On Wages and Hygiene: Surfacing Bad Management in Public Data

On Wages and Hygiene: Surfacing Bad Management in Public Data

Do restaurants with poor health ratings also have wage violations? Bad health and inadequate wage payments for employees might be caused by a common factor - bad management. But that's not easy to quantify.  To get answers, we merged Florida restaurant inspection data with federal wage compliance data. Ultimately we concluded that the likelihood of a restaurant improperly paying its employees is three times larger if that restaurant also suffers from food violations.

How we merged restaurant and wage data

As it turns out, a lot of public data suffers from its own failures in management and hygiene. Wrestling good information out of messy tables is often a rough-and-tumble process.

Because the two datasets do not share a common ID, such as an employer identification number, we used a combination of address geocoding and string distance comparisons on company names to find matches. We used the Census' TIGER geocoder to get the geolocation (latitude and longitude) for all establishments included in the two datasets. When multiple establishment names were associated with the same location, i.e. restaurants in a mall, we used string similarity to choose the best match.

We were able to geocode around 70% of addresses within both the restaurant and wage inspections. Out of the initial 36,432 restaurants within the food inspections, we found 440 that also showed up within wage inspections. (To the best of our knowledge, there were no biases in geocoding; the mean number of health or wage violations for restaurants that we could not geocode were not statistically different from the mean of those we could.)

A run-down of our analysis 

Restaurant inspections are conducted by city, county, or state authorities and check for compliance to food-related sanitary standards. Most cities require that restaurants are inspected at least once per year. The wage and hourly compliance inspections are carried out by the Department of Labor (DOL) and check for violations of the Fair Labor Standards Act – a federal law requiring employers to comply with minimum wage, overtime pay, payroll and hourly records, and limits on certain types of underage labor. 

The wage inspections are primarily initiated by complaints from workers, and are guaranteed to be confidential. In contrast to the universal coverage of hygiene inspections, only a very small share of establishments (around 2% of restaurants in Florida since 2011) end up being inspected by the Labor Department.

Exploratory analysis of the matched data suggests that if a restaurant has one kind of violation, it is more likely to have others, meaning it is run poorly on multiple fronts. First, we found that restaurants with poor health scores are more likely to have received a wage inspection, though not necessarily a violation. This can be seen in Figure 1. 

We divided Florida restaurants into quartiles based on their health inspection score (total number of health violations) and averaged the match rate with the wage and hourly inspection data in each of the bins. A match in the wage and hourly inspection dataset reflects the fact that a restaurant has been inspected for violations by the DOL (but not necessarily sanctioned), most likely as a result of a complaint from employees. 

As Figure 1 shows, the probability of receiving a wage inspection is three times larger for restaurants with a large number of food violations (i.e in the bottom 25%) than for restaurants with a small number of violations (in the top 25%).

Wage Food Inspections Chart 1

A wage inspection is very likely to lead to financial costs. On average, 80% of the restaurants that do receive a wage inspection have at least one violation, resulting in an average fine of $300 and back-wages of $5,000.

Next, we investigated whether the health scores of restaurants that have already received a wage inspection influence the likelihood of those restaurants also receiving a wage violation. This is essentially a measure of the probability that a given inspection will result in a violation. We again binned the restaurant hygiene score data into quartiles and computed the proportion of restaurant inspections that lead to at least one violation. We did not observe any relationship with health scores in this regard (Figure 2). That is not surprising given that the restaurants in the wage inspection dataset are already members of a high-risk subset of restaurants. From this we concluded that the rate of wage inspections, not the likelihood of wage violations, correlates to health violations.

Wage Food Inspections Chart 2

Combining these two pieces of analysis suggests that the likelihood of being identified with at least one wage violation strongly increases with bad food inspection scores. Figure 3 shows that restaurants with bad hygiene scores (i.e. in the bottom 25%) have nearly four times more chance of being identified with wage violations as restaurants with good scores (i.e. in the top 25%). Bad food inspection scores often result in fines and payment of back-wages.

Wage Food Inspections Chart 3

We will follow up with a statistical analysis of this relationship and a more nuanced view into the nature of health and wage violations driving the results. Stay tuned!