What role for economics in predictive analytics?

In the age of big data, Robert Lucas's (Nobel Laureate 1995) critique is again relevant. When we change the incentives of firms and consumers, they will change their behavior. Our insights from old data, even when N=all, no longer apply.

Transcript

1. What role for economics in predictive analytics? Mikl´ os Koren CEU MicroData miklos.koren.hu @korenmiklos Budapest BI Forum 2014
2. What role for economics in predictive analytics?
3. How would Amazon decide on a radical innovation? 1
4. Outline 1. How models help 2. Insights from economics 3. Lessons for implementation
5. 1. Models
6. 1. Models Models help think about changes that have never happened before. 3
7. More car owners switch insurers during the November campaign (Kiss, 2014) 4
8. Is it cheaper for them, or are they more attentive? Structural e↵ects on switching costs vs inattention Andr´ as Kiss (CEU) Salience and switching Nov 7, 2014 44 / 56 5
9. 2. Economics
10. 2. Economics Economics helps business, economic and social decisions with two key ideas.
11. 2. Economics Economics helps business, economic and social decisions with two key ideas. 2a. People respond to incentives If you succeed in changing people’s behavior, insights from old data no longer apply.
12. 2. Economics Economics helps business, economic and social decisions with two key ideas. 2a. People respond to incentives If you succeed in changing people’s behavior, insights from old data no longer apply. Robert Lucas (Nobel Laureate 1995)
13. 2. Economics Economics helps business, economic and social decisions with two key ideas. 2a. People respond to incentives If you succeed in changing people’s behavior, insights from old data no longer apply. Robert Lucas (Nobel Laureate 1995) 2b. Systems adapt Not all great ideas scale.
14. 2. Economics Economics helps business, economic and social decisions with two key ideas. 2a. People respond to incentives If you succeed in changing people’s behavior, insights from old data no longer apply. Robert Lucas (Nobel Laureate 1995) 2b. Systems adapt Not all great ideas scale. (equilibrium feedback effects)
15. Side effects of PROGRESA: non eligibles also gain (Angelucci and de Giorgi, 2009) 7
16. Predictive tools change our behavior 8
17. 3. Implementation
18. 3. Implementation Keep it simple. (Atul Gawande: Checklist Manifesto) Look out for the incentives of people making the change.
19. Why don’t Pakistani football makers adopt a simple innovation? (Atkin et al. 2014) Figure 1: “Buckyball” Design Figure shows the standard “buckyball” design. It combines 20 hexagons and 12 pentagons 10
20. Better arrangement of pentagons would reduce waste by a third Figure 11: Cutting Pattern for “O↵set” Four-Pentagon Die                             39.5      54     5       Figure displays the cutting pattern for the four-panel o↵set die. 11
21. Conclusions
22. Conclusions 1. Models help evaluate counterfactual decisions. 2. Economics (incentives work, systems view) improves business and social decisions. 3. Social sciences (economics, sociology, psychlogy) help implementation.
23. Conclusions 1. Models help evaluate counterfactual decisions. 2. Economics (incentives work, systems view) improves business and social decisions. 3. Social sciences (economics, sociology, psychlogy) help implementation. Even Amazon needs economists! 12