Supplement 6 Sep 26 - The Forecasters Astrology, a 2007 Gilt Yield and the Productivity That Was There All Along

"The only function of economic forecasting is to make astrology look respectable." - John Kenneth Galbraith

The quote pertains to the deep dive, as ever, and this week it pertains to the whole edition, because the week has been an argument about forecasting. Neil Woodford has spent a fortnight on his Noise Cancelling platform arguing that the people who set your interest rate and your tax bill have been working from a productivity number that was wrong by half, and that the same people who missed inflation on the way up in 2021 are now missing it on the way down. The gilt market, meanwhile, has spent the week pricing the opposite: a ten-year yield at a level not seen since August 2007, and a Bank Rate hike fully priced by Christmas. One of those two camps is going to look like an astrologer by the spring. I have a lot of sympathy with Woodford, as you will see, and one significant reservation, which you will also see. Galbraith, for what it is worth, said it about his own profession, which is the only kind of forecasting joke worth making. Regulars know I love a dad joke - what passes for banter in this game is nobody’s business.

As the summer's twenty-seven-year high (millenium high!?) on the thirty-year gilt is joined by a nineteen-year high on the ten-year, and the Budget countdown ticks over into the fifties (52 days to the red box at publication, for those keeping score), our next Property Business Workshop is live and tickets are selling. This is the really juicy one for taking big steps forward in your property businesses: JVs and M&A, joint ventures done properly, and buying and selling property businesses, portfolios and companies, not just houses. If this week's Supplement has a single practical message, it is that the cost of money is going to be argued over for the rest of the year and the deals that get done will be the ones structured to survive that argument. Book in on the next Property Business Workshop with myself and Rod Turner - Thursday 1st October - Manchester - https://tinyurl.com/pbwoct26. The VIP dinner seats are down to the last one or two, so if you want that extra time with Rod and myself, now really is the time. We are proud that this workshop is powered by Roma Finance, who in my experience really do think differently about bridging and development lending (they don't just say that, like some lenders do!), and they will be represented on the day. Early Bird pricing, at more than 10% off, is in its final stretch.

Welcome back to Trumpwatch. If last week was the week the markets decided "economic warfare" was the cheaper kind of war, this week was the week the other kind came back. A month without military action ended on Sunday night, and by Wednesday the President was standing in the Oval Office explaining that the renewed bombing campaign would not last "too long", which is roughly what he said in March. Three things to take in order, as usual: the war and the oil price, which have rejoined each other after a fortnight apart; the US jobs data, which arrived on Friday with a number nobody had pencilled in; and the Federal Reserve, which now has a coin flip that has become a sixty-forty. Then the bit that matters to you, which is what all of this did to a gilt market that touched a level last seen in the month the credit crunch began.

The war first. US Central Command began striking Islamic Revolutionary Guard Corps targets on Sunday, and again on Tuesday afternoon Washington time, citing attempted Iranian attacks on commercial shipping in the Strait and on American service members. The targets, per the Pentagon and per the President's own description, were the kit Iran has been rebuilding along its southern coast since July: radar, air-defence sites, mine-laying capability, and what Trump described as "a rocket that drops mines". Bandar Abbas, Chabahar, Qeshm, Jask and Sirik all reported explosions. Iran replied with drones and missiles at US bases across the region, which is the pattern of the whole six-month war, and accused the Americans of hitting a wedding at Sirik, with the Iranian Red Crescent putting the dead at four or five and the injured near seventy; CENTCOM's spokesman said the US military never targets civilians. Two supertankers were hit by projectiles trying to leave the Strait on Monday. And then, on Tuesday night, the line that tells you where this is going: "I couldn't care less if they sign a worthless, to them, agreement. I like our position now much better, with almost total control of the Hormuz Strait, and their economy totally collapsing." Read that twice. The President is no longer asking for a deal. He is saying, in terms, that the current arrangement suits him. That is a very different war from the one the market was pricing at $88 last Sunday.

And yet, as ever with this conflict, the physical picture and the headline picture disagree. The Energy Secretary, Chris Wright, told CNBC that more than 17 million barrels of oil transited Hormuz on Monday, and CNN reported that the US Navy escorted 40 vessels carrying 18 million barrels through on Tuesday, the day of the strikes. Traders' own estimates for recent weeks are more like 6 to 8 million barrels a day, against a pre-war 20 million. So on the good days the Strait is running at most of its old capacity under armed escort, and on the bad days at a third of it, and the market is paying for the uncertainty rather than the average. Brent went from $88 a week ago to a six-week high above $96 by Wednesday, eased on Trump's "not too long" comment, and sits at $96 as I go to print, with WTI at $91. European gas did the thing that actually matters for your January bill: the TTF benchmark touched its highest since January 2023 mid-week, around €72 a megawatt-hour, double where it was a year ago. Not 2022, when it peaked at €339. But the Ofgem window for the January cap is being filled right now, and it is being filled at these prices. More on that in the Macroscope and the Deep Dive. It’s the first time I’ve looked at the winter forecast (for what the long-range forecasts are worth this year) but those who have been plugged into this will know that the supersized El Nino has dominated the chatter this year, and we are likely to have heavy rain and wind, rather than freezing or heavy snow. But when the facts change here, we change our minds, of course - just like the weather. 

Elsewhere in the theatre, and briefly. Putin gathered leaders in Kyrgyzstan to condemn the strikes and the sanctions, which he will have enjoyed. Mark Carney, of all people, accused the Trump administration of "doing memes" instead of serious trade talks, which is the sort of thing a former Governor of the Bank of England says when he has run out of patience; Canada's tariff position is unchanged, so the memes are the news. Treasury Secretary Bessent trailed yet another round of sanctions on Tehran. The attempt to remove Governor Lisa Cook from the Fed Board, reinstated last week, rumbles on. Standard Fare, all of it, with the exception of the bit where the President said he no longer wants an agreement, which is not standard at all.

Now the jobs number, because it turned the week on its head on Friday afternoon. US payrolls rose 162,000 in August against a consensus of 53,000 to 56,000 depending on whose survey you read: roughly three times what anyone expected, the strongest month since March, with July's loss of 23,000 revised away into a gain. Unemployment held at 4.1%. The household survey backed it up: 569,000 more people employed, 683,000 more in the labour force, participation up two tenths, and the broad underemployment measure down to 7.7%, its lowest since June 2025. Hourly earnings rose 0.3% on the month and 3.1% on the year, a tenth above forecast. Food and drink establishments added 59,000 of the 162,000, which is the sort of detail that makes you wonder about seasonal adjustment in a summer that was, by all accounts, very hot on both sides of the Atlantic; information jobs fell 23,000 and have been falling at 8,000 a month for a year, which is the AI-in-the-labour-market story continuing quietly under the headline. Take the report at face value, though, and the "softening labour market" that was supposed to stay the Fed's hand in September has, for one month at least, stopped softening.

