The mailer contained the graphic below, which claimed that the median tax bill was $6,962 in 2011 and in 2016 it was $10,516, an attention-grabbing 51% increase.
It also claimed that the part of the town budget funded by taxes had increased 15.4%. Essentially, this says that the total taxes collected (the levy) only grew by 15.4%, which seemed inconsistent with the 51% increase in the median tax bill.
Reading further, the following statement appeared:
It seemed very unlikely that the median and the average (mean) tax bills would match to the dollar.
I checked our tax bills, and they had not increased by anywhere near 51%, so I began to suspect the numbers in the mailer were wrong.
At the first Town Council meeting in June, I said during public comment that the numbers in the mailer were wrong. I was politely told they were not, and it was suggested that I could set up a meeting with the finance director and she would explain the numbers.
At this time, the position of finance director was in transition, and Councilman Schwager suggested I meet with the tax assessor instead. This made sense because the median tax bill graphic referenced the following footnote:
Subsequently I met with the tax assessor, and it was immediately clear that she had no idea where the numbers in the mailer had come from. She gave me copies of what she had provided to the finance director and town manager, starting with the following email to the IT director at New England Revaluation, the contractor who hosts the town's real estate database:
It was interesting that she had apparently been asked for this on short notice, in the week before the mailer was sent out. The response was:
I put together this YouTube video, which explains why the 51% number is wrong and calculates an estimate using the valuations the tax assessor had given me:
After this video was posted, I discovered a document posted on the town website called the "Tax Roll", which was the perfect data source for this analysis because:
Same house differences are a better measure of rates of change of valuations and tax bills than median values. If you use the median valuations, chances are you are comparing two different homes at the two points in time you are interested in.
Retailers use "same store sales" to estimate trends. This produces an estimate of the trend that is not biased by changes in the population of stores. Instead of comparing the mean or median sales across all stores, you calculate the percentage change for each store and take the median of those percentages. If possible, this is the best way to estimate trends in real estate valuations and taxes.
New construction does not have the same price distribution as older homes, so if you use medians, they are the medians of two different populations of homes, and if new homes are more expensive than existing homes, the median home value will go up, even if every property value stays the same.
Using the Access to Public Records Act (APRA), I requested the tax rolls for FY2016 and FY2011, which the town suppllied in .pdf format.
These were converted to text files (.txt), which were read with a python script (in the form of a Jupyter Notebook). Single family homes (state code 01) appearing in both tax rolls were matched by parcel ID (inner join on parcel id where state code = 01).
The matching process produced 3,818 records, each containing the FY2016 and FY2011 valuations, examptions, and tax bills. House by house, the percentage change between the FY2016 and FY2011 tax bills was computed. The median percentage change was reported as the median of these 3,818 individual house percent changes.
I put together this YouTube video, which explains why the 51% number is wrong and calculates an estimate using the median of the 3,818 same-home percentage increases:
Faced with irrefutable evidence that the 51% number was way off, the town council president agreed to meet to discuss revising the official estimate.
On December 20th, 2017, Anne Musella and I met for four hours with the council president and vice president, and a member of the community. Town officials presented their sources for the original numbers, and insisted that we recompute the percentage growth including condos. This was an obvious face saving gesture by the town officials, who wanted to be able to claim that their number and our number were both wrong.
We were able to make the required changes to the Jupyter notebook during the meeting and run the percentage increase number including single family homes and condos, which turned out to be 14.2%. The town agreed to publish this number.
The reason why there are two numbers is that one (14.2%) includes single family homes and condos (state codes 01 and 23) and the other (13.3%) does not.
The results for just single family homes are available in a spreadsheet that contains the valuations, the exemptions, the tax bills, and the percentage increase from FY2011 to FY2016 for each of the 3,818 single family homes that were matched.