Presentation on theme: "Age, Women and Hiring: A Labor Market Experiment 2005 by Joanna Lahey Research supported by the MIT Schulz fund, NSF Doctoral Dissertation Grant # 238."— Presentation transcript:
Age, Women and Hiring: A Labor Market Experiment 2005 by Joanna Lahey Research supported by the MIT Schulz fund, NSF Doctoral Dissertation Grant # , NIA grant #T32-AG00186 and the NSF Graduate Research Fellowship for funding and support. Special thanks also to Lisa Bell, Joshua Campoverde, Faye Kasemset, Jennifer La’O, Dustin Rabideau, Vivian Si, Jessica A Thompson, and Yelena Yakunina for excellent research assistance.
“A doubling of the over-65 population by 2035 will substantially augment unified budget deficits and, accordingly, reduce federal saving unless actions are taken … policies promoting longer working life could ameliorate some of the potential demographic stresses…in choosing among … options, policymakers will need to pay careful attention to the likely economic effects.” --Alan Greenspan, August 27, 2004 FRB Opening Remarks
Motivation 1.“Fixing” Social Security: Get people to work past the normal retirement age −Previous work has focused only on Supply: How do we get older people to work? −I focus on Demand: Will firms hire older workers? 2.Workers live longer, are healthier, many indicate they would like to continue to work: Will they be able to?
Labor Demand through Hiring Note: term “discrimination” refers to differential treatment based on group characteristic Easiest way for firms to discriminate is through hiring –Difficult to prove discrimination in any individual case –Under ADEA, Compensation = being hired (possibly + lawyer fees) → not likely to sue
Two Questions 1.Is there differential hiring by age? 2.If so, what can explain this differential hiring? Answers (from my labor market experiment): 1.Yes, younger workers in both cities studied are more than 40% more likely to be called for an interview than older (similar to findings on race and gender). 2.No evidence for taste-based (animus) discrimination, suggestive evidence for statistical (group characteristics).
Measuring Discrimination Most economics studies do not measure “discrimination” but only a “residual” Most psychology studies measure discrimination directly, but –Subjects are undergraduates –Manager studies are generally hypothetical, often transparent
Why an Experiment? Compare job leavers/changers of different ages –Cannot control why left job Displaced older workers take longer to find a job than displaced younger workers (Diamond and Hausman 1984, Chan and Stevens 1999, 2001 etc.) –Could be discrimination… but… –Cannot rule out worker characteristics, such as higher reservation wages
Brief Literature Review Audit methodology –Began in the 1960s testing racial preferences in the housing market –Send paired testers “matched” on all characteristics except variable of interest Examples –Bendick et al. (1999) Consulting study. Looks at age in Fortune 500 firms. Not well controlled. –Bertrand and Mullainathan (2004) Sends resumes with “black” and “white” sounding names –Neumark et al. (1996) Examines differences in hiring for gender in the restaurant industry using both resumes and people
Experimental Setup Resume Audit (Correspondence review) –Sent resumes with different ages (as indicated by date of high school graduation) to job openings and measured the response rate by age. –40 want ads per week per city –10 “call-ins” per week per city –Sent two resumes per job –3996 firms total: about 4000 observations for each city
Setup Cont. Cities: Boston, MA and St. Petersburg, FL Looked only at women –Linda Jones and Mary E. Smith –Ages listed: 35, 45, 50, 55, 62 –Realistic work histories (last 10 years only) –Recent work experience in very entry-level positions
Setup Cont. Targeted only “entry-level” jobs –Anything that takes a year or less of post-high school education and experience combined –Jobs found through the Boston Globe, St. Petersburg Times or through calling firms in the yellow pages and asking if they are hiring (without replacement). –Examples: Clerical work, customer service, LPN, CNA, HHA, sales, laborer, waitress, nail tech, a/c duct cleaner etc.
Resume Creation Resumes representative of actual job seekers –Created using a computer program which randomly chose objectives, work histories, hobbies etc., from a list created from actual resumes –Addresses were chosen from middle class neighborhoods which according to realtor.com had wide variation in income and other demographic characteristics (eg. Somerville) –High schools chosen from small Midwestern college towns (DeKalb, IL and Ames City, IA)
Results for differential hiring Two kinds of responses measured –Positive or “call-back” –Interview Response rate overall of: 8% in MA and 10% in FL Response rates differ somewhat by method of application. Want ads are more likely to get positive responses than call-ins. Faxes slightly more likely than s. Results are similar with interview requests. Reminder: observable characteristics are roughly the same in proportion for each age (by design)
Tests for differential response by age Age is correlated negatively with the likelihood of being called back or being interviewed. Significant for the interview variable. –In MA: Every 10 years of age translates to an employee having to send out 1 more ad to receive a callback and 3 more ads to receive an interview –In FL: Every 10 years of age translates to an additional 3.5 ads for a callback and 4.5 more ads for an interview
Tests for differential response by age Split into two groups: Older and Younger, where Older = 50+ tests on positive and interview responses by younger and older –For positive, there is a difference of 1.5 pct pt, this is a difference of 20% in MA and 1.3 pct pt and 15% in Florida –For interview, these differences are 1.6 pct pt and 44% for MA and 1.8 pct pt and 43% for FL
Older/Younger Cont. Translation (on average): –In MA, a younger seeker needs to file 11 ads to get one callback; an older, 13. A younger seeker needs to file 19 ads for one interview request, an older, 27. –In FL, a younger seeker needs to file 9.6 ads to get one callback; an older, 11. A younger seeker needs to file 16.4 ads for one interview, an older, 23.