Which brings us to the Fed, and to Warsh's work-to-do speech from Wyoming a week ago that now has a payrolls print to sit alongside it. Futures had a September hike at around 50% before the jobs data and a little over 60% after it. Three voices to weigh. Governor Barr said on Tuesday he would back a hike if inflation does not moderate soon enough. Governor Waller, on Thursday, pulled the other way: he would favour staying on hold as long as the monthly inflation reports show prices moderating, which is the sentence that gave the global bond market its one good afternoon of the week. New York's Williams is in "wait-and-see" mode. So the committee is split three ways, the Chair has told them he has work to do, and the data that decides it lands next week: producer prices on Thursday and the August CPI on Friday. US CPI was 3.4% in July against our 2.9% (or really against our 3.1% because our CPIH is closer to their CPI than our CPI is, so I always feel obligated to point that out), and if it prints hot, the 15th-16th September meeting becomes the first Fed hike of this cycle at a moment when the ten-year Treasury is already at 4.79% and the thirty-year at 5.27%. If it prints cool, Waller wins and the Treasury market gets its breath back. I would not put real money on either, and I note that the futures market, which does, has moved ten points in a week on one jobs number. That is not conviction. That is a market waiting for a CPI print. One thing to get your head around - and this is the same in the US and the UK - a hike right now is likely to bring mortgage rates and bond yields DOWN, because it is the action that the hawks are waiting to see happen and almost daring the central bankers NOT to do, if that makes sense!

Phew - stepping away from the transatlantic macro-storm, let's get back to the reassuringly measurable reality of the real-time UK property market. Reassuringly measurable, that is, apart from the bits I had to get by private message.

As is customary, Chris Watkin has been relentlessly crunching the portal numbers and then publishing them at Property Industry Eye. His analysis for Week 34 of 2026 - the week ending 30th August - is where it is at, as always. If you want to know how the macroeconomic gridlock translates to the local high street, and the REAL property market on the ground, look no further. Chris's headline this week asks the question I get asked most at the moment: the UK property market is not struggling, so why does it feel like it is? His answer, which I am going to borrow and then argue with slightly, is choice. Buyers have more of it than they have had for a decade, and a buyer with choice is a buyer who can wait. That makes sense. We always offer 2 options to vendors where we can, when there is responsiveness to a creative solution to make a deal work - even 3 we find is too many. 

Supply first. 33.4k new listings this week, up from 31.8k the week before, against a 2026 weekly average of 36.3k and a ten-year average for week 34 of 33.2k. So the tap, which I said last week was running slower than normal, has this week run almost exactly at the long-run norm for the last week of the school holidays. The year-to-date picture is the one to hold onto: 1.231m new listings so far in 2026, which is, to the rounding, the same as 2025 (1.232m), 2.7% ahead of 2024, 10.5% ahead of 2023, and 10% above the 2017-19 pre-Covid average of 1.119m. Regular readers will remember my year-plus adage of "10% more stock than a normal market" as the ready reckoner. After drifting above 12% in the early summer and settling back to the reckoner level a fortnight ago and a fraction below it last week, this week it is sitting on the reckoner to the decimal point. Three readings in a row at or around 10% is the nearest thing to a stable equilibrium this series has offered all year. I would not call it mean reversion yet. I would call it the bath finding its level.

Demand, then. 23.2k homes went sold subject to contract in Week 34, essentially flat on the 23.4k of the week before, against a ten-year week-34 average of 24.7k and a 2026 weekly average of 24.5k. So we are running about 6% below the long-run norm for the week and about 5% below our own average for the year, which is the wobble I flagged three weeks ago neither deepening nor healing. The year-to-date gross figure is 832k homes sold subject to contract, which is 6.8% below 2025's 893k, 1.1% below 2024, 11% ahead of the limp lettuce that was 2023, and 5.7% above the pre-Covid 2017-19 years. And here is the number I keep coming back to: the decade average for this point in the year is 833k. We landed EXACTLY on it last week; this week we are a thousand homes under. Decade-average transactions, at decade-average-plus-ten-per-cent stock, with the Five Ds - death, debt, divorce, downsizing and the diddy ones - doing what they always do. Is that a struggling market? Not on the volume. It is a market where the same number of buyers as usual are choosing from ten per cent more houses than usual, and, as Chris says, choosing slowly. If you are selling anything at the moment, this isn’t a market to be going stale in. 

Net sales: 17.7k for the week, identical to last week, against a ten-year week-34 average of 18.3k and a 2026 average of 19k, taking the year-to-date to 645k. That is 5.4% below 2025 (682k), 0.7% below 2024, 14% ahead of 2023 and 3.9% above the pre-Covid average of 621k. Note again the gap between the gross and net comparisons to 2025: gross is 6.8% down, net only 5.4% down, because fewer of this year's agreed sales are falling over. That line has now held for more than a month and I am prepared to call it a feature rather than noise.

Now the friction, and this is where the answer to Chris's question actually lives. In July, 79.5k homes exchanged and completed while 75.6k withdrew from agents' books unsold. On the current cut, 51.3% of homes leaving the books in July sold and 48.7% gave up. The seven-year average is 57.6%, and I said last week and will say again that I worry about how much 2021 and 2022 flatter that average, but even against a more sober benchmark, half of everything that leaves an estate agent's window in this market leaves without a sale. I keep quoting this stat weekly because it is the single best corrective to the "market is fine" and "market is dead" headlines alike. Both are true. It is fine for correctly priced stock and dead for the rest, and Chris's show this week puts a number on "correctly priced": around eight in ten of the sellers who have achieved a sale this year did so WITHOUT reducing their asking price. In other words, the reductions column is not where the sales come from. The sales come from the homes that were priced to sell on day one. The reductions column is where the homes go that were not. The agents - often incentivized, especially at the corporates, by getting listings - are sensible and good at their job - they know when they can go on at the price that sells it because they have a seller who is motivated, and they know when they’ve got to puff the price up in order to get that listing (and they probably have already written that one off in their head, in the current market, but they will take their listing bonus regardless). 

The pricing evidence this week is the sharpest it has been for a month, and the anomaly I flagged twice in August has reversed exactly as promised, if a week earlier than I expected. 20.9k price reductions this week on 767k homes for sale. 13.7% of the entire stock was reduced during July, down a touch from 14.3% in June, with the 2026 year-to-date average at 13% against a six-year long-term average of 11.2%. And the gap between the average asking price of all listings and the average asking price of the homes actually going sale agreed has blown back out to 12.6% on Chris's current measure (£393k listed against £349k agreed), from 7.8% last week and 9.9% the week before, against a longer-term average he puts at 16-17%. Let the pedants, like me, note that 12.6% is the gap measured on the sale-agreed base; on the listing base it is 11.2%. Either way it is a five-point move in a week, and what happened is simple enough: the late-August listings were the cheap ones and the September listings are not. I said do not conclude too much from a late-August reading. I will now say do not conclude too much from an early-September one either, but the direction is the one that matters: the stock coming to market in September is priced further from where buyers are transacting than the stock that came in August. Whether that is agents' optimism, sellers' optimism or simply the mix of what sells in a school-holiday week, it is hard to say; probably all three. Next week that number will be heading towards 25% (depending on when the London schools are back) - but this is why we don’t get overexcited week-to-week. 

