- Saint Louis County housing market and tax appraisal errors from 2000 through 2014
- Single-family housing
- Price appreciation over-hyped?
- The county's legal compliance
- Room for improvement
June 14, 2016
A Quick Overview
Legal and Professional Requirements
- Saint Louis County's legal challenges (State of Missouri)
- Standards of the International Association of Assessing Officers (IAAO)
- State of Missouri State Tax Commission Assessor’s Manual Chapter IV (STC)
- Appraisals for tax assessment
- Market value arms-length transactions
Previous Findings
- Saint Louis non-compliance until 2007
- Saint Louis County and City
- Older units
- Every other county
Data Collection
- Saint Louis County tax assessment rolls
- Saint Louis County GIS cds
- Their judgement and COV
- Personal cleaning
Saint Louis County Transactions
Transactions by (My) Grade
The Saint Louis Housing Market
- Is heating up
- Caution about median values
- Appreciation and constant quality measures
- Case-Shiller, Census Bureau, Federal Housing Finance Agency, Zillow? - and tax assessment
- The hedonic approach
The Hedonic Approach (Quickly)
- Rationale: bundle of homogeneous services
- Estimates shadow prices
- Consider a semi-log hedonic model
- \(lnP_{th} = \sum_{c=1}^C \beta_{c}z_{cth}+ \sum_{t=1}^T \delta_{t}d_{th}+ \varepsilon_{th}\)
- Notice: peaks, seasonality, and general fluctuations
Single-Family Housing Prices
There Are Limitations to the Hedonic Approach
- We can chat about it.
- Notice the source of the difference
Single-Family Housing Quality
Appradial Errors
- Difference of the logs: \(ln(Appraisal) - ln(Price)\)
- Similar to ratio
- Over and under appraisals are equidistant from zero
- Missing hedonic sales dropped
- Errors greater than 3 * IQR dropped (IAAO - extreme)
- X and V don't make a difference
Appraisal Error Distribution
Should be Fair
- County assesses every odd year
- Expected appraisals patterns over time
- Should start at zero and drift
- Start with a smoothed estimate (not hedonic)
Appraisal Errors Over Time
Appraisal Errors Over Time (Ratio)
Appraisal Errors Should be Random
- Unpredictable
- Given available data (not making house visits)
- Look at error by logged appraised value
Appraisal Errors by Appraisal Value
Doesn't Look Exciting But!
- Consider distressed sales
- Less predictable
- County's sales validity code 5 and Z
- Start with a constant quality measure
Distressed Sales Perform Poorly
Appraisal Distressed "Errors"
Appraisal "Errors" by Appraisal Value
Bring It All Together
- Using County data to predict County errors
- One regression (OLS) with three parts
- Unit characteristics
- Month
- Zip code
Predicted Appraisal Errors
| Model 1 | ||
|---|---|---|
| Intercept | -0.3494 (0.0172)*** | |
| Age in Decades | 0.0055 (0.0008)*** | |
| Age in Decades ^ 2 | -0.0015 (0.0001)*** | |
| ln(Lot Size in Acres) | -0.0011 (0.0009) | |
| ln(Living Area in Sq.Ft.) | 0.0252 (0.0022)*** | |
| Stories | 0.0214 (0.0016)*** | |
| Bedrooms | -0.0060 (0.0008)*** | |
| Full Bath | -0.0041 (0.0009)*** | |
| FEMA 100 Year | 0.0059 (0.0031) | |
| R2 | 0.3146 | |
| Adj. R2 | 0.3135 | |
| Num. obs. | 153486 | |
| RMSE | 0.1535 | |
| p < 0.001, p < 0.01, p < 0.05 | ||
Predicted Appraisal Errors over Time
Predicted Appraisal Errors and "Errors"
Predicted Appraisal Errors over Space
Conclusion
- Main: the County is legal, but has room for improvement with existing data
- Housing market hype? Need more data!
Thank You
Questions?