{
  "id": "2001-septemberoctober-ratio-study",
  "slug": "2001-septemberoctober-ratio-study",
  "url": "https://dwillis.github.io/uplink/article/2001-septemberoctober-ratio-study.html",
  "year": 2001,
  "month": "September/October",
  "monthNum": 9,
  "headline": "Ratio study",
  "kicker": "PROPERTY ASSESSMENTS",
  "author_name": "Mark Houser",
  "author_title": "Reporter",
  "affiliation": "Pittsburgh Tribune-Review",
  "summary": "The Pittsburgh Tribune-Review used computer-assisted ratio studies to evaluate whether Allegheny County's $24 million property reassessment was accurate, finding significant flaws including overassessment of affordable neighborhoods and underassessment of expensive homes, contrary to the county assessor's own positive findings.",
  "keywords": [
    "assessment",
    "properties",
    "reassessment"
  ],
  "topics": [
    "Business & Economy",
    "Investigative Methods",
    "Data & Databases"
  ],
  "technologies": [
    "Excel"
  ],
  "full_text": "When its ancient, inaccurate property assessments sparked several lawsuits and a judge ordered things put right, Allegheny County coughed up $24 million to an Ohio firm to reassess the nearly 600,000 real estate parcels in Pittsburgh and its suburbs. Simply put, a ratio study compares assessed values with the actual prices properties sell for in the most recent transactions, to see how well assessors are predicting the market. But after the results came in, homeowner hordes practically stormed the ap- peals office, protesting that their new val- ues were off by a mile. Were they right, or just complaining? Did taxpayers get their money's worth from the reassessment? Ratio study The Pittsburgh Tribune-Review set out to answer those questions this spring. To do so, we used computer-assisted report- ing and a standard analysis tool of the as- sessment industry, called a ratio study. Simply put, a ratio study compares assessed values with the actual prices prop- erties sell for in the most recent transac- tions, to see how well assessors are predict- ing the market. Ratio studies involve thou- sands of calculations, property by prop- erty, which means they are an ideal job for your whirring desktop assistant. The county assessor's office conducted its own ratio study, using old and filtered sales data provided by the contractor, Sa- bre Systems and Service. Big surprise - it found a few isolated problems, but over- all, good work. Gold star sticker, guys. Flaws in assessment But the Trib, using industry standards the county didn't follow, found serious flaws in the reassessment. In the city of Pittsburgh, one of every five properties sold for more than double, or less than half, the new as- sessed value in 2000. Throughout the county, the Trib found the least expensive neighborhoods were frequently overassessed, while pricier homes were routinely assessed at hundreds of thousands of dollars less than what they sold for. To do a ratio study, you need a database of property tax rolls. The information must include: 1. The assessed value of each property. 2. The date of the property's most recent sale, and the price. 3. Information about the sale to deter- mine whether it was an \"arm's length\" trans- action on the open market. Weeding out The last is crucial. Many properties sell at prices that don't reflect their true value. An elderly mother sells her home to her son for $1, sometimes called a \"love and affec- tion\" sale. An owner fails to pay real estate taxes, and the property is seized and sold at a sheriff's sale at a bargain basement price. A seller includes an 18-foot fishing boat in the deal. Sales like this can't be used to establish the market value of homes in a neighbor- hood, SO they have to be weeded out in a ratio study. The computerized tax rolls we used contained a code describing the type of sale; we used this to select only county-qualified arm's length transactions of $10,000 or more for our study. Because the new assessments are supposed to reflect a property's value in 2000, the Trib looked at sales from 2000. (The assessors' own study used sales from 1996-99, the very sales Sabre used to come up with its values in the first place.) In the end, we narrowed the tax rolls down to an Excel spreadsheet of 16,502 properties that sold last year in Allegheny County, including 4,338 in Pittsburgh. The newspaper used two common mea- sures of accuracy in ratio studies: median sales ratio and coefficient of dispersion. Finding the median Median sales ratio shows if properties are typically underassessed, overassessed or as- sessed just right. A property's sales ratio is its assessed value divided by its sale price, typically expressed as a percentage. Less than 100 percent means assessors lowballed the value compared to what it sold for; more than 100 means the opposite. The Trib cal- culated sales ratios for every property in a school district (or Pittsburgh ward) and found the median, to find out whether homes were more likely to be over- or un- derassessed in that neighborhood. There are 43 school districts in Allegh- eny County. Schools get the bulk of prop- erty tax revenue, so they made ideal \"neigh- borhoods\" on which to base our measure- ments - small enough to isolate problem areas, big enough to have sufficient sales for statistically sound findings. To speed