Here is this week's back-of-the-envelope, and I will hedge it as rough working as ever, because it stacks three of Chris's numbers that were not designed to be stacked. The sell-through rate, the share of homes on agents' books going sale agreed in a month, was 14.2% in July against a pre-Covid average of 15.5%. Compounded over a typical twelve-week sole agency, that is roughly a 37% chance of going under offer inside the contract, as it was last week. Now add Chris's eight-in-ten. If 14.2% of a 767k book is around 109k homes going under offer in a month, and around eight in ten of those did it without a reduction, then roughly 87k sold at their original price and roughly 22k sold after a cut. In the same month, around 105k homes were reduced. So, in the roughest of round numbers, of the homes that took a price reduction, something like one in five found a buyer inside the month, against the one in seven for the book as a whole. That is better odds after a reduction than before, which you would expect, but it is not the transformation sellers imagine when they finally agree to "come down a bit". The homes that sell are overwhelmingly the ones that never needed to. If you are selling, that is the memo: the price you launch at is the sale, and the reduction is the consolation prize. If you are buying, the 22k are your people: a seller who has already moved once has, in my experience, usually decided to move again. The second reduction is nearly as good as the second fall-through (when that happens) - but much more regular of course. 

The quietly encouraging line continues: the fall-through rate came in at 23.7% against a decade average of 24.5%, so BELOW the norm for the second week running, and only 5.32% of homes sold subject to contract in July actually fell through, against a 2025 average of 5.3% and a ten-year average of 5.8%. Fewer offers, but better offers, from buyers who have done their sums, and the deals are sticking. Stock stood at 767k on 1st August against 760k a month earlier and 763k a year ago, flat, and the pipeline of agreed sales at 487k against 508k twelve months ago: the pipeline is 63% of the stock against 67% a year ago. Same size bath, thinner outflow, and the agent knows it, which is your negotiating context if you are on the buying side.

On pricing itself, July's agreed sales averaged £345.41 per square foot, 1.2% ahead of twelve months ago and 11.9% ahead of five years ago. Sub-inflation nominal growth with CPI at 2.9% is a real-terms fall of about 1.7% on the year, which is improving affordability by the back door. Chris's £-per-square-foot at sale agreed matches the Land Registry index with 98% accuracy five months in advance, which is why it matters more than any of the headline indices you will read about in the Macroscope below, one of which has just quietly rewritten its July number.

The rental side: £1,770 per month was the average asking rent in week 33, with August 2026 at £1,788 against £1,800 in August 2025, so essentially flat to fractionally down year on year on Chris's portal measure of asking rents on new listings, the most real-time cut there is. Against that, £1,394 in August 2021 tells you what the last five years did, and the 2026 year-to-date average of £1,765 tells you that this year has been the plateau. Availability continues to improve at the margin: 323k rental properties available in August against 319k a year ago, with 135,928 new rental listings in the month against 128,821 and 111,080 in the comparison periods. The stock-shortage era is easing at the edges, rents have stopped rising nationally on the new-listing measure, and the ONS index of all tenancies is still catching up with rises already banked. Affordability ceiling, or supply constraints easing? A bit of both, still, and neither is going away by Christmas. Proof of last week’s individual landlord numbers surprisingly still rising, in spite of the tax position and the likes (we do need to remember how many are unencumbered, of course) - without limited companies being measured in those Government figures. Stock is on the rise, or at least on the basis of all the evidence we can see (demand is down too, which we do know from Zoopla et. al. and more people are staying at home for longer from anecdotal evidence, although more research and measurement is required there). 

Chris - this is my weekly appreciation paragraph, and this week it comes with an extra thank you for sending the numbers over from wherever you are. Thanks for what you do! If you want some help positioning yourself as a local market expert - as an estate agent or any form of property professional - give Chris a shout! Either way give his channel www.youtube.com/@christopherwatkin a follow and some love, please! This week Simon Gates joins Chris to pick apart the Week 34 figures and ask why a functioning market feels like a struggling one, before heading to Oldham to run the data ruler over the town's estate agents, which of them are actually getting their sellers moved, and then the lettings market as well. Watch the Week 34 show here: https://youtu.be/M5sduCfU7iQ.

Dust off the Macroscope, then. A proper week after the bank holiday lull: Nationwide's August house price index, with a footnote that rewrites July; the Bank of England's Money and Credit report for July, with the lowest mortgage approvals since January 2024 and the dearest new mortgages in over a year; the final August PMIs, my darlings of the real time economy, all four of them, with the construction one doing what the construction one does. Bringing up the rear the contract states that we have to talk about the gilts and swaps, and will do so for many years to come, I'm sure, and this week they earned the space.

OK. Nationwide. UK house prices rose 0.2% in August on a seasonally adjusted basis, with annual growth at 1.6% and the average price at £275,465. That is "little changed" in Robert Gardner's words, "subdued" in the headline, and about what the sale-agreed data has been telling us since June. But read the notes to editors before you read the numbers, because Nationwide has revised July. A "system change" led to some cases being left out of the original July sample, and the corrected July figures are a monthly change of minus 0.1% rather than plus 0.1%, and annual growth of 1.4% rather than the 1.8% I reported to you at the start of August. So the August 1.6% is not a slowdown from 1.8%. It is a small pick-up from a July that was weaker than anyone, including Nationwide, thought at the time. The village watchman strikes again, and it is worth noting that he has struck twice in a fortnight now, once on Ofgem's typical household and once on a house price index, and both times the revision made the recent past look worse and the present look better by comparison (and there’s plenty more to come on productivity figures, don’t worry about that!). I am not suggesting anything sinister. I am suggesting you keep a copy of the number you were given, because the number you are given next month may be talking about a different July.

Underneath the headline, the average price has fallen around £3,400 from its April peak of £278,880 and sits at its lowest since February, which is a more useful sentence than "up 1.6% on the year". Gardner's commentary is worth a moment, because it is the most explicitly two-handed a banking sector chief economist has been all year. On the one hand: geopolitical tensions high, the conflict exerting upward pressure on energy prices and market interest rates, and Bank Rate expectations "volatile". On the other, and this is the bit I would underline: "encouraging signs" that the energy shock is not feeding through to underlying price pressures, private sector wage growth easing further, which "should give policymakers breathing space" before deciding how much tighter policy needs to be. That is a mortgage lender telling the MPC, politely, not to hike. He also notes that affordability is improving because prices are growing well below earnings, with some of the gain given back through higher mortgage rates, and that activity "should regain momentum" if the energy shock wanes and market rates fall back towards pre-conflict levels. Two ifs, and neither of them was looking good on Wednesday. The special report on National Parks (a 24% premium, £66,500 in cash terms on the Q2 average; 6% within five kilometres; 14% in a National Landscape) is charming and, I suspect, mostly a planning story: you cannot build in a National Park as a rule, so the houses that are there are worth more. Scarcity, again, dressed as scenery.

