This analysis is part of a series on the state of the short-term rental industry in Door County.

To dive into the data on the growth of the short-term rental industry in Door County, we submitted a public records request to the Door County Tourism Zone Commission, asking for a list of current room tax permit-holders.

For each permit, we requested that the data include the business type, the year the permit was issued, the property’s physical address, the business name (if applicable), and the owner’s name and address.

Tourism Zone Commission administrator Juliana Behme explained that deactivated permits are not included in the commission’s main data file, and the commission tracks those previous permit-holders separately. The same property could appear on both lists if the property had changed hands or stopped and then restarted renting, she said, because the commission would issue a new permit in that case. We requested a copy of the previous permit-holders list, as well.

We performed some basic data organization on the current permit-holders list. For instance, the commission’s permit numbers are split into four sections, separated by hyphens. Three of the four sections contain numerical codes corresponding to the rental’s municipality, business type and vacation rental management company or agent, if the rental has one. The fourth section is a permit ID number.

We copied the four sections of those permit numbers into separate columns, so that we could run queries on the data set based on municipality, business type and agent.

The data aren’t perfect: Some of the earlier permits from the late 2000s and early 2010s didn’t have dates listed for when the permit was issued, part of the reason we chose to start the bulk of our analysis in 2018. The fact that the permit ID numbers are sequential, meaning higher permit ID numbers were issued later, was helpful in making sure we had the data in the right order.

To determine the number of new short-term rental businesses, among current permit-holders, that have been created since 2018, we needed to cross-reference the lists of current and previous permit-holders.

We wanted to account for any possible cases in which a short-term rental had been originally created before 2018 but had either changed ownership or stopped and restarted renting in the years since. In such cases, we might not be able to tell that the rental had originally been created earlier without cross-referencing the two lists, because the current permit-holders list would only show the current permit with the more-recent date it was issued.

To perform that analysis, we used a program called csvdedupe, created by Chicago-based civic technology company DataMade, that allows for finding matches when comparing two spreadsheets.

We input two spreadsheets into the program: one containing all permits in the cottages, cabins and houses category that were issued since the start of 2018, and another containing all previous permits. We matched the two lists based on the rental’s street address and ZIP code and found a little more than 100 matches.

Because the previous permit-holders list does not include dates and in many cases does not include permit numbers, we couldn’t tell whether a rental that appeared on both lists was originally created before or after the start of 2018. We removed such rentals from our analysis for this question, treating them as if they were created before 2018.

The Tourism Zone Commission’s permit-holders data is manually maintained, meaning there could be typos or other data entry errors, such as slight differences in how addresses are spelled between the current and previous permit-holders lists.

Such issues could result in false negative matches, so we used a feature of the csvdedupe program that allows for training a machine learning model with human input to be able to account for variances in the data sets.

To create an animated map showing the location of each current short-term rental and the date its permit was issued, we used a geocoding tool to translate each rental’s address into latitude and longitude coordinates.

Because the data and the geocoding process aren’t perfect, it’s possible some addresses were imperfectly located on the map. We found a handful of addresses that were mistakenly placed outside Door County and manually corrected the latitude and longitude for those by looking them up on a map and copying down the coordinates.

Read the story:

This analysis was performed in reporting a story on the growth of the short-term rental industry in Door County. Read the story by clicking here.