memorising derivations.
What ECON 2515 covers
ECON 2515 runs across ten teaching topics that build from what econometrics is, through the OLS estimator and multiple regression, into the full inference toolkit (t-tests, p-values, F-tests) and the model-specification problems of multicollinearity and heteroskedasticity, closing with an applied model-critique review.
How ECON 2515 is assessed
| Component | Weight | Format |
|---|---|---|
| Quizzes (Quiz 1-4, online timed) | subject to confirmation | Online Canvas quizzes, MCQ + short numeric, timed 20-50 min, due across the term (raw 10 pts each) |
| Group Assignment 1 - Written Report | subject to confirmation | Group empirical written report using real data in R (raw 80 pts), due ~Week 5 |
| Group Assignment 2 - Video Presentation | subject to confirmation | Group video presentation of an empirical analysis (raw 80 pts), due ~Week 10 |
| Final Examination | subject to confirmation | 2 hr face-to-face closed-book: Part A 15 MCQ x2 marks + Part B 2 worked-answer questions (40 marks) |
Multiple regression: interpret a coefficient, find the marginal effect and the turning point
- +3(a) EDUC enters linearly, so β̂₂ is a level-level effect: holding experience (and its square) constant, one more year of education is associated with about $28.50 more in weekly earnings. State 'ceteris paribus' explicitly — in multiple regression every slope is a partialled-out, other-things-equal effect.
- +2(b) Because experience appears as EXPER and EXPER², its effect is not constant. Differentiate: ∂EARN/∂EXPER = β̂₃ + 2β̂₄·EXPER = 11.2 + 2(−0.160)·EXPER = 11.2 − 0.32·EXPER.
- +2(b, cont.) Evaluate at EXPER = 6: 11.2 − 0.32(6) = 11.2 − 1.92 = $9.28 of extra weekly earnings per additional year of experience at that point.
- +3(c) Earnings peak where the marginal effect is zero: 11.2 − 0.32·EXPER = 0, so EXPER* = 11.2 / 0.32 = 35 years. Beyond 35 years, extra experience is associated with lower earnings (β̂₄ < 0 gives a hump shape).
Key terms
- Econometrics
- Using statistical methods on real-world (usually observational) economic data to quantify relationships, test theories, forecast and evaluate policy — with a focus on causal, not merely predictive, questions.
- Ceteris paribus
- 'All else equal.' The thought experiment that defines a causal effect: how y changes when one x changes while all other relevant factors are held constant.
- OLS estimator
- Ordinary least squares — the rule that picks the intercept and slopes to minimise the sum of squared residuals Σ(yᵢ − ŷᵢ)²; unbiased and BLUE under the Gauss-Markov assumptions.
- Zero conditional mean (exogeneity)
- The assumption E[u|x] = 0 (SLR.4 / MLR.4). It is what makes OLS unbiased and the estimate causal; it fails under omitted variables, reverse causality or measurement error.
- Omitted variable bias (OVB)
- The bias in an OLS coefficient when a variable that belongs in the model and is correlated with an included regressor is left out; direction = sign(effect on y) × sign(correlation with the included x).
- R-squared / adjusted R-squared
- R² = SSE/SST is the share of variation in y explained by the model; it only rises as regressors are added, so adjusted R² penalises extra variables and is used to compare models of different size.
- Functional form
- How variables enter the (still linear-in-parameters) model — level-level, log-linear, linear-log, log-log or quadratic — which fixes how the slope and elasticity are interpreted.
- Elasticity
- The percentage change in y for a 1% change in x, (∂y/∂x)·(x/y); read directly as β₁ in a log-log model.
- t-statistic
- t = (β̂ − c)/se(β̂) ~ t(n−k−1), measuring how many standard errors the estimate sits from the null value c; the basis of single-parameter hypothesis tests and confidence intervals.
- p-value
- The probability of a test statistic at least as extreme as the one observed, assuming H₀ is true; reject H₀ when p ≤ α. It measures evidence against the null, not the probability the null is true.
