Here is also a full copy of the #stunting conference report for those who are interested in the details and in references to the most recent research currently being conducted: goo.gl/LcV1S2
2/10
Context: 155 million children around the world are stunted. Stunted children are too short for their age because they experience slower growth because of poor nutrition and high rates of infections. #Stunting is concentrated in Africa and South and Southeast Asia today.
3/10
Context (cont.): Stunted children perform poorly on cognitive tests, receive fewer years of schooling, have lower income as adults and are more likely to live in poverty. Eradicating stunting is a very important policy goal and is part of SDG Goal 2. un.org/sustainabledev…
4/10
There were four main lessons from the conference about how to move #stunting research and policy forward to deal with this important global challenge.
5/10
Lesson 1: #Stunting is not a recent phenomenon. Some currently developed countries had high levels of #stunting in the past (c.f. Japan in the graph below) and it took decades of strong improvements in economic development and health to eradicate #stunting.
6/10
Lesson 2: We need more research on interventions in adolescence. Are the effects of child #stunting on health and development permanent, or could adolescence be a critical window for interventions to mitigate the effects of being stunted? doi.org/10.3945/ajcn.1…
7/10
Lesson 3: Age matters. For most children, growth faltering occurs between 6 and 24 months and remains flat between 24 and 60 months. Associations between #stunting and wealth, income and parental education are stronger for children over 24 months. doi.org/10.1371/journa…
8/10
Lesson 4: Interdisciplinary research and perspectives are vitally important. The causes of #stunting are multidimensional. To construct complete models or design experiments researchers need to consider medical, economic, social and cultural drivers of stunting.
I present 10 examples of collider bias in economic history research.
I focus on cases where authors overcome or mitigate the bias. The cases illustrate how to diagnose bias and come up with creative solutions to manage it.
And now on to some examples:
2/n
Example 1: Sample-selection bias
Bodenhorn et al. 2017 argue that since we only observe historical heights for men who enlisted in the army or committed crimes, changes in current labour market conditions will bias inferences from trends in height. This is collider bias!