Summary: Eec - Overall

Read the summary and the most important questions on EEC - overall

• 1 Week 1: Potential Outcomes and Causality

  This is a preview. There are 5 more flashcards available for chapter 1
  Show more cards here

• When adding control variables, for OLS to provide unbiased and consistent estimators, which assumption needs to hold?

  The zero mean conditional assumption

• The ATE and the ATET are the same if.... (3 scenarios)

  1. Entire population get treated (All students participate in remedial teaching)
  2. Entire population benefits the same from treatment (all students benefit the same from remedial teaching)
  3. A random sample of the population is assigned to treatment (a random sample of the students is assigned to remedial teaching)

• Explain the difference between ATE and ATET (or ATT)

  ATE: you let outcome be the divider; compare the outcome of treated with the outcome of untreated
  
  ATET; you compare hypothetical scenarios; if you are treated we wanna consider the outcome if you had been untreated.

• Provide the formula for the Difference-in-means estimator

  (note I understand why its ATE, not why its ATET...)

• Explain sample selection bias

  Sample selection bias is when your sample is not representative of your population. 
  
  Example; polling for elections to predict who will win; if you end up sampling heavy from one side (e.g. Republicans) your estimates will be off (aka biased).

• Selective participation leads to bias if (1 reason):

  Selection is based on unobserved characteristics which are correlated with Y

• Explain the self-selection problem and how to solve for it

  Self selection essentially means that the 2 groups, treatment and control, are not similar enough in nature for us to isolate any causal effect.
  
  for example; weaker students participate in remedial teaching.
  
  if they are the ones being treated, and the control group is the rest, then you will not be able to measure the treatment effect accurately.
  
  Randomization can solve this

• 2 Week 2: Instrumental variables

  This is a preview. There are 4 more flashcards available for chapter 2
  Show more cards here

• What is the difference between LATE and intention to treat?

  LATE
  

  • Average effect of the treatment for compliers
  • Average earnings returns to completing medical school
  • Average decrease in likelihood to receive benefits after getting a job-search period

  
  
  Intention to treat
  

  • Average effect of assignment to treatment group, regardless of whether the individual actually receives the treatment
  • Average earnings return to winning the first lottery for medical school
  • Average decrease in likelihood te receive benefits when assigned to a caseworker with the default to give a job-search period

• What are the 4 assumptions wecan make, if the zero mean condition assumption holds?

  1. Regressors are exogenous 
  2. OLS estimators are unbiased
  3. OLS estimators are consistent
  4. Coefficients can now considered to be causal effects (rather than correlations or associations)

• What is a bad instrument?

  When the instrument is not exogenous (aka its correlated with the error term