Does Haze Really Move Property Prices in Singapore? What Past PSI Spikes Show

In June 2013, Singapore's 3-hour PSI hit 401 — the highest reading ever recorded — and in that same year, HDB resale prices reached their all-time peak. Six years later, in September 2019, the haze returned and HDB resale prices ended the year up 0.1%. In October 2023, the haze came back again, and resale prices rose by roughly 5% for the year.

Three haze episodes. Three different price outcomes — including one where the worst air quality on record coincided with the most expensive resale flats Singapore had ever seen.

That inconsistency is not a data problem. It is the answer. If haze genuinely moved Singapore property prices, we would expect the effect to scale with how bad the haze got, and to point in the same direction every time. It doesn't. What we mostly see instead is a market behaving exactly as it does in any other September or October — a seasonal lull being retrofitted with a dramatic explanation.

This piece works through what the past PSI spikes actually show, why any visible dip is far more likely seasonal and behavioural than causal, and what buyers and sellers should realistically do with that knowledge — including the one place where haze genuinely does hand buyers a small edge.

The Short Answer

Here it is up front, so you can decide whether to read the next 4,000 words or go check the air purifier filter.

• Haze can plausibly nudge transaction volume in the weeks it is worst. Viewings get postponed. Agents reshuffle appointments. This is a timing effect, not a demand effect.
• Haze has no established mechanism for moving price levels. Price is anchored to affordability, financing rules, supply and comparable transactions. Weather touches none of those.
• The months when haze typically bites — September and October — are already the softest months in Singapore's resale calendar. This is the single biggest reason people misread haze as a market driver.
• The dosage-response test fails. 2013 was far worse than 2023, yet 2023 was the stronger price year. A real causal factor would scale with intensity.
• Monthly transaction samples are too thin to support the claims people make from them. A district with a dozen resale deals in a month cannot tell you anything about a 1% price effect.

The rest of this article is the evidence and the reasoning behind those five points.

Why Haze and Singapore Property Prices Keep Getting Linked

The link feels intuitive, and intuition is exactly why it survives. Haze is unpleasant, visible and memorable. It arrives, the market goes quiet for a fortnight, and a story writes itself: the haze killed the market.

Three things make that story stick.

First, the timing overlap. Singapore's haze episodes cluster in the second half of the year. The 2013 episode peaked in June with follow-on episodes through August and September. The 2015 episode peaked in late September. 2019's episode landed in mid-September. 2023's landed in early October. Meanwhile, the transacted volume of resale flats and private homes also tapers through the second half of the year. Two things that share a calendar will always look correlated.

Second, haze is a narratively convenient explanation for a slow month. Nobody wants to say "viewings were down because it's October and everyone's saving for December." Saying "the haze" is shorter and more interesting. This is not a conspiracy — it is just how market commentary works when nobody is forced to control for seasonality.

Third, air quality is genuinely a liveability factor, and buyers do care about liveability. The confusion is that caring about something is not the same as pricing it. Singaporeans care about which floor a unit is on, whether the block faces west, and how far the walk to the MRT is. Some of those are quantifiably reflected in price. Haze exposure, so far, is not — at least not in any way that shows up cleanly in published transaction data.

The calendar nobody mentions

Before attributing anything to haze, it helps to know what a normal Singapore resale year looks like.

The resale market has a rhythm:

• Q1 (Jan–Mar) tends to be reasonably active, with post-CNY momentum and new-year decision-making.
• Q2 (Apr–Jun) is frequently the strongest stretch, particularly May and June.
• Q3 (Jul–Sep) is the soft patch. The seventh lunar month — colloquially Ghost Month — typically falls in August or early September, and a meaningful share of Chinese buyers and sellers avoid transacting then. School holidays in June also pull family decision-making out of the market.
• Q4 (Oct–Dec) starts with a modest bump after Ghost Month, then fades into the year-end holiday lull, with December routinely one of the thinnest months of the year.

Now layer haze on top. In 2015, 2019 and 2023, the worst of the haze landed squarely in the Q3 soft patch or the early-Q4 shoulder. Any weakness observed in those windows has at least three competing explanations before weather even enters the frame: Ghost Month, school holidays, and the general second-half slowdown.

The pattern you should notice is not that haze keeps showing up — it is that haze keeps showing up inside a window that was already soft for structural reasons. That is a textbook confounder.

What Past PSI Spikes Actually Show

Let's put the episodes side by side, with the property-market context that usually gets left out of the headline.

