The Evaluation of Immigrant Credentials Dietz, Esses, Joshi, Bonnet-Abu Ayyouh Issue: What role do three factors play in the evaluation of educational credentials: 1. Immigrant status 2. Accreditation (or not) of foreign credentials 3. Race (black vs white, proxied by name of applicant) Very important issue; difficult to address using secondary data analysis � Set up experiment � In addition, they ask whether racial biases affect credentials evaluation � (secondary data analysis)
The Experiment “In basket” exercise conducted with university students � Assumed role of V.P. of H.R � Based on info provided, they provide evaluation re: filling of sales executive job � Outcomes: » Evaluation of education credentials using two 7 point scales: re “suitability of education” and “quality of education” (averaged) Treatment: (experimental conditions) » Six levels that varied the three factors among job applicants: immigrant status, race (black/white) and accreditation status Participants: » 405 university students in business and psychology, some with employment/ supervising experience; randomly assigned to six treatment levels Set up a number of hypotheses
Key results: 1. No sign of blatant immigrant or racial bias re: the evaluation of the educational credentials: e.g. » No sig. difference in evaluation of credential between - Qualified Canadians and Qualified Immigrants (with accreditation) - Qualified White Canadians and Qualified Black Canadians - Qualified White Immigrants (accredited) and Qualified Black Immigrants (accredited)
2. Signs of more subtle bias People maintain remnants of biases against ethnic groups/immigrants, but they are suppressed by anti-prejudice social norms However, when a seemingly non-bias based justification (i.e., lack of accreditation) is available, these biases are expressed e.g. - Qualified Immigrants (not accredited) received lower evaluation than all others (not surprising) - Qualified Black Immigrants (no accreditation) received lower evaluation than Qualified White Immigrants (no accreditation) - Latter finding marginally significant
3. If foreign credentials were recognized (accredited) then they are not discounted relative to Canadian credentials 4. The more evaluators harbour biases against immigrants, the more negatively they will evaluated credentials of immigrants without accreditation relative to all others
Comments 1. Thoughtful paper…. attempted to account for potential effects of numerous characteristics; assess effects of subtle prejudices 2. Are these results generalizable to decision makers in real companies?? Important if results to influence policy » They argue “yes” in paper – Some participants with business experience – Findings consistent with previous research – Realistic experimental context (in-basket) » But biases likely underestimated in this experimental context because: (a) A generally liberal group of participants (young university students) (b) People doing an in-basket test know they are being monitored/ evaluated More likely to act in “politically correct” manner � » Could think of results as producing lower bound re: biases
3. Could attempt to administer this same experiment to population of actual decision makers 4. Regarding call for accreditation of foreign credentials. This would be a very reasonable and useful step » Currently all university degrees treated as equal » Been called for a number of times » This experiment provides more support for such an initiative
5. Methodology a) Not clear in paper why the need to control for background characteristics of participants in a regression; they are randomly assigned to the six treatment levels. Why not simply compare means? b) Not clear in paper how the reported means for different “treatment” groups are determined? Raw Means? Predicted values from regression? Is sample size sufficiently large to “see” difference in outcomes (i.e. biases) How large does the difference have to be to be statically significant? – What is a large or small difference in evaluation score? Step at statistically sig or not, no discussion of magnitude of effects
c) Choose other groups in addition to blacks (South Africa) as the ethnic group? Relatively few immigrants (6%) are black, and very few from South Africa… These results may not be generalizable to all ethnic groups: biases may differ re: blacks, Chinese, East Asia d) Accreditation of foreign credentials… what body was “employed” to accredit credentials… how realistic was the transmission of this key piece of info to participants?
“Why Do Recent Immigrants to Canada Struggle in the Labour Market: A Field � Experiment”… P. Oreopoulis » Experiment where outcomes is “call-back rate” for distributed resume’s » Finds higher levels of bias » Ethnic group (as determined by name) matters, as does having a foreign degree (compared to Canadian) » But difference between foreign and Canadian experience the major determinant of lower call-back rates Overall, a very thoughtful and novel paper. Fun to read and think about �
Evaluating the Source of Low Returns to Immigrants Foreign Experience and Credentials: Problems with Recognition or Reflection of Productivity? Are immigrants’ foreign experience and education undervalued, or is this lower � valuation a reflection of the productivity of these workers? Compute relative productivity of workers with (1) Canadian and foreign � experience, and (2) Canadian and foreign credentials. Compare these relative productivities with the wage gaps (relative wages) for the same groups Obviously an important and useful topic �
Comments Very ambitious project… pushes the methodology and the data to the limit � Comment on establishment (firm) based research � Assumption in model: » Proportion of workers for the establishment a a whole (e.g. % male/female) apply across all types of workers (e.g. by occupation, marital status, education) » Relative productivity of workers at firm level (e.g. males/females) is same across all other cells (e.g. by occupation, --- etc.) » If education and occupation variables used, could be a problem – E.g., shares of workers by education not likely to be the same across all occupation, by gender, by age – Relative productivity of workers by education level not likely to be same across low and high skilled occupation – Need to know the implications of these assumption – Can relax the constraints test some of these assumptions in final paper --- encourage this
Data Issues Significant issue… will require substantial sensitivity testing and comparisons � with other data sources Final regression equation straight forward; requires data at establishment level � (the WES) » Value added output » Capital stock » Shares of workers in various categories (e.g., immigrants with foreign credentials, immigrants with foreign experience, education levels, occupations, gender, etc.) Data challenges with all three types of data �
Value-added output Have variables � » Gross payroll » Gross operating expenditures » Gross revenues Could compute estimate of value-added output, but at least 50% of the values � are imputed (20% non-response, other values falling outside reasonable range Do have imputation flag; conduct some tests � Link WES establishments to Annual Survey of Manufacturers; output data far � superior Capital stock » Data do not exist at establishment/location/firm level » Forced to use proxies, methods of distributing industry capital stock to establishments (not firms)
Worker shares Need shares of workers in various groups at establishment level � » Estimates at establishment level only for occupation (6 categories) and gender » Need to go to worker data for age, education, immigrant status, marital status, etc. » But only 1 to 24 workers per establishment » Hence, perhaps can estimate shares at industry level, prorate to establishment in same way? » Will need to convince readers that the data are up to this task – Sensitivity tests – Comparisons with other data services » Good topic; very innovative approach; some interesting challenges
Recommend
More recommend