Next up, the Bank of England's Money and Credit report for July, which landed on Tuesday and which the trade press called a two-year low and I would call the retail buyer stepping back at precisely the moment the £-per-square-foot said prices had stopped falling in nominal terms. Net mortgage approvals for house purchase fell to 56,100 (56,053 unrounded) from 58,200 in June, the lowest since January 2024 and about 8% below the six-month average of 60,800 (and well adrift of my “healthy market” figure of 65k pcm, and this was not war-related, remember - this cooling started at the start of 2026 - it’s easy to conflate the higher interest rate with the market cooling and give Mr Trump the lion’s share of the blame, but this history of that does not quite stack up - although it has of course been a contributory factor). Net borrowing of mortgage debt fell to £4.3bn from £7.7bn, below the six-month average of £5.3bn, although the annual growth rate of mortgage lending held at 3.6%. Remortgage approvals nudged up to 34,500 from 34,100. And the number that connects this section to the last: the effective rate on newly drawn mortgages, the rate people actually paid, rose to 4.45% in July from 4.35% in June, the highest in more than a year, while the rate on the outstanding stock ticked up to 3.97%. So in the same month, fewer people were approved, they borrowed less, and those who did borrow paid ten basis points more than the month before. Approvals are the buyer count a little time ahead of Chris's sold-subject-to-contract number, which is itself four to five months ahead of the Land Registry, so what July's approvals are telling you is roughly where the SSTC data might be in August/September and where the official price indices will be at the end of the year or towards the end of the winter. Subdued, is the word, and I would expect the September approvals print to be worse than this one given what the swap market did this week - it’s almost a guarantee, especially as the bond markets wobble and the Prez has changed the chatter to “no deal” (although perhaps that’s just his next step in getting a deal, of course). 

The rest of the release rhymes. Consumer credit rose £2.0bn, a touch above its six-month average, with credit cards at £0.9bn and other lending, car finance and personal loans, at £1.1bn. Households put £2.2bn into ISAs and pulled £3.5bn out of interest-bearing sight deposits, which is the savings ratio being drawn down at the margin, or people moving cash to where the rate is, or both. Companies raised close to no net finance in July after repaying £2.4bn in June, and the broad money supply measure M4ex fell £10.1bn on the month, driven by the non-bank financial sector shrinking its holdings by £14bn. I am wary of reading a single month of M4ex, and Woodford, who reads it every month, has been arguing for a year that the monetary aggregates are quietly signalling something better than the surveys. July is not one of those months. It is a month in which the household sector borrowed a little more and saved a little less and the housing market borrowed a lot less, which is a consumer holding up and a buyer holding back, and that is a reasonable one-line description of the whole summer. But why draw down in July on savings? Holidays. Tax bills. That’s why. So as usual, don’t sweat too much over one month’s number.

Onto the PMIs, my darlings of the real time economy, and the final August readings which I trailed last week. Manufacturing first, revised UP to 51.7 from the 51.5 flash: a tenth consecutive month above 50, output and new orders and export demand all rising, business optimism at a six-month high, and employment growth at its fastest for two years. Two years. In a sector that spent 2024 and 2025 shedding people, that is the number I would want explained before I believed any of the others, and the explanation the survey gives is order books and confidence. Services, revised DOWN to 52.5 from the 52.8 flash, still the highest since April, still up from 52.1 in July, with new work rising for a second month and business expectations at their highest since February, "close to those seen just prior to the Middle East conflict" in Tim Moore's phrase. Two things underneath that. Employment in services fell again, at the slowest pace since October 2025, but that is now twenty-three consecutive months of falling service sector headcount, the longest run since the survey began in 1996, and firms are explicitly citing hiring freezes and "efforts to boost productivity through automation". Keep that phrase for the Deep Dive; it is Woodford's thesis in a survey respondent's own words. And prices: input cost inflation accelerated sharply from July's five-month low on fuel and transport, and output charge inflation accelerated for the first time in four months as firms protected margins. So the services sector is growing, shedding staff, and putting its prices up. Composite at 52.5 from 52.2, the highest since April, and the all-sector reading including construction at 51.8, a six-month high.

Then construction, which fell to 44.3 in August from 44.7 in July against a consensus of 45.5, the twentieth consecutive month of contraction. Housing was the reason: the residential index fell to 37.6, the only one of the three sub-sectors to contract faster than in July, while commercial (47.8) fell at its slowest rate since January and civil engineering (40.5) at its slowest since March. Survey respondents cited "a reduction in new projects, especially house building starts". Two softer lines: new orders fell at their slowest pace since September 2025, and job losses were the slowest since September 2025 too, with subcontractor usage rising for the first time in around two years. And one line that matters for the CPI: construction input price inflation fell to its lowest since February. Reuters made the obvious political point, that a housebuilding index at 37.6 is a challenge for a Prime Minister who wants to expand the public housing stock, and I made it at length in the 9th August construction special so I will not repeat it. I will just note that the manufacturing PMI is hiring at a two-year high and the construction PMI is in its twentieth month of contraction, and that those two facts describe the same economy at the same time. Whatever is being built in Britain this year, it is not houses.

Gilty, or not Gilty? The court reconvenes, and this week the accused is the front page. The five-year gilt closed the week at 4.630% against 4.638% last Friday, down a basis point or so on a Friday-to-Friday basis, having spent Wednesday morning at around 4.78%. The thirty-year closed at 5.775% against 5.80% last week, also fractionally lower on the week, having touched around 5.94% on Wednesday. The ten-year, which is where the headlines were, touched 5.29% on Wednesday, the highest since August 2007, and closed the week at 5.13%. So the shape of the week was a spike and a full retracement at every point on the curve, and if you only read Friday's closes you would think nothing had happened. Twelve months ago the five-year stood at a little over 4.1%, so we remain 50-odd basis points on the wrong side, and the fifty-two week range on the five-year (3.668% to 4.807%) tells you we are comfortably in the top quarter of the year's range and have been since March. The thirty-year's range is 5.033% to 5.944%; Wednesday was that high. The Treasury bus, once again: we did not vote for the driver, we cannot reach the pedals, and this week he took a roundabout at speed and then slowed down again and everyone upstairs was sick anyway.

