Do dwelling prices follow wages, interest rates, the stock market? The study sets 20 outside series beside 8 housing series, quarterly from 2,008 to 2,026, and checks every link three times: whether it holds with an interval, whether it holds once we allow for the best lag having been picked, and whether the indicator improves a forecast a year ahead.
Every link is between annual growth rates, never levels. In levels anything that grew "correlates": SBITOP and dwelling prices both roughly doubled since 2015, which says only that both went up.
The series
Four families: housing (the SURS price index, rents), the economy (GDP, wages, employment, prices), interest (mortgages, Euribor, the government bond) and equities (SBITOP, two European indices). Pick a family.
Quarterly means. Wages: the labour cost index; the jump at the end of 2025 is the public-sector pay reform. ETN rent only 2012–2024 and for quarters with 400+ new leases. HICP rents include non-profit rents. SBITOP before 2010 is the OECD share price index for Slovenia, rescaled to SBITOP.
Source: SURS, Eurostat, ECB, LJSE, OECD, STOXX, ETN · bloki.si
Prices are part of the business cycle
Sentiment and activity move most closely with price growth: construction confidence 0.77 (90% interval 0.59–0.88), services confidence 0.68, GDP 0.64, employment 0.63, economic sentiment 0.59, retail trade 0.58. Unemployment moves the other way (−0.50).
None of them leads prices: the strongest links sit in the same quarter or one quarter apart. Prices are part of the cycle, not a forecaster of it. The links hold for used flats in Ljubljana and elsewhere, for houses, and for the ratio of price to GURS value.
Correlations of annual growth rates, not levels. 90% interval, block bootstrap with 8-quarter blocks, because annual growth rates overlap. The strongest lag is picked from 17 and so overstates on its own; the test that allows for it is in the next chart.
Source: SURS, Eurostat, ECB, LJSE, OECD, STOXX, ETN · bloki.si
Who leads whom
With 17 lags to choose from, the best one will always look good. So every best correlation is compared with the best one the same series gives when it is circularly shifted by three years or more: that keeps both series' persistence and breaks only their alignment.
This test removes wages. Wage growth reaches 0.67 against prices, but with prices leading wages by a year, and random alignments of two smooth series come out as high (p = 0.16). Pick an indicator.
Band: 90% interval. p: how often a random alignment of two series with the same persistence (a circular shift of three years or more) gives a correlation at least as strong at any lag. The smallest attainable p is about 0.02.
Source: SURS, Eurostat, ECB, LJSE, OECD, STOXX, ETN · bloki.si
Interest rates rise with prices
The annual change in the mortgage rate moves positively with price growth, 0.47; three-month Euribor 0.49. Prices and rates answer the same cycle: when the economy grows the central bank raises rates, and that reaction outweighs the dearer credit.
In 2017–2026 the same-quarter link is nearly gone (0.10) and at a two-quarter lead it turns weakly negative (−0.21). That is the only hint of the textbook sign, and it is not reliable. New mortgage volume correlates with nothing (0.07).
Both series standardised over their common span (deviation from the mean in standard deviations). Equities are shifted two quarters forward. Correlation is not a cause.
Source: SURS, Eurostat, ECB, LJSE, OECD, STOXX, ETN · bloki.si
Equities lead by two quarters
SBITOP moves with prices at 0.45 in the same quarter, but at 0.64 two quarters earlier (interval 0.25–0.85, p = 0.02). The lead holds in both periods: 0.47 in 2008–2016 and 0.43 in 2017–2026.
The European indices show nothing in the same quarter (EURO STOXX 50 0.14, STOXX 600 0.06) and 0.44 and 0.37 two quarters earlier.
Each period has only 32–38 overlapping quarters, so the intervals are wide.
Source: SURS, Eurostat, ECB, LJSE, OECD, STOXX, ETN · bloki.si
A correlation is not a forecast
Does an indicator improve a forecast of price growth a year ahead, once we know how fast prices grow today? The model is refitted every quarter on what was known then; 46 forecasts from 2014 to 2025.
Only the three equity indices improve it: SBITOP cuts the error to 0.95 (interval 0.93–0.97), EURO STOXX 50 to 0.94, STOXX 600 to 0.94. Every activity, sentiment and rate series makes the forecast worse or leaves it unchanged. Moving together is not foresight.
