Alert! Alert! Respectful online discussion about a consequential issue.

From “Heckman curve” update: The data don’t seem to support the claim that human capital investments are most effective when targeted at younger ages. by Andrew.

Excellent example of civil discourse on a consequential but also highly disputed topic with grave empirical constraints.

Heckman is a magnificent economist (Nobel winner) and researcher. Simultaneously he is a passionate and ardent advocate for pre-k schooling. His arguments for pre-k have always seemed to me to be both motivated and normative. He wants it to be true that there is a large return and feels that there is a moral element (we need to look after the poorest) and therefore his arguments are structured to support the outcome he seeks rather than reflecting the full balance of the empirical record.

Which, from my reading, is far more nuanced, flawed, disputed, and context specific than is often acknowledged. Measures are flawed and incomplete, defined variantly, confounding variables uncontrolled, etc.

Heckman has made the claim that there is a declining returns curve - money invested early in a child's life has higher returns than money invested later.

Click to enlarge.

The research being discussed in this post contradicts the Heckman curve hypothesis. It finds that the existing body of evidence is oddly scattered but no power law.

Click to enlarge.

But then the conversation goes a lot of different directions. Which studies were included? Effect sizes? Do the studies invalidating the curve also call into question whether there is even a positive return? Why do some of the studies indicate such anomalously large returns not otherwise seen in social science?, etc.

Some quite interesting points are raised that are rarely addressed.

If you do multiple interventions, it is hard to disaggregate one from another. If a child was in pre-k and had a family intervention at 15, is any improvement, if there is any, due to pre-k, due to the teen intervention, or due to the combined pre-k with teen intervention?

The Heckman curve is indirectly a double advocacy position 1) Invest in pre-k and 2) Invest in pre-k at the expense of later interventions.

Interventions structured for the most marginalized children are not extendable to the average child in terms of potential beneficial consequences.

The discussion about the appropriate statistical treatment is different from the discussion about the experimental design - both are relevant but in different ways.

"It’s a general problem that ideas become fixed in social sciences with minimal and often highly unpersuasive evidence behind them."

A good discussion, a wide-ranging discussion, a respectful discussion on an important and consequential topic. A discussion that tends to reinforce that as eager as we are to do something effective and positive, empirically we still don't have the evidence we want and need to support our preferred solutions.