Swaps did not retrace, and that is the sentence that matters for anyone with a completion this autumn. The five-year SONIA swap finished the week at 4.36% against 4.30% last Friday, having touched 4.46% mid-week; the three-year at 4.29% against 4.23%. A month ago they were 4.22% and 4.16%. A year ago they were 3.82% and 3.74%. So the mortgage-pricing curve is 14 basis points higher than it was at the start of August and 54 higher than a year ago, and while the gilt market gave back Wednesday's panic by Friday, the swap market kept about half of it. Lenders price off swaps, not gilts, and a fortnight of repricing letters is, I would suggest, now baked in. Rightmove's tracker already has the all-in average fix above 5%; I doubt it comes back under this side of the MPC. Bank Rate, for the avoidance of all doubt, remains at 3.75%. The 4.00% you keep hearing is the level three MPC members voted for in July, the level the curve now has priced for December, and, per Huw Pill's remarks this week, the level the Chief Economist would rather go to now than be forced to go past later. It is not the current rate - but it may well be by the end of the month. Without panic, there will be some 6% debt still around, but it is sensible to price in 6.25% as the price of your limited company 5-year fixed debt right now, with the usual caveats for “quirky” (add 0.5% - 1%). 

What DID change this week, and it is the most important thing in this section, is the market's timing, in the wrong direction. Last Sunday I told you the oil sell-off had pushed the expected next Bank of England move from late 2026 into 2027, with less than four basis points priced for the 17th September and a roughly 15% chance of a hike. This Sunday the market has a full hike priced by year-end and another by March 2027, on the back of one week of oil, one gas spike, one Fed Chair with work to do, and one Chief Economist who has voted for a hike at every meeting since April and said out loud why. My read of the 17th, held with less conviction than last week: about 66-34 between a hold and a hike, 5-4 either way, although 6-3 hold remains the third potential option. The doves' case has not changed and Nationwide made it for them on Tuesday: private sector wages easing, underlying inflation not showing the energy pass-through, a housing market at its lowest approvals in two and a half years. The hawks' case has got louder: gas at a 2023 high, services firms putting prices up, and a market that has now moved to meet them. Friday’s close only gives 11% chance of a hike, and that feels a little skinny at this stage, although I suspect the market is searching for who the fifth “hiker” would be given the dovish nature of a few members of the committee and the Governor’s position, up until now, being fairly clear that he wants to hold. The QT envelope decision the same day still matters more for long-end gilts than the vote itself, and I will say now what I said a month ago: if the Bank trims the active sales of long gilts, Wednesday's 5.94% on the thirty-year may prove to have been the top. If it doesn't, it wasn't. I’ve made no secret over the past years that I think the Bank has done a terrible job on the quantitative tightening and wasted an absolute fortune - so I don’t expect them to draw this parallel this year either, to be honest. What this means in practice, as ever: if you have completions between now and the Budget, the case for securing rates rather than surfing the curve got stronger this week, not weaker, and the swap market has already sent that memo to your lender. This is where things differ from June 2023 - at that point, I felt quite clear that the top of the curve was with us when the 5 year was dancing around 5% yield - I’m not as clear about that right now, especially if every time Burnham speaks, the bond markets drift. Boring advice, again, but boring is useful I think. It keeps ageing well, and this was a week in which it aged about a fortnight in five days. Remember - Stress tests now exist around 7.5% for stock you are holding in your portfolio - that number is higher than I’ve been on it since the days of Truss, and that wobble passed very quickly - this one may not. 

OK. Here endeth the lesson on current rates - here comes the Deep Dive. The through-line this week is forecasting, and specifically whether the people whose forecasts set your interest rate and your tax bill have been working from the wrong numbers. Neil Woodford (notorious fund manager, for those who don’t recognise the name - who is back in the spotlight with a new LinkedIn Newsletter and accompanying YouTube channel which reminds us, apart from anything else, just what a great talker he is) says yes, at length, and the ONS has, in its own careful way, half agreed with him. The Treasury's monthly comparison of independent forecasts says the consensus is at 3.3% for Christmas, not the 4-and-5s you hear on the radio. The Resolution Foundation, no friend of Woodford's politics, has run the productivity numbers and found the same thing he did. And the gilt market, as we have just seen, has spent the week betting against all of them. Let's get into it.

This is somewhere I have been meaning to go for a while, and Neil Woodford has given me the excuse. Four documents, all published in the last three weeks, all circling the same question: has Britain's economy been better than the people measuring it thought, and if so, what does that do to the inflation and interest rate forecasts that are, right now, pricing your next mortgage? We start with Woodford's own case, made across three pieces on his Noise Cancelling site. Then the ONS productivity release of 18th August, the primary document he leans on, read in full rather than in his summary of it. Then the Treasury's monthly compendium of independent forecasts, which tells you where the consensus actually sits as opposed to where the radio says it sits. And then the Resolution Foundation's quarterly outlook, which has run the same productivity numbers from a very different starting point and arrived, to my slight surprise, at the same place.

Source 1: Neil Woodford, Noise Cancelling: "The great UK productivity myth" (28th August), "Bond hysteria, a leaky Hormuz and fiscal incontinence" (4th September) and "Inflation myths: the surge that never came" (4th September)

Theme 1: The Measurement

The Summary: Woodford's central claim is that the ONS has understated UK productivity for at least two years because the Labour Force Survey has overstated the number of people in work, and therefore hours worked, which is the denominator in the productivity calculation. He traces his own paper trail back to November 2024 ("businesses don't pay people more to produce less") and dates the divergence between the LFS and HMRC's PAYE Real Time Information to around mid-2023. The turning point in his account is the ONS's 18th August release, in which the statistics office says it is developing a "component-based" approach aligned with RTI employment. On the RTI measure, output per hour rose 0.7% in the year to Q2 2026 and output per worker 1.4%; on the LFS measure, output per hour fell 0.2% and output per worker rose 0.4%. Against the 2019 average, RTI has output per hour up 4.6% and the LFS 2.3%: twice the growth.

The Propenomix Perspective: I am predisposed to agree with him on the LFS, having grumbled about that survey in the labour market section for two years, and the ONS numbers, which we read for ourselves in a moment, support him. A caution about the word "half", though. The 2.3% against 4.6% is measured from a 2019 base and the two series only parted from mid-2020, so "twice the growth since 2019" is true and "twice as productive" is not: it is a gap of around two percentage points in the LEVEL of output per hour, built over three years, which is a lot in productivity terms and not much in any other. And it is a story about the labour input, not output. GDP is identical in both series.

Theme 2: What It Does to Policy

The Summary: Woodford argues the mismeasurement propagates into both arms of policy. For the Bank, productivity growth is an increase in potential supply that raises the economy's non-inflationary "speed limit" and feeds its estimate of the neutral rate; an understated series therefore makes current policy look less restrictive than it is. For the OBR, he points to last November's downgrade of medium-term productivity growth from 1.3% to 1.0%, which lowered the growth forecast, worsened the end-of-parliament fiscal position and, in his account, led directly to Rachel Reeves's second round of tax rises; the OBR assumes trend productivity grew 0.3% in 2024 and 0.7% in 2025 and does not reach 1% until 2030. His prescription: the Bank should recalculate its speed limit and cut later this year; the OBR should rebuild its growth forecasts; and a Burnham government would, in his view, spend any headroom rather than return it.

