You can't use statistics without a model. The problem is that you have a shitty model and it's annoying when you keep parroting it like you're so logical. Your model as inferred from this comment is that there's some way to quantify the current applicants, and that applicants below a certain value are always rejected. The remaining applicants are "above the bar" and could be hired. The current "above the bar" group is not 50/50, therefore changing it means accepting people below the bar.
The problem is that none of your assumptions hold water. Technical interviews are about as accurate as flipping a coin. Mundane factors like whether or not the interview is after lunch have a major impact on pass rates. Performance on the interview has little to no predictive value for future job performance.
You can't brush this off as social ignorance, this is simply ignorance of the facts.
The problem is that none of your assumptions hold water. Technical interviews are about as accurate as flipping a coin. Mundane factors like whether or not the interview is after lunch have a major impact on pass rates. Performance on the interview has little to no predictive value for future job performance.
You can't brush this off as social ignorance, this is simply ignorance of the facts.