ETC4420: Microeconometrics

2x STATA exercises (20% in total) – Some questions and replicating some models on STATA – not very hard but quite lengthy
1x Assignment (20%) – Includes writing a report after estimating a few models taught in the lectures – not very hard as well, but very lengthy and hard to finish, especially for those who haven’t used STATA much.
1x Final exam (60%) – Students thought it was quite lengthy (I thought it was alright tbh), but aside from that, not the hardest exam ever.

This unit is by and large an extension of ETC3410 – Applied Econometrics, with more models, and more weird things to consider in applied economic modelling and practice. We investigate extensions of the models in ETC3410 (binary, instrumental variables, panel data) to imperfect datasets, which are listed below:

1)Discrete choice – when the data is one of a few choices, e.g. ratings on Amazon (which is ordered), or preference in transport (which is unordered) – we look into multinomial choice models such as the ordered probit, multinomial logit, some of their extensions.
2)Counts – when the data is built based on a count random variable (e.g. number of hospital visits) – we look into count regression models such as Poisson or negative binomial models – and some extensions.
3)Data from non-representative samples (censoring, truncation, limited samples) – say for example, we only collected wage information from employed individuals, or recorded any wage below $X as <$X. We use the Tobit model, Heckman sample selection models (which are extensions of instrumental variable estimation in ETC3410) and associated extensions to deal with these defects.
4)Binary panel data – where the dependent variable is a binary choice (e.g. employment status of an individual across 2013-2021). Not too fancy.

Alongside the above we also have concepts such as efficient estimation, set identification, etc. that’s driving current econometric research where we have no definite answers to.

The unit is called microeconometrics because the models taught are usually specifically applied to microeconomic data – data that concerns individual units (people, businesses, and other microeconomic individual units), and how variables have causal effects on one another. Personally, I’m not too interested in these topics but I thought the unit was quite good in introducing advanced, open questions in econometrics that I haven’t thought of. Combining the conceptual, economics knowledge in this unit with more statistics/machine learning concepts and you’ll be a well-rounded practitioner of data analytics and modelling.

I thought the underlying philosophy of this unit is conveyed quite well – in that real-life data is often quite shit, which makes our models usually quite shit, even with some level of sophistication, and improving these models for better outcomes, instead of piling on complexity and hope for the best is an active area of research right now. Whether this unit is taught well is very subjective, but I thought it was alright.

Professor Xueyan Zhao – she’s very nice, and genuinely cares about her students. I’m more of a self-learner, and I thought I needed the textbook to supplement her lecture slides. Her explanations were a bit long-winded but that’s okay

One

3.5 out of 5

Yes, with screen capture

Econometric Analysis – William Greene – only 3 (very long) chapters are needed for this unit, specifically:

Part IV:
17 – Binary Outcomes and Discrete Choices,
18 – Multinomial Choices and Event Counts,
19 – Limited Variables, Truncation, Censoring and Sample Selection.

I thought the textbooks were useful in adding detail and mathematical explanations to the lecture slides, but it’s very long-winded and boring to read through.

1x 3 hour lecture per week

S1 2021

92 HD