The Propenomix Perspective: This is the part I find hardest to argue with and easiest to worry about. The mechanism is right: wrong denominator, wrong speed limit, an MPC more worried about capacity than it needed to be. Where I would hedge is the leap from "the OBR's assumption was too low" to "the tax rises were unnecessary". The November downgrade took the assumption to around 1%, and, as Source 4 will show, the Resolution Foundation's read of the corrected data is that productivity is growing at about 1.1% a year, which would make the OBR right for the first time in its existence rather than wrong by half. The downgrade may have been the correct number arrived at from the wrong direction. That is not nothing, but it is not the £20bn error Woodford implies. On the Bank I am closer to him. Huw Pill has voted for a hike since April on the grounds that the economy has less slack than it looks; if the series he was reading understated supply by two points, the slack was there and the vote was wrong, and this week's remark about hiking now to avoid hiking more later reads differently in that light. But the case that it should not be hiking got stronger, and it was made by a statistics office rather than a fund manager, which is the bit that ought to worry the hawks.

Theme 3: Inflation, and the Yields

The Summary: The forecasters, Woodford says, predicted at the outbreak of the Gulf war that UK inflation would pass 4%; it fell instead, to 2.6% in June before the Ofgem-driven 2.9% in July. Now, he argues, the same forecasters say the country has entered a new inflationary age, and the data says the opposite. In the Weekly he accepts inflation "will likely pick up a bit" from 2.9% but calls talk of 4-5% by year-end "ridiculous", noting that Brent peaked at $120 earlier in the year and inflation fell anyway; his July briefing had CPI peaking a little over 3% in September, below 2% in the second half of 2027, and Bank Rate below 3% next year. On yields, he rejects the view that gilts are high because the market dislikes fiscal policy: ten-year gilts are high because ten-year Treasuries are, Treasuries are high because US inflation is 3.4% and the new Fed Chair has said he is worried, and the 40 basis point spread on Thursday (5.16% against 4.76%) has been stable enough to prove it.

The Propenomix Perspective: This is where my sympathy is strongest, because I lived through 2021 in this newsletter and I remember who got it wrong and in which direction. Woodford's argument is that it is now wrong on the way down by the same mechanism, forecasters extrapolating the last shock into the next, and the war evidence supports him more than I expected: $144 physical crude in April on Woodford's account, $120 Brent, and a CPI that went DOWN through the summer. My own view sits nearer his 3% than the radio's 4%: I have said since August that the Bank's 3.2% peak looked about right and possibly generous (and remember, this had been revised downwards from forecasts from the Bank earlier in the year), and nothing in the services PMI's price lines this week moves that by more than a couple of tenths. On yields I am with him on the mechanism and wary of the conclusion. Gilts go where Treasuries go, and this week proved it again. But a 40 basis point premium is not nothing; it is roughly the cost of being a country whose Chancellor has to give a speech on Monday explaining why the fiscal rules still hold. The Treasury sets the level; the premium is ours; and the premium is the bit the Budget can move. It’s ridiculous in comparison to the US, frankly, although we do share some of the same problems including the demographic pyramid. Both need to take control of the spending side - and that seems more likely in the UK than the US right now, but still unlikely in the US unless Healey reads the runes and manages to persuade Burnham, because he clearly doesn’t get it at the moment. 

Theme 4: The Synthesis, and the Hole

The Summary: Read together, the three pieces make one argument. Productivity has grown at twice the rate the official series showed; the Bank and the OBR have set policy off the wrong number; inflation is therefore lower-trend than the hawks believe, and the forecasters, wrong in 2021 in one direction, are wrong in 2026 in the other; gilt yields are imported from Washington rather than earned in Westminster; and the correct response is lower rates and no further tax rises. Woodford acknowledges in passing that the Prime Minister "is focused on doing what he can" on the cost of living and might take VAT off gas, which "would of course keep inflation down too". He does not fold that into his inflation arithmetic or into his case against the forecasters.

The Propenomix Perspective: And that is the hole. Woodford's case is that inflation will undershoot the forecasters because the economy is stronger and the shock smaller than they think. But the July CPI at 2.9% already includes the April removal of most of the electricity levy into general taxation; the October print will include VAT coming off electricity from 1st October; and if Burnham does to gas what he did to electricity, the January print will include that too. None of that is economics. All of it is policy, funded from general taxation, that mechanically lowers the measured price of the thing going up. Woodford’s pushback will be that these prices are half the contributor to inflation that they were in the 1980s, although he almost rubbishes “second order effects” because food inflation has NOT reflected the price of fuels increasing as much as they have (again, that’s correct, but of course it was c.20% within recent enough memory which was worse than any other inflation numbers we can remember - it made rental inflation look low - so there is some protection baked in there, you’d think). Woodford may well be right that inflation lands nearer 3% than 4%, and the reason may be that the Prime Minister has decided it will, which is a different claim from "the data says the opposite". The politics of “fixing the cost of living” - and Woodford, so far, has stayed away from the politics - are having a non-zero impact. More intervention brings the headline down further; it also pushes the reversal into April 2027, when the VAT goes back on in the same month the uprating figures are set, which I flagged last week and will keep flagging. I am not trying to ascribe politics to Woodford's forecast. I am saying that a forecast of measured CPI in a year when the government is actively managing measured CPI is a forecast of the government's intentions as much as the economy's, and the honest version of the argument says so. Where does that leave me? Sympathetic to the number, sceptical of the reason, and more convinced than I was on Monday that the MPC is looking at a headline it does not fully control. We'll see.

Source 2: Office for National Statistics, "Productivity flash estimate and overview, UK: April to June 2026 and January to March 2026" (18th August 2026)

The primary document, then, read in full, because a fund manager's summary of a statistics release is a fund manager's summary, however good, and the release says things the summary does not.

Theme 1: What the Numbers Actually Say

The Summary: The ONS's recommendation is unambiguous: until its component approach is published, the Real Time Information-based measure is "the best measure of changes in labour productivity". On that measure output per hour was 0.7% higher in the year to Q2 2026 and output per worker 1.4% higher; on the LFS measure output per hour fell 0.2% and output per worker rose 0.4%. Against the 2019 average, RTI has output per hour 4.6% higher, a 7.1% rise in gross value added against a 2.4% rise in hours; the LFS has it 2.3% higher, the same 7.1% rise in output against a 4.7% rise in hours. RTI is an administrative near-census of payrolled employees, supplemented by LFS self-employment data; it does not collect hours, so both measures multiply an LFS average-hours figure by their respective headcounts.

