Historical Property Value Analysis
A geospatial data science study of six years of residential property values, combining spatial comparisons, Python automation, sensitivity analysis, and reproducible reporting.
Why it matters
A local comparison group puts nearby appreciation in context. Sensitivity checks show when a few openings dominate the summary, helping readers distinguish descriptive patterns from conclusions that would require stronger causal evidence.
My contribution
I designed distance-based parcel groups and before-and-after appreciation comparisons, using surrounding parcels as a local benchmark for wider market changes. I later tested how case selection and extreme values influenced the retained summaries, and documented the assumptions and limits in reproducible reports.
Result
Across 18 business openings, changes in home-value growth were mixed. From before to after each opening, nearby homes’ growth improved relative to surrounding homes in nine cases and weakened in nine. A few large positive results made the overall average positive.
Tools
- Python
- ArcGIS Pro
- ArcPy
- pandas
- NumPy
- Matplotlib
- Difference-in-differences
- Sensitivity analysis
These are descriptive comparisons, not causal findings. The later review validates retained results and reports; it is not a fresh GIS analysis.
Project gallery
Project figures and report excerpts showing study coverage and descriptive category comparisons. 2 images. Select an image to view it full size.
Study design and analysis
Six years of residential values (2020–2025) support a comparison between parcels within 500 feet of business openings and those more than 500 and up to 2,000 feet away. The descriptive difference-in-differences measure tracks how the appreciation gap between the two groups changed from before to after each opening.
The original Python and ArcGIS workflow selected parcels and calculated appreciation; the later review checked those existing results, documented exclusions, and produced reproducible reports without rerunning the spatial analysis.
Category comparisons group results by business type. Sensitivity checks change which cases are included in the average; magnitude screening changes the cutoff used to flag results. These checks do not establish statistical significance.
Findings and interpretation
Of 63 business records, 37 openings fall within 2021–2024 and 18 have usable before-and-after comparisons. Each result describes one business opening and the groups of homes around it, not a single property. The study uses appraised values, not sale prices.
The results are nine positive and nine negative. Positive means nearby homes gained ground in annual value growth relative to surrounding homes after opening, compared with the pattern before it. Negative means they lost ground. Both groups can still rise in value. If growth slows more for nearby homes than for surrounding homes, the comparison is negative.
The average (mean) is +3.10 percentage points. Equal counts do not cancel out because the sizes of the results differ: large positive results can outweigh several smaller negative ones. The middle value (median) is +0.18.
Excluding three cases with unusually large average annual changes in one property group changes the average to −0.38 percentage points. This shows that a few cases strongly influence the summary; it does not prove those cases are wrong or that the remaining average measures a causal effect.
Recorded local checks include 33 passing validation and regression tests. The findings remain descriptive; the original parcel geodatabase is unavailable.
Technical details
Geospatial Data Science · Residential values, 2020–2025
Did residential property appreciation near selected business openings differ from appreciation in the surrounding area?
- Residential parcel values for 2020–2025 are appraised values, not sale prices. Nearby parcels lie within 500 feet of an opening and comparison parcels 500–2,000 feet away; the historical spatial assignments were not reverified during the later review.
- Each opening gets one measure: the post-opening nearby-minus-comparison appreciation gap minus the pre-opening gap, built from parcel-level annual changes.
- Eighteen of the 63 business records have a usable comparison; the other 45 are documented. Filtered subsets and an extreme-value exclusion show how much a few cases drive the summary. This is not a statistical significance test.
- There are no standard errors, confidence intervals, or parallel-trends tests, and parcels can fall near more than one opening. The original parcel geodatabase is unavailable, and the tested calculation code has not regenerated the historical GIS estimates.