The Brazilian Housing Liquidity Index: Measuring the Ability to Transact in Brazil's Residential Market

Radar Perene Working Paper Series, No. RPWP-004 · Version 2.0 · July 2026 House price indices are the most visible statistic produced about any residential market, and in Brazil they are effectively the only statistic most observers ever see. Yet price answers a narrow question — at what level did the marginal transaction clear — and remains silent on a question that matters at least as much to households, lenders, and researchers: whether a transaction can be completed at all, in reasonable time, at reasonable cost of financing. A housing market can reprice and freeze simultaneously. Between 2014 and 2016, Brazilian residential prices drifted slowly in nominal terms while mortgage concessions collapsed, financed transaction volumes contracted, and the effective ability of an ordinary household to buy or sell a property deteriorated dramatically. No widely published Brazilian statistic captured that deterioration as a single, interpretable number. This paper documents the Brazilian Housing Liquidity Index (Índice de Liquidez Imobiliária, ILI), a monthly composite that measures the ability to transact in Brazil's residential market rather than the price at which transactions clear. The index aggregates eight components organized into four pillars — credit conditions, market pricing (rental yield relative to long real rates), transaction activity, and credit stress — built from public institutional data: Central Bank of Brazil statistical series, FipeZap price and rent indices, ABECIP financed-transaction data, and the market real-rate curve. A fifth pillar covering supply and construction is a disclosed, planned extension. Components are normalized by expanding-window percentile rank — each month ranked only against data available up to that month, with a 24-month burn-in per component — and aggregated to a 0–100 scale read through five bands, from frozen (0–20) to euphoria (80–100). The official published series is monthly from December 2015 through May 2026 — approximately 126 observations, adopted in production in July 2026 and verified against the production database as of that date (latest reading: 52.0, "balanced"); the burn-in consumes 2014–2015, underlying inputs reach back to 2011 and earlier, and the earlier full-sample and prototype vintages are retained as methodological history. Validation is descriptive and deliberately modest, and its headline is computed on the official expanding-window series itself: the index's Spearman correlation with forward FipeZap price changes rises with horizon — +0.57, +0.61, and +0.66 at three, six, and twelve months — against +0.63, +0.66, and +0.72 for the full-sample version on the identical window, quantifying the in-sample bias at approximately 0.06 of rank correlation; forward six-month outcomes bucketed by index level show monotonic separation in both frequency and median of subsequent nominal price changes; and a monetary-stance benchmark test shows that an inverted expanding percentile of the housing-finance rate alone carries near-zero forward correlation with prices, while the index shares only 0.42–0.48 of rank correlation with that benchmark and retains forward rank correlations within every tercile of the rate cycle — the index is not a restatement of the monetary cycle. Version 2.0 audits, corrects, and extends that record against the production database (July 2026). Two published claims are retracted or corrected in the text: the research-prototype bucket table and its 2011-era calendar anchor could not be reconstructed and are retired — the 2011–2013 "heated" characterization now rests on the underlying inputs alone, not on any computed vintage — and the earlier statement that the series' opening readings were its historical minimum is corrected: the official series opens deep in the weak band (27.6 in December 2015) and prints its all-time low (23.0) in April 2016. Version 2.0 also adds the tests the earlier versions committed to: block-bootstrap confidence intervals that exclude zero at all three horizons while making the effective sample explicit (approximately 41, 20, and 9 independent windows); and a race against trivial single-series benchmarks, which the composite wins with intervals excluding zero against the concession and rate benchmarks at every horizon — together with the unfavourable result reported beside it: an equal-weighted mean of the same eight components ties the proprietary weights, so the index's value lives in the component set and normalization, not in the weight vector. The index is experimental, standalone, and display-only by internal doctrine: it does not feed the Radar Perene regime engines and will not until it passes a defined validation gate. The index is a descriptive instrument for reading the state of housing-market liquidity within its own history; it is not a forecast of prices, not a market-timing signal, and not investment advice. Keywords: housing liquidity; residential real estate; Brazil; composite index; mortgage credit; SFH/SBPE; FipeZap; transaction volume; market regimes ---