Analysis by independent researchers revealed the Krueger-Card report, which was based on a phone survey in which fast food restaurant managers and assistant managers were asked about their staff size, to be deeply flawed. The Employment Policy Institute analyzed the phone survey results against actual payroll data from the restaurants and concluded that "the data set used in the New Jersey study bears no relation to numbers drawn from payroll records of the restaurants the New Jersey study claims to cover."
According to the Krueger-Card data set, a Burger King in New Jersey went from zero to 29 full-time workers after the minimum wage hike between February and November of 1992, while a Wendy's in Pennsylvania reduced its workforce from 30 to zero full-time workers during the same nine-month period. Truly radical — indeed, implausible — shifts in a business's employment strategy. When compared to actual employment records, the EPI analysis found that in one third of the restaurants surveyed, Krueger-Card even got the direction of employment change (whether staff was cut of added) wrong.
A subsequent analysis published by the National Bureau of Economic Research based on payroll records of fast-food restaurants during the same period revealed that Garden State workers experienced a 4.6 percent decrease in employment after the minimum wage hike compared to the Pennsylvania control group. In other words, they confirmed the commonsense economic principle that when something costs more, people can afford less of it. Or in the case of a minimum-wage hike, when workers cost more to employ, businesses can afford to hire fewer workers.
As someone who handles data on a daily basis, it's pretty clear that either Kreuger didn't know how to handle a survey like this or he knew that if he scrubbed the data properly he would have had a result that was contrary to his initial bias. Why did he include the outliers that were clearly wrong and skewing the data? Why did he rely on the memories of assistant managers at fast food restaurants? Does he really think they can remember how many people they hired or fired over any stretch of time? The fact that one third of respondents even got the direction of employment change wrong when compared to actual payroll data is pretty horrible. So I just have to wonder, did he know his study was bogus or was he such a data novice that he didn't realize its flaws?