Gera US Price Momentum (GUSPM) — Methodology

Data source

Step-by-step formula

1. 1

   Download the FHFA HPI datasets

   Fetch hpi_at_state.txt and hpi_at_metro.txt from https://www.fhfa.gov/data/hpi/datasets. No API key required. Files are plain tab-delimited text. State file: columns = state_code, year, quarter, hpi_value. Metro file: columns = metro_name (quoted), cbsa_code, year, quarter, hpi_value (or "-" for no data). Data goes back to 1975 for some states.

2. 2

   Parse and build time series per area

   For states: read each row, skip if hpi_value is not a float. For metros: skip rows where hpi_value is "-". Build a per-area array of {yr, qtr, hpi} objects sorted by (yr, qtr). Metros with fewer than 8 valid data points are excluded from the published dataset.

3. 3

   Identify the latest quarter

   The latest data point for each area is the maximum (yr, qtr) pair with a valid HPI value. As of this computation, the latest available quarter for all 51 states is 2026 Q1.

4. 4

   Compute pct1yr

   Find the data point with yr = latestYr − 1 and qtr = latestQtr (exactly 4 quarters prior). pct1yr = (latestHpi − hpiOneYrAgo) / hpiOneYrAgo × 100, rounded to 2 d.p. If no matching point exists, pct1yr = null ("insufficient data").

5. 5

   Compute pct5yr

   Find the data point with yr = latestYr − 5 and qtr = latestQtr. pct5yr = (latestHpi − hpiFiveYrAgo) / hpiFiveYrAgo × 100, rounded to 2 d.p. If no matching point, pct5yr = null.

6. 6

   Compute volatility

   For each of the last 10 quarterly periods, compute the 1-year HPI % change (current quarter vs same quarter 1 year prior). Take the population standard deviation of these 10 values (std dev = sqrt(variance)). If fewer than 4 such periods exist, volatility = null. Volatility band: "low" if < 1.0%, "medium" if 1.0–2.5%, "high" if > 2.5%.

7. 7

   Compute GUSPM raw score

   If pct1yr is null: guspm = null. Otherwise: guspm_raw = 0.5 × pct1yr + 0.3 × (pct5yr ?? 0) / 5 + 0.2 × max(0, 3 − (volatility ?? 3)). This blends 1yr momentum (50%), annualized 5yr momentum (30%), and an inverse-volatility stability bonus (20%). Areas with volatility ≥ 3% get zero stability bonus.

8. 8

   Normalize to 0–100

   guspmNorm = (guspm_raw − min_raw) / (max_raw − min_raw) × 100, rounded to 1 d.p. Normalization is across all 461 entries (states + metros) in the current dataset so that scores are comparable. The range updates each quarter when new FHFA data is published.

Pseudocode

# Input: hpi_at_state.txt, hpi_at_metro.txt
# Output: GUSPM_ENTRIES array (normalized 0–100)

for each area_file in [state_file, metro_file]:
  for each row in area_file:
    if hpi_value == "-" or not valid_float(hpi_value): skip
    series[area_key].append({ yr, qtr, hpi: float(hpi_value) })

for each area in series:
  series[area].sort_by(yr, qtr)
  if len(series[area]) < 8: skip  # insufficient data

  latest = series[area][-1]
  oneYrAgo  = find(series[area], yr=latest.yr-1, qtr=latest.qtr)
  fiveYrAgo = find(series[area], yr=latest.yr-5, qtr=latest.qtr)

  pct1yr = oneYrAgo  ? (latest.hpi - oneYrAgo.hpi)  / oneYrAgo.hpi  * 100 : null
  pct5yr = fiveYrAgo ? (latest.hpi - fiveYrAgo.hpi) / fiveYrAgo.hpi * 100 : null

  changes = [(series[area][i].hpi - series[area][i-4].hpi) / series[area][i-4].hpi * 100
             for i in range(4, len(series[area]))][-10:]
  volatility = stdev(changes) if len(changes) >= 4 else null

  if pct1yr is null: guspm_raw = null
  else:
    guspm_raw = 0.5 * pct1yr
               + 0.3 * (pct5yr ?? 0) / 5
               + 0.2 * max(0, 3 - (volatility ?? 3))

  results.append({ ..., latestHpi: latest.hpi, pct1yr, pct5yr, volatility, guspm_raw })

# Normalize
min_g = min(r.guspm_raw for r in results if r.guspm_raw is not null)
max_g = max(r.guspm_raw for r in results if r.guspm_raw is not null)
for r in results:
  r.guspmNorm = (r.guspm_raw - min_g) / (max_g - min_g) * 100 if r.guspm_raw else null

GUSPM_ENTRIES = results.sort_by(guspmNorm, descending=True)

Limitations & caveats

• Index-based, not price-based: FHFA HPI is a repeat-sales index. It measures % change in prices, not absolute price levels. An index value of 495 means prices have risen 395% since the base period (1980Q1=100 for All-Transactions NSA), not that homes cost $495.
• Enterprise mortgages only: The All-Transactions index covers only mortgages acquired or securitised by Fannie Mae and Freddie Mac, plus FHA/VA appraisals for refinancings. It excludes cash sales and jumbo (above conforming limit) mortgages, which can be significant in high-cost metros.
• GUSPM weights are illustrative: The 50/30/20 split is a judgment call; changing it would reorder areas. The normalized score is for comparison within this dataset, not across datasets with different methodologies.
• Volatility window: The last 10 quarterly periods (≈2.5 years) capture recent cyclical variation. Longer windows would smooth structural changes; shorter windows would amplify recent events. Areas with short FHFA series show “insufficient data” for volatility.
• Not investment advice: GUSPM is a statistical summary of FHFA HPI data. It is not a property valuation, mortgage advice, or investment recommendation.