Lecture 4 Instrumental Variables IV Estimation of the Return

Lecture 4: Instrumental Variables IV Estimation of the Returns to Education The Ability Bia e Suppose where E(Usleduci, ability) 0 Ability is undefinable, therefore not included. RCT not possible This fits the general IV framework. . Write with V Bability Ui Then (using the LIE) E(Vleduc) E(ability leduc) The PRF thus does not match the causal equation: . An IV for Education? We need a variable that is . 1. Correlated with educ; (relevant) 2. Uncorrelated with ability, (exogenous)

Solution

Randomized Controlled Trial - People are allocated at random for an intervention

The conditional expectation for an undefined variable

The difference between the groups is the variable being studied.

Ability value conditional to education considers other variables as well; it is not unconditioned only on the kind of education, or, the number of years of education. Ability bias, in this case, is the omitted variable. Therefore, randomized control trial is not possible. Random assignment of intervention taking certain groups will not give the results as the results will vary within a group, and, outside a group. The control groups will need to keep other variables in control to create groups for which values can be compared.

As ability is undefinable, the study can be undertaken only by considering variables that can quantify & be a proxy to ability.