Welfare Implications from the BOE* Reminder: FL = demography of the future Hurt depends on occupation, location –To get one interview, # applications FL Average: 17 ads younger, 24 ads older FL Healthcare: 5.5 ads younger, 10 ads older MA Clerical: 32 ads younger, 72 ads older –# ads in paper each week varies by occupation Ex: In one week in FL: 32 LPN ads, 8 pre-school In general, half of ads are repeats from previous wk –Assume 7-10 interviews to land a job (optimistic) * Back of the Envelope
BOE and Welfare Cont. Occupational differences in time to job –Licensed Practical Nurse Younger: an offer in a week Older: 3 weeks –Clerical Younger: 3-5 weeks (all new ads), 6-10 weeks (half repeats) Older 7-10 weeks or weeks May be even longer, since within a 5 month period there are more repeat ads –=> Real welfare effects for low savings seekers
Is there Differential Hiring? Conclusion Yes, there is differential hiring by age. The effects are slightly stronger in MA than in FL and are stronger for interview requests than for callbacks. These results are similar to those found for race and gender.
Reasons for Differential Hiring: Background Different policy implications Taste-Based Discrimination (Becker 1971) –Employer –Employee (Co-worker) –Consumer Statistical (Arrow 1972) –See next slide
Statistical Discrimination: A Primer Judges an individual on “group” characteristics –Used when it is costly to gather information about an individual –Can be efficient for the employer (though not always the individual) –May have negative feed-back effects Not based on animus Groups have different underlying productivities
Statistical Discrimination, Cont. May or may not be legal Examples: –Want a high skilled worker, so employ a college graduate rather than a high school graduate (legal) –Want someone with technology skills, believe older workers do not have such skills → hire a younger worker (not legal) –Want someone who is energetic, believe smokers are less energetic → hire a non-smoker (legal only in some states)
Reasons tested in this study 1.Lack energy Sports vs. no sports on resume (tennis or racquetball) 2.Insurance and pension costs “I do not need insurance as I am already covered at home” This is the one item not found on real resumes surveyed. 3.Less flexible/adaptable “I am willing to embrace change.” “I am flexible.” 4. Health risks => absences Attendance award on previous job. 5. Knowledge and skills obsolescence –Computer degree from 1986, 1996, and 2002/ Also put in controls for volunteer work, typos etc.
Younger and Older get different responses to variables. –Statistical Discrimination: Older workers helped more than younger when information put on resume for both. Found for: Computer skills, but only in MA Volunteering, may imply up to date work skills or good health –In general, resume characteristics matter more for younger seekers Could be reverse statistical discrimination, more likely employers read younger resumes more carefully (=>ambiguous type of discrimination) Similar to Bertrand and Mullainathan’s finding about blacks Individually, not all resume characteristics matter. However, as a whole, they do matter.
Test for Employer Discrimination Look at companies with and without separate human resources departments –HR departments are trained in discrimination law => less likely to discriminate based on taste –HR departments have more experience in hiring => more likely to statistically discriminate Results: Firms with HR departments more likely to interview younger workers, but not significant => may statistically discriminate (need more research)
Employee Discrimination Assume that older workers less likely than younger to discriminate against older people Look at the employees in the place of work Match zipcode information from my dataset to place of work PUMA information on age demographics in the census (possibly too crude) Results: No effect of age of workforce on differential hiring by age
Consumer Discrimination Assume older customers discriminate less than younger customers do against older workers Look at demographics of consumer base Match company information from my dataset to age information from the census by zipcode –This is a good approximation to the group I want Also look just at sales and service occupations Results: No interaction between percentage of customers over the age of 50 and age of applicants –Areas with higher percentages of people over age 50 were more likely to call back or interview for both groups, but results are stronger for younger workers
Why is there Differential Hiring? Conclusion No evidence for taste-based discrimination –Probably not employer-based –Probably not consumer-based –Evidence not as strong against employee-based Some evidence for statistical discrimination –Might be lack of computer skills, might be fear that work skills are rusty (evidence only suggestive) –HR department evidence highly suggestive– need to finish collecting HR department information to be more certain
Conclusions There is differential hiring by age, at least in entry-level markets and against women This differential hiring may be due to statistical discrimination Any plan which requires older workers to continue working into later ages needs to consider these results