Research Journal

“The economic returns to schooling for Italian men. An evaluation based on instrumental variables”

Q1. What is the research question that the authors are interested in estimating?

The Organization for Economic Cooperation and Development (OECD) is composed by 37 world’s largest economies. Among these countries, Italy has a much lower upper secondary education attainment than the average of the OECD average. The researchers wants to find out the return of education on male household heads. The data used is a repeated cross-sectional data from the 1993 to 1995 waves of the BI survey.

Q2. What is the OLS regression estimating equation to answer the research question above that a novice Econometrician would run (that you know we should not run as it could be biased)? (1-line mathematical equation, though you can explain the variables in words.

  • The following OLS regression equation is what the researchers have proposed to use:

    • \[ ln(W) = X_i'\alpha + \beta S_i + u_i \]

      • “W” is the real hourly wage, “S” is years of schooling, “X” is a vector of observed attributes and “u” is a normally distributed error term with zero mean and finite variance.

Q3. Why would running the above naive OLS specification cause bias?

  • Ordinary last squares estimates of the return to education can’t accurate captures the relationship between earning and school attainment because either measurement errors in the schooling variables or because the error term includes other unobserved variables which can lead to omitted variable bias. For example, student’s individual ability that are correlated with total years of schooling they get as well as their future income. Thus, to address this type of omitted variable bias, instrumental variable method needs to applied.

Q4) What instrument (Z) can you use to get around and get true causal effects?

  • The first “Z” variable introduced is called D51, which is a dummy variable equal to 1 for individuals born since 1951, 0 if otherwise. Such an variable is aimed to pick the effect of 1969 educational reform in Italy which liberalized access to higher education. “This exogenous event has influenced the decision to pursue higher education by the cohorts of individuals born since 1951” (Brunello & Miniaci 8).

  • The additional “Z” variable introduced as instruments are a set of variables that are aimed to measure family backgrounds, including both the highest completed educational level and the occupation held by the parents of the interviewed household head.

  • Please give the first-stage and second-stage regression equations, and explain them.

    • First stage:

      • \[ \widehat{School_i} = \lambda_o + \lambda_1Edu_p + \lambda_2Job_p + \lambda_3D51 + \epsilon \]

        • λEdu_p : This variable is one of the two instrument variables introduced to capture the effect of family background. It measures the highest education attainment of the individual surveyed in our study. Intuitively, having parents who have college degrees would increased the likelihood of their kids earning a college degree. The authors of this study argues that such an influence only affects the wage of the surveyed individuals through affect their total years of schooling.

        • λJob_p: This variable is the second instrument variable included in capturing family background. It measures the occupation of the surveyed individual’s parents. The two authors of this studies argues that depending on what type of occupation the parents have, they might influence the parents’ decision in pushing their children to go to college as well.

        • λD51: This is the dummy variable that captures the educational reform in 1969 in Italy that widen the opportunity of obtaining college education to more individuals born after 1951. This variable is irrelevant to the wage level except if the individual attained college and obtained a college degree due to the governmental policy that made college entrance easier for students.

          • In 1969, individuals who were born in 1951 would turn 18 which is the standard age to go to college across all OECD countries.

    • Second stage:

      • \[ ln(W) = X_i'\alpha + \beta \widehat {School_i} + u_i \]

  • Potential violation of exogenetiy:

    • Although I don’t think exogeneity can be violated in our regression. However, I do want to point out that the researchers should specify what does the reform change to allow more students to go to college? Does the reform simply lowered the entrance standards or did it provide more resources to high school and middle schools so that students can achieve better scores. If the former is the case, then we might not see a substancial increase in wage of an individual due to the quality of a college degree decreased.