Year	Haze window	Reported peak 24-hour PSI	HDB resale price outcome	Dominant market driver that year
2013	June, with follow-on episodes into Aug–Sep	~246	Prices peaked, then began a six-year slide	TDSR (Jun 2013), MSR (Dec 2013)
2015	Late Sep–Oct	Reported in the 250–270 range	Prices continued drifting lower	Multi-year cooling cycle, rate normalisation
2019	Mid-September	~154	Roughly flat for the year	Market bottoming after 2018 cooling measures
2023	Early October	~123	Up around 5% for the year	Apr 2023 ABSD hike, supply-demand mismatch

A caveat on the numbers before anyone quotes them at a dinner party: PSI readings are published by NEA on different time windows — 1-hour, 3-hour and 24-hour — and contemporaneous reporting has cited slightly different highs depending on which window was being described. Treat the figures above as indicative published peaks, not laboratory-grade precision. The direction and rough magnitude are what matter for this argument.

2013: the year the worst haze met the highest prices

2013 is the cleanest demolition of the haze-price theory, because it is the one episode where the causal story should have been most visible.

That June, the 3-hour PSI hit 401. Schools closed. N95 masks sold out. The episode was, by a wide margin, the most severe Singapore has recorded in the modern PSI era.

And in that same year, HDB resale prices peaked. The subsequent six consecutive years of decline — running through 2019 — had nothing to do with air quality. They were driven by the Total Debt Servicing Ratio framework introduced in June 2013, the Mortgage Servicing Ratio for HDB loans that followed in December 2013, and a broader cooling-measures regime that reined in borrowing capacity across the board.

If haze moved prices, 2013 should have been the year the effect was unmistakable. Instead, the year's defining price event was a regulatory change to how much buyers could borrow. That is the honest ranking of causes.

2015: haze meets an already-cooling market

The September 2015 episode was severe — the 3-hour PSI reportedly peaked at 341 — and it overlapped almost precisely with the tail of Ghost Month and the start of the year-end slowdown.

Prices were already in a multi-year decline. Attributing any part of that decline to haze requires ignoring that the same downward drift was present in 2014, 2016, 2017 and 2018 — years with no meaningful haze.

The relevant question is not "did prices fall in the quarter the haze happened?" It is "did prices fall more than they would have otherwise?" Nobody has demonstrated that, and the trivial check — comparing the hazy quarter against the equivalent quarter in non-haze years — does not support it.

2019: haze meets the market bottom

The September 2019 episode pushed the 24-hour PSI into the unhealthy range, reportedly peaking around 154 — the most significant reading since 2015.

HDB resale prices for the full year 2019 moved by roughly 0.1%. That is the sound of a market that has stopped falling after a long correction, sitting quietly while it works out what happens next. The 2018 cooling measures had already done their work. A two-week haze episode was, in the scheme of things, a rounding error.

2023: haze meets a policy shock

October 2023 brought the 24-hour PSI into the unhealthy range again, reportedly peaking around 123 on 7 October — the highest reading in four years.

HDB resale prices ended 2023 up by roughly 5%. The dominant story of that year was the April 2023 round of cooling measures, which raised Additional Buyer's Stamp Duty across several buyer categories and tightened loan limits, followed by a market that simply absorbed the shock and kept climbing on the back of BTO supply-demand mismatch.

Look at that chart next to the outcomes in the table above. The severity ranking is 2013 > 2015 > 2019 > 2023. The price outcome ranking, from weakest to strongest, is roughly 2015 < 2019 < 2023 ≈ 2013. There is no relationship. That is the whole finding, compressed into two lines.

Meanwhile, here's what actually moved prices

For contrast, here is the annual change in the HDB resale price index over the last several years — the same stretch during which haze episodes came and went.

The swings here are enormous — from flat to double-digit growth and back — and every one of them lines up with a documented driver: pandemic-era stimulus and low rates, then rate normalisation, then a supply crunch, then cooling measures, then renewed demand. Not one of them lines up with a haze season.

Why the Mechanism Doesn't Hold Up

Correlation problems aside, the causal story has a mechanical problem: it doesn't specify how haze would move prices.

Volumes can dip. That's not the same as prices falling.

There is a defensible version of the haze effect, and it concerns volume, not price.

When the PSI climbs, people postpone things. A viewing scheduled for Saturday afternoon gets moved to the following weekend. An agent's open house draws fewer than half the usual crowd. That is a real friction, and it can absolutely show up in the number of transactions registered in a given fortnight.

But a postponed viewing is not a cancelled purchase. Unless the buyer's circumstances change — a loan expiring, a lease ending, a job move — the demand is still there, just later. Deferred demand typically produces a dip followed by a rebound. That creates a lumpy monthly volume chart and zero change to the underlying price level.

Price is set by affordability and comparables, not by the weather

Ask what actually determines the price of a resale flat or condo and you get a short list:

• Financing capacity — TDSR, MSR, LTV limits, and the prevailing interest rate environment.
• Supply — how many comparable units are on the market, and how many more are coming via BTO launches, en-bloc redevelopment or new launches.
• Income and employment — whether buyers can service the loan and expect to keep doing so.
• Comparable transactions — what the last similar unit in the same stack sold for, which becomes the anchor for the next negotiation.
• Buyer sentiment about the future — where people think prices will be in three years.

