Quantitative Economics

Scheduling: The three groups will meet at the following times:

• Tue 15:30--17:00: Guy, Henry, Joe

• Tue 17:00--18:30: Kavi, Max

• Wed 15:30--17:00: Akshay, Ian

All tutorials will be held in Corpus, in the "PPE Suite" teaching room, which is Room 10 in Staircase FB1 (Fellow's Building Staircase 1). [Look for "FB1" on this map: after entering the staircase, ascend two flights of stairs; the door to the room is immediately on your right.]

Problem sets: Each week I shall assign you problems from the tutorial worksheets posted on the course weblearn page. These must be submitted to me, via the Corpus Lodge, by 3pm on Sunday. I shall not grade your answers to all the problems that have been assigned: instead, I shall only grade a certain subset of these (to be decided upon after you have submitted your solutions); in the tutorials I shall give priority to going through the solutions to those questions that I do not grade.

I shall also usually assign some reading to complement the lecture slides, drawn from either Stock & Watson, Introduction to Econometrics [S&W]; Angrist & Pischke, Mastering 'Metrics [MM]; or Angrist & Pischke, Mostly Harmless Econometrics [MHE]. I've instructed both the Corpus and Queen's libraries to increase their holdings of these titles. [But I regard these readings as being very subsidiary to the problem sets.]

Tutorial 1 (Week 2)

Problems: all questions in worksheet 1

Readings: S&W, Ch. 1, Ch 2.1--2.4

Tutorial 2 (Week 3)

Problems: 1, 2, 3, 5, 8, 10, 11, 12, 14 in worksheet 2

Data-based questions: 16 and 'Causal Inference' question (on separate worksheet)

Readings: S&W, Ch. 2 and 3

Tutorial 3 (Week 4)

I will ask you to submit solutions to a smaller number of problems this week (and in subsequent weeks). As we discussed in tutorials, I accordingly expect that you write more detailed solutions to these problems: you need to explain what you are doing, rather than simply mechanically substituting numbers into formulae taken from the lecture notes. For example, if you are asked to conduct a hypothesis test, you should state each of the following clearly:

• 1. The null and alternative hypothesis under test (remember these must refer to parameters, not estimators)

  2. The test statistic that you will use

  3. The distribution (exact, or approximate in large samples) of the test statistic under the null

  4. The form of the decision rule (i.e. 'Reject the null hypothesis if ...')

  5. The significance level of the test, and the critical value(s) that will deliver a test with the chosen significance level

  6. The value of the test statistic computed from the estimates, and your accept/reject decision.

This is how you will be expected to answer such a question in the exam; just supplying (say) 1 and 6 will not be enough to score full marks.

Problems to submit:

• • Worksheet 2: 14. (For those of few you who gave an incomplete or incorrect answer to this last week, please make another attempt at it.)

  • Worksheet 3a: 2 and 3.

  • Worksheet 3b: 4, 7 and 8. (In 7(b), to compute the F-statistic correctly -- ignore the 'rule of thumb' comment -- you will need to look at Section 7.2 of S&W.)

Readings:

• • For regression: S&W, Ch 4, 5.1-5.3, 6, 7.1-7.2 [and optionally MM, Ch 2.]

  • For causal inference and RCTs (from last week): MM, Ch 1.

Additionally attempt, but do not submit: 1 in worksheet 3a; 1 and 3 in worksheet 3b

Tutorial 4 (Week 5)

To answer q3 and q4, you will need to run several regressions, and construct new variables on the basis of the estimates obtained. I strongly recommend performing these calculations in either R or Matlab, or some other statistical package. (Doing the calculations in Excel, while possible in principle, would be extremely tedious.)

• • R: follow the installation instructions given here on the QE weblearn page.

  • Matlab: all Oxford students have access to a licence; follow the instructions here. During installation, make sure to selection the "Statistics" toolbox. Matlab comes provided with very extensive tutorials and documentation.

You may submit work electronically, but if you do so please submit it as a pdf, so that there is some hope of it displaying correctly on my computer.

Problems: 2, 3, 4, 5 in worksheet 4

Readings: S&W, Ch. 7.3-7.5, 9.1, 9.2, 9.4, 12.1-12.3

Note about conditional expectations, relevant to q5.

Tutorial 5 (Week 6)

Problems: 1, 2, 3, 4 in worksheet 5a

Readings: MM, Ch. 3; S&W, 13.1-13.3

Tutorial 6 (Week 8)

Problems: 3, 4 in worksheet 5b; 1, 3, 4 in worksheet 6

Readings: MM, Ch. 6; S&W, Ch. 14, 16.4

Revision Classes

Scheduling: Weeks 1--3 of Trinity term, Wednesdays 10:30--12:00 (though I am happy to continue to 12:30 if there are sufficient questions), in the Large Discussion Room of the Social Science Library (Manor Rd). In Week 4, I will be available at that same time in my office (FB Attic Rm 1 at Corpus), to discuss any final questions you have about the material.

We shall proceed by reviewing solutions to past exam questions indicated below, but I am also happy to answer any other questions you have related to this material.

Class 1 (Week 1): Probability, hypothesis testing and regression

Lecture material: Weeks 1--3

Past exam questions:

• • 2016q4: properties of expectation and variance

  • 2017q4: conditional expectation and linear regression

  • 2015q3 and 2014q6: interpretation of regression estimates; hypothesis testing

Class 2 (Week 2): Applied regression, instrumental variables and RCTs

Lecture material: Weeks 4--5

Past exam questions:

• • 2016q5 and 2015q4: causal interpretation of regression; LATE; instrument validity

  • 2017q7: RCTs, treatments and compliance

  • 2017q5: omitted variable bias and bad controls

Class 3 (Week 3): Time series

Lecture material: Weeks 7--8

• • Tutorial 6, q2: forecasting,

  • Tutorial 6, q3: dynamic multipliers, breaks

  • 2016q2: inference with unit roots

  • 2016q9, corrected: cointegration

  • 2015q6: Granger causality, nonstationarity