- F-test
- A joint test of several restrictions at once, F = [(SSR_R − SSR_U)/q] / [SSR_U/(n−k−1)] ~ F(q, n−k−1); one-tailed, and equal to t² for a single restriction.
- Multicollinearity
- High (but not perfect) correlation among regressors; it inflates standard errors and makes individual t-stats insignificant (VIF > 10 flags it) but does not bias the coefficients.
- Heteroskedasticity
- Non-constant error variance, Var(u|x) ≠ σ² (violates MLR.5); OLS stays unbiased but the usual standard errors are wrong, so inference needs robust (White) standard errors or WLS/GLS.
ECON 2515 FAQ
How is ECON 2515 assessed?
Assessment combines online timed Canvas quizzes (Quiz 1-4, MCQ + short numeric), two group tasks (a written empirical report in R around Week 5 and a video presentation around Week 10), and a 2-hour face-to-face closed-book final exam. Exact weightings are set by the course profile each offering and are subject to confirmation, so check your official Canvas page.
Is there a final exam, and what does it look like?
Yes — a 2-hour, face-to-face, closed-book final. Part A is 15 multiple-choice questions worth 2 marks each (30 marks); Part B is 2 worked-answer questions with parts (40 marks) where you must show all steps. You may bring one A4 cheat sheet (both sides) and a calculator; statistical tables (t, F, normal) are provided.
What is the hardest part of the course?
Most students find the interpretation and inference judgement calls harder than the algebra: getting the log/linear/log-log/quadratic/interaction interpretation direction right, computing marginal effects and turning points, choosing one- vs two-sided tests with the correct degrees of freedom, and telling apart the three specification problems (OVB biases coefficients, while multicollinearity and heteroskedasticity affect only precision or the standard errors).
How should I prepare for the exam?
The exam rewards application, not recall, so practise reading R/STATA output and writing full worked answers: interpret every coefficient type, run the 6-step hypothesis-test template, compute t, F, CIs and marginal effects from given numbers, and always separate statistical significance from economic magnitude and from causal validity. Build your one-page A4 sheet early around the formula list, the assumption table and the functional-form interpretations.
Do I need to know R for this course?
Yes. Workshops, quizzes and the group assignments use R / RStudio to estimate models and run diagnostics (e.g. VIFs, Breusch-Pagan and White tests, robust standard errors). The exam itself is closed-book on paper but tests your ability to read and interpret the kind of R output you produced during the term.
What maths background do I need?
The course assumes comfort with basic probability (expectation, variance, covariance, conditional distributions) and single-variable calculus (power, exp and log rules; the chain and product rules) so you can derive OLS and compute marginal effects. Week 1 reviews these, but revisit them early if they are rusty.
Is this page official or affiliated with the University of Adelaide?
No. This is an independent AskSia study guide built to help you revise ECON 2515. It is not produced, endorsed by, or affiliated with the University of Adelaide, and it does not reproduce official course materials. Always confirm assessment details, dates and weightings on your official Canvas course profile.
How to study for the exam
Treat ECON 2515 as a skills course, not a memory course: keep pace with the pre-workshop videos and slides each week so the 3-hour workshop is spent practising in R rather than catching up, and do every quiz and worked problem by hand at least once so you can reproduce a t-stat, an F-stat, a confidence interval, a marginal effect and a turning point without the software. As the term builds, maintain a running one-page A4 sheet — OLS variance, R²/adjusted R², t/F/CI formulas, the MLR.1-6 assumption list and what each one buys, the functional-form interpretation table, and the 6-step test write-up — because that same sheet is your allowed exam aid. In the final weeks, drill past-style Part B questions end to end: read the output, interpret each coefficient with the correct log/dummy/quadratic/interaction rule, run the test with the right degrees of freedom and tail, then finish with a sentence on economic magnitude and whether the estimate can be called causal.