Haze does not appear on that list. It does not change what a bank will lend. It does not add or remove a single unit from the market. It does not change anyone's salary. It does not alter the last transacted price of the unit next door.

That diagram is the whole argument. Haze travels down the volume branch. It never reaches the price branch.

The exposure window is too short for the decision cycle

There's a time-scale mismatch that gets overlooked. Haze episodes in Singapore are episodic — typically days to a couple of weeks at their worst, sometimes recurring across a season. The average property purchase decision, by contrast, runs over weeks to months: browsing, shortlisting, viewings, loan pre-approval, negotiation, exercise, completion.

A two-week disruption inside a three-month process usually just shifts the internal schedule. It rarely kills the process. And because the buyer's financing position and budget haven't changed, the price they can pay hasn't changed either.

The dosage-response test fails

This is the most decisive analytical point, and it's simple enough to apply yourself.

If haze affects prices, the effect should scale with severity. 2013's peak was roughly twice 2023's. If haze cost the market even half a percentage point of price growth in 2013, it should have cost a visible fraction of that in 2023.

Instead, the price outcomes point in opposite directions across episodes, and the year with the mildest haze (2023) was among the strongest price years of the four. When a supposed cause produces effects with no relationship to its intensity, the cause is not doing the work. Something else is.

The Thin-Data Trap: Why One Hazy Quarter Proves Nothing

Even if the mechanism were plausible, the data most people use to argue it is far too thin to carry the weight.

Singapore's monthly resale samples are small

HDB resale transactions run in the low thousands per month across the entire island. Split that across 26 towns and multiple flat types, and a given town-flat-type combination might see a few dozen deals in a month. Private resale volumes are thinner still: at the district level, a quiet month can produce single-digit or low-double-digit transactions for a specific project.

Now consider that you're trying to detect an effect that might, in the most generous reading, be one or two per cent. Detecting a 1% effect in a sample of 30 transactions is not analysis; it is astrology with a spreadsheet. The confidence interval around that median PSF will be wider than the effect you're claiming.

Anyone who tells you "District X's PSF fell during the haze month" is almost certainly describing noise.

The single-quarter fallacy

The second trap is time aggregation. Pull a quarterly URA or HDB dataset, find a weak Q3 in a haze year, and the story writes itself. But Q3 is structurally weak — Ghost Month, school holidays, the pre-year-end lull. Comparing a hazy Q3 to a strong Q2 is comparing apples to a different season's oranges.

A weak Q3 tells you it was a Q3. That's it.

That flow is worth bookmarking. It will correctly classify the overwhelming majority of "haze hit the market" claims you encounter.

A five-point sanity check anyone can run

If you want to test the haze hypothesis yourself, here's a framework that will keep you honest.

1. Compare like with like. A hazy September should be compared to the previous five Septembers, not to the preceding June.
2. Use rolling three-month medians. Single-month medians in small samples bounce around for no reason. A three-month rolling median smooths that out.
3. Check the volume side first. A genuine demand shock shows up in volume before price. If volume dipped but the price level didn't shift, you're looking at friction, not repricing.
4. Segment by region. Haze is broadly islandwide, though NEA publishes regional PSI readings and some regions do run higher. If a "haze effect" shows up in only one or two districts, it's not haze — it's a specific project, a specific launch, or a specific seller.
5. Look for the rebound. Deferred demand returns. If there's no rebound after the haze clears, the cause was something else.

Does Haze Create Negotiating Room for Buyers?

Here's where the analysis gets practically useful, and where a small amount of truth hides inside the common wisdom.

If haze does suppress viewings, then the buyers who do turn up during a hazy fortnight face less competition. In a market where a well-priced resale flat can still pull multiple offers in the first weekend, that's not nothing. Fewer competing buyers means less pressure to bid up, more time to think, and an agent who is more attentive because you're one of three people who showed up rather than one of twenty.

But be precise about what that gets you.

Where the leverage really comes from

The negotiating room doesn't come from the weather. It comes from seller motivation colliding with low turnout.

A seller's willingness to accept below asking is a function of their own deadline — a lease ending, a school registration, a job relocation, a bridging loan running out, or simply fatigue after eight weeks on the market. Buyers who show up during a low-turnout window occasionally find a seller who is further along that curve than they would be in a busy month.

That's a genuine edge. It's just not a haze edge. The same seller would be equally motivated if the market were quiet for any other reason — a rainy fortnight, a school exam period, a Ghost Month, a sudden rate scare.

