1 discussion of schuh & stavins jon zinman dartmouth college 26 oct 2007

11
1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

Upload: leona-page

Post on 20-Jan-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

1

Discussion ofSchuh & Stavins

Jon Zinman

Dartmouth College

26 Oct 2007

Page 2: 1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

2

Acknowledgement

• FRBBOS great provider of public goods re: research on consumer payment choice– Conferences– AEA session

Page 3: 1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

3

Main Contribution

• “Beyond Demographics”– Adding payment chars reduces sig of demos– Adding payment chars dramatically increases

fit

• Can do more:– Show regression results with just demos– Tighten critique of what (little) we learn from

demo correls• High-frequency, not life-cycle, choice here• Economic and policy modeling: need primitives

Page 4: 1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

4

Tweaks to Existing Setup

Motivation:

• Paper is primarily interested in what’s driving declining check volume– Don’t we know that action is on intensive

margin? (85% of SCF hhs had checking account in 1995; 89% in 2004)

– Conclusion says that many hh’s have not reduced check use

• Sounds plausible, but what’s data source?

Page 5: 1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

5

Tweaks to Existing Setup

Econometric implementation: more outcomes?• Intensive margin: why is share right outcome

variable?– Skewed?– Denominator endogenous?– Try some specs with level or log, conditional on

income and other proxies for total spending?• Rejection (hazard out) on extensive margin?

– Any data on this?– Interesting re: response to low prob. events (fraud)– Interesting re: spending control on credit cards?

Page 6: 1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

6

Tweaks to Existing Setup

Econometric Implementation:• Reorganize/reclassify payment characteristics to

provide tighter links to model based on primitives?– Show survey questions– What’s in “cost”? (fixed, marginal, overdraft risk….?)– What’s in “convenience” (time cost? Portability?)– Where is acceptance?

Page 7: 1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

7

Tweaks to Existing Setup

Representativeness issue:

• Get “old” data from consulting/market research firms, and compare?

Page 8: 1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

8

Tweaks to Existing Setup

Exposition/interpretation

• Showing that perceived attributes correlate with choices sensible, but….

• Acknowledge possible justification bias

Page 9: 1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

9

Blue Sky for Future Surveys

Issues with existing survey data:• Too heavy on introspection (vs. revealed preference).

Worried about:– Noise– Justification bias

• Not enough latitude for heterogeneity in “payment chars”– over time (even at high frequencies)

• Credit revolving status• Overdraft risk

– across transaction types• Cash back• POS vs. billpay

Page 10: 1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

10

Blue Sky

Solutions for future surveys• Ask about recent transactions

– Whats not whys• Including acceptance

– Compare this elicitation method to subjective attribute perceptions: would be nice methodological contribution

• Could also tweak subjective questions to allow for heterogeneity in payment chars:– Do ever find [media] most [attribute/char]?– How often?

Page 11: 1 Discussion of Schuh & Stavins Jon Zinman Dartmouth College 26 Oct 2007

11

Blue Sky

Exercise for future surveys:

• What’s in the 70% unexplained variation?

• How can we collect data on these factors?