The Propenomix Perspective: Woodford has the numbers right, to the decimal. But keep hold of the last sentence: RTI does not measure hours. Both series still take the LFS's word for how long each person worked, and the LFS is the survey whose phone room was under-resourced for six weeks in the very quarter under discussion. The ONS is clear the correction for that is small, and the Resolution Foundation, in Source 4, puts the hours effect at about six minutes a week, which is reassuring (but non-zero). The 7.1% rise in output since 2019 being identical in both series is, again, the tell: the whole argument is about how many hours it took. Which is the argument that matters for the Bank, so fair enough.

Theme 2: The ONS's Own Explanation

The Summary: The ONS's account of the divergence is not quite Woodford's. It says enhancements to the LFS since January 2024 mean recent movements reflect "both underlying developments in the economy and improved survey quality, including obtaining more responses from employed people", and that as survey improvements bring measured employment into line with other indicators, "recent LFS measures of employment growth are likely to be temporarily higher while recent LFS measures of productivity growth are likely to be temporarily lower". The LFS, in other words, is overstating employment GROWTH because it is catching up with people it previously missed, not necessarily because its level is wrong now.

The Propenomix Perspective: Read carefully, this is a subtler story than "the survey was broken and the tax data was right", and it matters for how much weight you put on the correction. Woodford's version is that the LFS overstated the number of people working. The ONS's version is that the LFS under-counted them for a while, has been fixing that since January 2024, and the fix shows up as artificial employment growth for as long as the catch-up lasts. The two stories point the same way for the last two years of productivity GROWTH, which is why RTI is the right series to use, but different ways for the LEVEL: if the survey was missing employed people before, the pre-2024 productivity level was, if anything, overstated, and the true gain since 2019 sits somewhere between the 2.3% and the 4.6%. I am not entirely sure which reading is right and, refreshingly, neither is the ONS, which is why it is running the linked-data test. So the direction of Woodford's argument is now official and the magnitude is not, and the methods article the ONS has promised this month, before the Budget and before the Blue Book on 30th October, is a release worth putting in the diary. Let the pedants, like me, argue about the level; the MPC should be arguing about the growth rate, and there the ONS has conceded the point.

Theme 3: The Quarter Nobody Mentions

The Summary: Table 1 of the release shows the RTI series quarter by quarter. Output per hour against its 2019 level stood at 3.9% in Q2 2025, 4.3% in Q3, 4.3% in Q4, 5.5% in Q1 2026 and 4.6% in Q2 2026, a fall of 0.8% on the quarter in the latest reading; the annual growth rate ran 1.9%, 3.1%, 1.4%, 2.1%, 0.7% across the same five quarters. Output per worker, by contrast, rose 0.6% on the quarter and 1.4% on the year, its level against 2019 rising steadily from 3.2% to 4.6%. The ONS notes that new GVA estimates are "more volatile on a quarterly basis, especially in production industries", and advises users to focus on long-term trends because productivity is a structural feature of the economy.

The Propenomix Perspective: Here is the bit that does not make the podcast. On the measure the ONS now prefers, productivity FELL in the second quarter of 2026, by 0.8%, and annual growth dropped from 2.1% to 0.7%; the Q1 reading, 5.5% above 2019, now looks like the peak. Two explanations, and I suspect both apply: Q2 was the quarter of the war and the $120 Brent, and output per hour in an economy whose energy bill has just doubled is going to wobble; and an LFS phone-room under-resourcing error sat inside the quarter, which the ONS says accounts for a tenth of it. But the honest reading is that the productivity revival is real over two years and was not accelerating in the latest quarter. Why does it matter? Because Woodford's inflation call rests partly on a supply side growing faster than the Bank thinks, and if the latest quarter says 0.7% not 2%, the speed limit is higher than the MPC assumed but not by as much as the LSE's 1.6% or Bloomberg's 2% suggest. The argument is right, and the most recent data point is the weakest it has offered. I would rather know that than not. The output-per-worker series, which does not depend on the LFS hours figure at all, kept rising, which is some comfort and possibly the better guide until the hours are sorted. No-one can genuinely make a case for a cut in the base rate right now - the bond markets would lose their minds. 

Theme 4: The Synthesis: Where the Productivity Is

The Summary: The industry breakdown, on the LFS basis for Q1 2026 against 2019, shows information and communication making the largest upward contribution to output-per-hour growth (2.8 percentage points) and human health and social work the largest negative (minus 0.9 points). Growth in information and communication was driven mainly by higher GVA rather than fewer hours. The next release is 17th November; the component-approach article is due in September; the quarterly national accounts on 30th September and the Blue Book on 30th October will revise the output side.

The Propenomix Perspective: So the productivity Britain "found" is disproportionately in information and communication, the fastest-growing part of the economy for a decade, and the productivity it "lost" is in health and social care, where the state has been putting its people. That is not an AI story yet, whatever the podcast titles say; the Resolution Foundation is blunter about it in Source 4. It is a composition story in which the private sector's most digital industry carries the number and the public sector's most labour-intensive one drags it. What does this mean for the 17th? The MPC will not have the methods article in time, will have the RTI flash it already has, and will be looking at a services PMI whose respondents are cutting headcount and citing automation. If I were Pill I would read that as an economy that can grow without hiring, therefore without wage pressure, therefore without a hike. If I were Mann I would read the same survey's output-price line and vote the other way. Onwards, to the forecasters.

Source 3: HM Treasury, "Forecasts for the UK economy: a comparison of independent forecasts", No. 469, August 2026 (19th August)

Every month the Treasury collects the published forecasts of around two dozen City and non-City houses and prints them side by side. The August edition contains 16 forecasts received between 3rd and 14th August plus older ones from June and July, so it predates the Ofgem cap, Jackson Hole and this week's strikes. The next edition is out on 16th September, the day before the MPC.

Theme 1: Where the Consensus Actually Is

The Summary: The independent average for CPI in the fourth quarter of 2026 is 3.3%, down from 3.4% in July, with the forecasts received in August averaging 3.4%. The range across forecasts made in the last three months runs from 2.1% (Bloomberg Economics, a December forecast) to 4.3% (Liverpool Macro Research, June). Of the twenty-six institutions listed, only two are in the fours, both dated June: Liverpool Macro and the OECD, at 4.3% apiece. The City houses cluster tightly: Barclays 3.2%, Deutsche 3.1%, Pantheon 3.1%, UBS and NatWest 3.3%, HSBC, JP Morgan and KPMG 3.4%, Capital Economics and EY 3.5%, Nomura 3.6%. Among the non-City houses, NIESR and the CBI are at 3.6%, the British Chambers at 3.8%, CEBR at 2.9%. For Q4 2027 the average is 2.3%, ranging from 1.6% (Nomura) to 4.1% (Liverpool again). The medium-term annual averages run 3.1% for 2026, 2.7% for 2027, 2.2% for 2028 and 2.0% by 2030.