The gain is modest, a few per cent of a 7.7 percentage-point error, and the three indices are one finding, not three. It is a weak signal, not a predictor. SBITOP's annual growth runs between 35% and 41% in 2025–26 against 6% to 9% for dwelling prices: the largest equity signal in the series, which by its weight in the forecast implies little.
A forecast of annual growth four quarters ahead from today's growth, with and without the indicator; both models are refitted every quarter on what was known then. 90% interval. With 20 indicators, one false 'reliable' is expected by chance.
Source: SURS, Eurostat, ECB, LJSE, OECD, STOXX, ETN · bloki.si
Wages follow prices, not the other way
In the same quarter wages move with prices at 0.38. The strongest link has prices leading wages by a year, and in 2017–2026 the same-quarter correlation is even negative (−0.35). Wages caught up with prices; they did not drive them. The rolling correlation shows how links change over time; pick a group.
Correlation over the 20 quarters to the date shown. Five years of overlapping annual growth rates are about five independent observations: the chart shows direction, not level.
Source: SURS, Eurostat, ECB, LJSE, OECD, STOXX, ETN · bloki.si
Rents follow inflation
HICP rents move with inflation at 0.75, with the cycle more weakly (employment 0.65, construction confidence 0.62) and with equities not at all (SBITOP 0.16). Part of that is mechanical: the HICP includes non-profit rents, which are indexed to prices.
Market rents of new leases (ETN, 2013–2024) are robustly linked only to inflation (0.53, p = 0.02). Their other "best lags" sit at the edge of the window and fail the test: a few dozen quarters of a noisy median cannot carry a lead–lag analysis. Slovenia has no market rent index.
Same-quarter correlation. HICP rents include non-profit rents; ETN rents 2013–2024. Price/value: the median ratio of price to the GURS generalised value in flat sales.
Source: SURS, Eurostat, ECB, LJSE, OECD, STOXX, ETN · bloki.si
Real prices
Deflated by the HICP, prices keep their links to the cycle (economic sentiment 0.68, construction confidence 0.71, GDP 0.61) and the SBITOP lead (0.61). They lose the link to inflation (0.08): the nominal one was the 2022 price surge, not a channel.
§ How it is calculated
Growth, not levels. Every index series enters as annual log growth in per cent, rates and unemployment as a four-quarter change, sentiment surveys as a level.
Intervals come from a moving-block bootstrap with blocks of 8 quarters, because annual growth rates overlap and neighbouring quarters are not independent observations.
Choosing the lag. Lags run from −8 to +8 quarters. The best is the largest of 17, so its p compares it with the largest under every circular shift of the indicator by three years or more. There are only about fifty such shifts, so the smallest attainable p is about 0.02.
The forecast. Growth 4 quarters ahead from today's growth (the baseline) and from today's growth plus the indicator; both models are refitted every quarter on the past only.
Sources. The SURS dwelling price index (2026 restated from the new 2025 base), Eurostat, the ECB, the Ljubljana Stock Exchange (before 2010 the OECD share price index for Slovenia, which is SBITOP on another scale), STOXX, and the real-estate transaction register for the price/value ratio and rents. ETN rent is cut at the end of 2024, because from 2025 only a sixth of new leases are filed while the median jumps by two fifths: a change in who still files, not the market.
What this cannot say. Eighteen years, one bust and one boom: every interval is wide. Correlation is not a channel. Rates are tested as changes, not as the burden of a repayment; that is another question.
Can the Bloki server answer questions like these?
The server holds the property market, not the economy: the price index, sales, rents and adverts for every municipality. It holds no GDP, rates or stock indices, and it makes no forecasts.
- Yes
How have prices moved in my municipality?
The price index (2023 = 100), national or by municipality, flagging which quarters are reliable, and sale medians with their n.
price_indexmarket_stats - Yes
How much is sold and let in my municipality at all?
Sales and lettings per 1,000 building parts a year, dwellings and business premises apart, with their n.
market_activity - Yes
What is a flat bought in 2019 worth in today's money?
Carries the price with the index and states the index's measured bias.
carry_price_forward - No
Will prices rise next year?
The server does not forecast. It gives the index to the last reliable quarter; this article shows how little even the best indicator improves a forecast.
price_index - No
How do interest rates, wages or the stock market affect prices?
The server holds no economic series. The answer is in this article, from SURS, Eurostat, the ECB and the exchanges.