Brooklyn vs. Philadelphia
Condo / co-op housing values · ZIP 11201 versus ZIP 19107 · Independent model selection and October 2026 forecasts.
Brooklyn
Philadelphia
Philadelphia · 19107
Lower backtest MAE relative to September market value (0.170% vs. Brooklyn 0.285%).
Indexed condo/co-op values
January 2020 = 100 · normalized to compare changes, not price levels.
Source: user-transcribed monthly Zillow ZHVI condo/co-op series. This chart is computed directly from the 81 supplied observations per market.
Backtest leaderboard
| # | Model | MAE | RMSE |
|---|---|---|---|
| 1 | Holt damped trend | $3,718 | $5,739 |
| 2 | Last value baseline | $8,993 | $14,467 |
| 3 | Drift | $10,056 | $14,483 |
| 4 | Linear trend (36m) | $99,990 | $116,737 |
| 5 | Seasonal naive (12m) | $111,923 | $152,773 |
Backtest leaderboard
| # | Model | MAE | RMSE |
|---|---|---|---|
| 1 | Holt damped trend | $439 | $567 |
| 2 | Last value baseline | $702 | $938 |
| 3 | Drift | $805 | $1,071 |
| 4 | Seasonal naive (12m) | $6,757 | $8,368 |
| 5 | Linear trend (36m) | $6,995 | $8,298 |
October 2026 forecasts · uncertainty
| Measure | Brooklyn | Philadelphia |
|---|---|---|
| September 2026 observed | $1,306,042 | $257,791 |
| October 2026 forecast | $1,308,132 | $257,048 |
| Empirical residual 80% range | $1,303,929–$1,312,637 | $256,355–$257,687 |
| Monthly projected change | +0.16% | -0.29% |
| Rolling backtests | 44 | 44 |
The empirical range uses the 10th and 90th percentiles of signed rolling forecast residuals. It is not a calibrated predictive confidence interval.
Methodology & QA
Both series cover January 2020 through September 2026 with 81 unique month-start dates, no missing prices, no missing months, no nonpositive values, and no duplicate dates. Five models were evaluated over the same 44 expanding-window one-month-ahead forecasts; each city selected the model with the lowest mean absolute error. The initial training window contains 37 observations; validation targets run from February 2023 through September 2026. Selection and error reporting share these validation windows, so these results are not an independent post-selection test. Models are re-fitted to all available data to forecast October 2026.
Interpretation: Brooklyn leads in projected home value; Philadelphia leads in the scale-adjusted MAE proxy. Relative MAE here is backtest MAE divided by the latest observed market value, not MAPE. No causal inference or investment advice is implied.