up find- ing the medians in each of these neighbor- hoods, I used an Excel Subtotals trick. The newspaper used two common measures of accuracy in ratio studies: median sales ratio and coefficient of dispersion. Experts I talked to from the Chicago-based International Association of Assessing Offic- ers said a good rule of thumb for using MSR is anything below 95 or above 105 is a problem. Our study found wealthy neighborhoods as low as 88, and three poor city wards above 120. Coefficient of dispersion Coefficient of dispersion is a much more difficult concept. Basically, it de- scribes how many percentage points each property's sales ratio is off the median sales ratio. In a non-cookie-cutter neighbor- hood, the national standard COD is 15 or less for houses, or 20 or less for a mix of residential and commercial properties. COD measures consistency, which makes it an important companion to the MSR when discussing assessment accu- racy. The city of Pittsburgh provides a perfect example. Some got the shaft Pittsburgh's MSR was a perfect 100, mean- ing properties were no more likely to be un- derassessed than overassessed. But its COD was 24, meaning the average property was either underassessed or overassessed by 24 percent. Translation: A lot of owners were getting big breaks. A lot were getting the shaft. There are formulas for calculating COD, but they are complex, and I didn't want to strip a brain gear or melt my hard drive. Luckily, I got help from an expert. Peter Davis, a Kansas Department of Revenue of- ficial and the guy who helped write the lat- est ratio study standards for the IAAO, let me use a free Visual Basic demo program the department has posted on its website: (www.ink.org/public/kdor/pud/pvdratio.html) (It comes bundled with instructions.) The program, which Davis invites any other interested reporter to use, calculates COD from an Excel spreadsheet of assessments and sale prices. It also figures a confidence interval, use- ful for neighborhoods with CODs close to the line. Davis also patiently provided much valu- able advice on constructing the study fairly. For example, before calculating COD it is standard industry practice to \"trim\" off properties with the most extreme sales ra- tios. Unlike MSR, COD can be affected sub- stantially by just a few bad assessments. The IAAO-approved standard method of deal- ing with this discards outlying ratios more than 1.5 times beyond the range of the middle 50 percent of the sample. It took some time on the spreadsheet, but it was worth it to keep the results as watertight as possible. To read the Trib's main story on its ratio study, go to http://www.triblive.com/ tribinfo/assess.html Q & A by Mark Houser You have an Excel sheet with thousands of homes and their sales ratios, grouped by their neighborhood. You want the median sales ratio for each neighborhood. How can you do this quickly and easily? Trick Microsoft. You can use the Subtotal function, even though it's not built to do medians. Go ahead and use Data/ Subtotals to find the sum of sales ratios at each change in neighborhood. Then, highlight the column with the subtotals, and pick Edit/Replace. Replace \"SUBTOTAL (9,\" with \"MEDIAN(\". Hoo-hah! Faked you out, Bill Gates. Once we had the numbers, we fleshed them out with details. We talked with homeowners on one street where homes sold for a quarter of a million and were assessed at half that. We talked the neighbor a few blocks away whose $45,000 house was taxed as if it were worth $63,100. We got a county overseer to admit that his assessors were murmuring privately about \"so many mis- takes it's unbelievable.\" Sabre Systems never responded to our re- peated calls and e-mails. Some county offi- cials criticized the Trib's methods. Others waved Trib clips as evidence of a major break- down. Since our analysis appeared, several other ratio studies have been done - one by the county controller, another by a consultant for county council, yet another by our cross- town competition. All of them have found serious accuracy problems with the new as- sessed values, contradicting the findings of our county executive. Meanwhile, 89,000 homeowners have appealed their new values, far more than the county ever expected. The vast majority are getting reductions. Allegheny County now expects to pay $20 million in refunds, more than double what it budgeted. And several county elected officials are calling for Sabre Systems to refund some of that $24 million. Mark Houser can be reached by e-mail at mhouser@tribweb.com.",
  "provenance": {
    "extracted_by": "pre-v2",
    "extracted_at": "",
    "verified": true
  },
  "similar": [
    {
      "slug": "2001-septemberoctober-calculating-bills",
      "score": 0.263
    },
    {
      "slug": "2000-june-exposing-flaws",
      "score": 0.173
    },
    {
      "slug": "2000-june-covering-tax-assessments",
      "score": 0.139
    },
    {
      "slug": "2002-november-december-analysis-finds-juries-lack-color",
      "score": 0.127
    },
    {
      "slug": "2004-july-august-property-data-analysis-finds-extra-tax-credits",
      "score": 0.113
    }
  ],
  "html_url": "https://dwillis.github.io/uplink/article/2001-septemberoctober-ratio-study.html",
  "json_url": "https://dwillis.github.io/uplink/article/2001-septemberoctober-ratio-study.json"
}