Issue in Brie
Such a pattern would work in the opposite directionfrom the cohort eect by raising the cognitive func-tion level of the respondents in older age groups.In light of these problems, it is useful to analyzedata that follow individuals longitudinally. The
Healthand Retirement Study
(HRS) is an excellent source forsuch analysis of cognitive variables.
Since it beganin 1992, the HRS has surveyed every two years anationally representative sample of more than 22,000Americans over the age of 50.
Our study examined several HRS questionsinvolving tasks measuring cognitive function. Forexample, we studied an immediate word recall task inwhich the interviewer reads a list of 10 simple nouns,and the respondent is immediately asked to recall asmany of them as possible, in any order. At age 51, theaverage performance is 6.2 words out of 10. By age90, the average (controlled) performance is 3.0 wordsout of 10.We analyzed the responses to the HRS questionsusing two parallel tracks. First, we undertook a“naive” analysis that simply plots mean performanceby age, ignoring the potential role of cohort and selec-tion eects. Second, we conducted a “controlled”analysis that traces out the performance trajectory us-ing
intra-individual dierences. Figure 3 showsthe results for the word recall task; results for otherquestions showed a similar pattern.The HRS results paint a clear picture of declin-ing cognitive function with age. They also suggestthat selection eects may be more important thancohort eects. Since older cohorts tend to have fewer
educational advantages, cohort eects are predicted tocause the naive proles to fall more steeply than thecontrol proles. Selection eects, in contrast, shouldcause the naive proles to fall less sharply than thecontrol proles, since the individuals with the poorestcognitive function tend to exit the sample. Selectionbias seems to be more important in the HRS data,since our controlled proles are consistently steeperthan our naive proles.
Financial Performance andAge: The Inverse U-Shape
This section summarizes evidence from our full studyshowing that the prices people pay in dierent credittransactions vary by age, exhibiting an inverse U-shaped curve pattern. The main example highlightedbelow covers credit card balance transfer oers.Other transactions we examined include home equityloans and lines of credit, car loans, mortgages, andseveral other types of credit card transactions. In eachcase, we conducted a regression analysis that identi-es age eects and controls for observable factorsthat might explain the patterns by age.
The resultsconsistently conrmed an inverse U-shaped pattern.
“Eureka” Moments: Optimizing the Useof Credit Card Balance Transfers
Credit card holders frequently receive oers to trans-fer account balances from an existing card to a newcard charging a substantially lower annual percent-age rate (APR) for an initial period, from six monthsto a year or more (a “teaser” rate). The catch is thatall payments on the new card are applied rst to thetransferred balance and are applied to new purchases(which are subject to a higher APR) only after thetransferred balance has been paid o.The optimal strategy for the cardholder during theteaser-rate period, then, is to make all new purchaseson the
credit card and none on the new card untilall the transferred balances have been paid o. Inour analysis, we categorize cardholders by the speedwith which they converge on this optimal strategy.
F . R I W R T HRS, -
Authors’ calculations from University of Michigan,
Health and Retirement Study (HRS)