What haze does not change

• The comparable transactions. If the last three units in the stack sold at S$1,750 PSF, that anchor is still there. Weather doesn't reset it.
• The seller's cost basis. Their mortgage, their outstanding loan, their minimum acceptable number.
• Bank valuations. A valuer looks at recent comparable transactions, not the PSI.
• Supply. The same number of units are competing for the same buyers.

So the realistic summary is: haze might widen your negotiation band by a small amount for a short period, in the specific case where the seller is already motivated and you're the only serious buyer at the table. It will not produce a market-wide discount.

The behavioural risk nobody talks about

There is a flip side worth flagging. Low-turnout months do strange things to sellers' heads. A seller who gets zero offers during a hazy fortnight may conclude — wrongly — that the market has turned against them, and either drop their price aggressively (good for a sharp buyer who moves fast) or pull the listing entirely and wait (bad for everyone).

There's also the risk of the buyer misreading their own success. If you negotiate a 3% discount during a quiet October, ask yourself whether you got it because of the haze or because the market was quiet. If you don't know, you can't repeat it next time.

What Buyers and Sellers Should Actually Do

Strip away the haze framing and the practical advice is refreshingly boring.

For buyers

• Use hazy months to view, not to decide. Fewer people at viewings means more unhurried inspections and better access to the agent. Bring the questions you'd normally rush through.
• Check liveability factors that haze exposes. Cross-ventilation, floor level, whether the unit faces the prevailing wind, how well it holds air with windows closed, and whether the building's common areas get stuffy. These are real quality-of-life variables that only become obvious when you're forced indoors.
• Anchor your offer to comparables, not to the weather. Pull the last three to five transactions for the same project and similar units. That's your negotiating range.
• Watch the rebound. If volume dips during the haze and recovers within a month or two, you're looking at deferred demand. Expect competition to return, and don't assume the quiet window will last.
• Ignore single-month PSF moves. Track rolling medians. A single month in a thin sample is not a signal.

For sellers

• Don't price around the weather. You'll be wrong for the same reason everyone else is: haze affects turnout, not value.
• If your listing window overlaps a hazy fortnight, adjust your expectations on time, not price. Expect fewer viewings and a longer path to offers. Expecting a lower price would be a mistake.
• If you have a genuine deadline and turnout collapses, that's a decision point — but it's about your deadline, not the haze. The relevant question is whether you can carry the property for another two months. If you can, wait. If you can't, discount — and accept that you're discounting because of your timeline, not because the market re-rated.
• Use the quiet period productively. Better photos, updated floor plan, sharper listing copy, a re-look at the asking price relative to fresh comparables. Low-turnout windows are a good time to fix the things you can't fix in a crowded month.

What Actually Moves Singapore Property Prices

Since we've spent this long on what doesn't, it's worth restating what does — and how to tell them apart from weather noise.

Factor	How it hits prices	Typical lag
Cooling measures and borrowing limits	Directly changes how much buyers can pay	Immediate to one quarter
Interest rates	Changes monthly instalments and affordability ceilings	One to two quarters
Supply pipeline (BTO, new launches, en-bloc)	Changes competition for the same buyer pool	Two to six quarters
Income and employment	Changes long-run ability to service loans	Slow, persistent
Buyer sentiment about the future	Compresses or expands willingness to transact	Weeks, but noisy
Haze	No established mechanism	None observed

The gap between the top five rows and the last one is not a matter of degree. The top five operate on affordability, supply, or expectations. Haze operates on convenience. And convenience, in a market where the decision cycle runs for weeks, mostly just reschedules things.

Even in markets that take weather far more seriously — think hurricane season in Florida or flood-risk zones in the UK — the pricing effect comes from permanent changes in risk perception and insurability, not from a bad fortnight. Singapore's haze is episodic, seasonal and widely understood to be temporary. That's precisely why it doesn't get capitalised into price.

Food for Thought

1. If haze really suppressed demand, why was 2023 — the mildest of the four episodes — among the strongest price years? What does the absence of a dosage-response relationship tell you about the claims being made?

2. How much of what we call "market sentiment" is actually just the calendar? If Ghost Month, school holidays and the year-end lull produce the same quiet-period pattern every year, how much market commentary is really just seasonal patterns being given dramatic names?

3. If you negotiated a discount during a hazy month, would you know whether you earned it or got lucky? What would you need to record — turnout, days on market, competing offers — to be able to tell the difference next time?

4. What else do we attribute to visible, memorable causes simply because they're easy to remember? A rainy weekend, a bad news cycle, an election period, a viral social media post about a "cooling market." How many of those survive the same five-point sanity check?

5. Does a liveability factor have to move PSF to be worth caring about? Air quality, noise, afternoon sun, ventilation — some of the things buyers value most are hard to find in a price index. If the market doesn't price it, does that mean it doesn't matter, or just that it isn't measured?