The Propenomix Perspective: So here is the robustness check on Woodford's claim that "the same people say we have entered a new inflationary age". On the Treasury's own panel, they do not. The centre of gravity is 3.3-3.4% for Christmas, a couple of tenths above the Bank's 3.2%, and the only forecasts in the fours are two months old and from the two houses with the least skin in the mortgage market. So who has been saying 4s and 5s? NIESR’s blog in March suggested that inflation would be 5-6% if the conflict wasn’t resolved soon, by the end of 2026. April’s Bank of England forecast suggested 3.5% or a little over, apart from “Scenario C” which I detailed at the time which said 6.2% by year end. EY’s item club suggested 6.4% inflation by year end if Hormuz stays shut - the counter would be that of course it has not stayed shut, as detailed in Trumpwatch. Woodford’s point - that the oil price is drastically overweighted in these forecasts - does have some validity - which leaves him arguing with a consensus that is closer to him than he says. That is not a criticism of the argument; 3.3% is still above his "a little over 3%", and the direction of the miss is the point. It is a criticism of the framing. The forecasters are not predicting a surge. They are predicting a plateau a few tenths above where Woodford would put it, and the difference between 3.1% and 3.4% in December is, roughly, one VAT cut on gas. I’d have thought that to make a big deal about it he would be predicting closer to Bloomberg’s very hopeful 2.1% by year end. All this takes us straight back to the hole.

Theme 2: Rates, Oil, and the Four Houses at 4.00%

The Summary: The panel's average for Bank Rate at the end of 2026 is 3.7%, median 3.75%. Four houses have it at 4.00%: JP Morgan, NatWest Markets, Beacon Economic Forecasting and Heteronomics. Liverpool Macro has 3.25%; Berenberg, Bloomberg Economics and the ITEM Club 3.50% on older forecasts; everyone else 3.75%. For end-2027 the average is 3.46%, from 2.75% (Liverpool) to 4.25% (NatWest), with Capital Economics, Berenberg and RSM at 3.00% and NIESR and Pantheon at 3.75%. Oil assumptions for 2026 average $87 a barrel, from $75 (Capital Economics) to $96 (Barclays). Q4 2026 unemployment is forecast at 5.2%, and productivity, output per hour, at 0.7% for 2026 and 0.6% for 2027.

The Propenomix Perspective: Two observations. First, the market has run a long way ahead of the forecasters this week. The panel has Bank Rate at 3.75% at Christmas with a hawkish minority at 4.00%; the swap market, as of Friday, already has two hikes priced in. Second, look at the oil assumption: $87 for the year. Brent is $96 as I write and was $88 a week ago. If the war stays at this week's temperature the oil assumption is fifteen dollars light and the inflation forecasts a few tenths light with it; if it goes back to last week's, they are about right. The 3.3% consensus is an $87 oil consensus, and the fours are a $110 oil scenario the two June forecasters happened to be running when Brent was near that. Both are forecasts of the war, not of the economy. And the productivity line made me smile: the panel has output per hour growing 0.7% this year, which is exactly the ONS's preferred number for the year to Q2 and below the Resolution Foundation's 1.1%. On this variable the "consensus" is already halfway to agreeing with Woodford. Nobody has told the gilt market.

Theme 3: 2021, and the Man Who Was Right

The Summary: 2021 is the clearest recent case of the whole forecasting community missing an inflation turn in the same direction at once. In the spring of 2021 the Bank's Monetary Policy Report projected inflation rising temporarily above target and returning within a year or so. Transitory inflation. The Treasury's comparison at the time had the Q4 2021 consensus in the low twos. The outturn was 5.4% in December 2021 and 11.1% at the October 2022 peak. The dissenter was Andy Haldane, then Chief Economist, who voted alone in May and June 2021 to curtail asset purchases and, shortly before leaving the Bank, warned that inflation could be nearer 4% than 3% by year-end. This week Haldane was reported advising the Prime Minister to avoid the "hat-trick", a third consecutive tax-raising Budget.

The Propenomix Perspective: I keep quoting 2021 because I lived it in these pages, and I was nearer the Haldane camp than the consensus and still nowhere near the 11% we got until early 2022 when I suggested double digits were on the cards, a long way before the consensus did. The lesson I took was not "the forecasters are useless" but "the forecasters are consistent", which is worse. They erred together, in the same direction, because they used the same models with the same assumption that shocks fade, and the shock did not fade. Woodford's argument is that they are doing it again in reverse, and the HMT table gives him half a point: the consensus drifted DOWN over the summer, from 3.4% to 3.3%, while the market moved up, and the forecasters were closer to right than the market in June and July. So which do you believe now? My answer, hedged as ever: the forecasters are more reliable than the curve on inflation and the curve more reliable than the forecasters on rates, because the curve is money and the forecasts are opinion, and the MPC reads both. The tiebreaker in 2021 was Haldane, and what made him right was looking at the money supply and the savings pile and saying demand would meet a supply that could not respond. Woodford is looking at the same savings pile five years on and saying demand is meeting a supply bigger than anyone measured. Same method, opposite conclusion, both against the consensus. The reason I forecast FIRST before I look at the consensus is, I’m sure, obvious at this point and whilst it risks me looking stupid, it also stops me making the same mistakes as everyone else (I can then make my own independent mistakes!).

Theme 4: The Synthesis: The Forecasters and Your House

The Summary: The compendium also carries the panel's forecasts for the variables that matter directly here. House price inflation, fourth quarter on fourth quarter, is forecast at 4.1% for 2026 on the independent average, from 2.5% (Pantheon) to 5.3% (Beacon), and 3.9% for 2027. Public sector net borrowing for 2026-27 averages £119.0bn, from £79.8bn (ICAEW) to £159.7bn (CEBR), and £103.1bn for 2027-28. Q4 2026 unemployment averages 5.2%, average earnings 3.6%, and real household disposable income growth 0.2%.

The Propenomix Perspective: Four point one per cent. I read that number three times to make sure I had the right column, and I have. The forecasting profession's central expectation, as of mid-August, is that UK house prices will be 4.1% higher in the fourth quarter of this year than last. Nationwide has them at 1.6% and falling from an April peak; Chris's £-per-square-foot at sale agreed, which leads the Land Registry by five months, has them at 1.2%; Zoopla had 0.9% last week; Rightmove has asking prices down 1% on the year. The only way from here to 4.1% by December is a fourth quarter of about 3% growth in three months, in a market with the lowest approvals since January 2024 and a five-year swap at 4.36%. Galbraith's astrologers, in other words, are alive and well and working on the housing line. The borrowing figure is the one to watch on the 16th: a £40bn range around a £119bn average, with the OBR's forecast build under way and every basis point on the thirty-year adding to the debt interest line. If the September average moves up from £119bn, that is the Budget getting harder in public, three weeks before Healey opens the box.

Source 4: Resolution Foundation, "The Macroeconomic Policy Outlook: Q3 2026" by Stephen Hunsaker and Simon Pittaway (24th August 2026)

The Resolution Foundation is not Neil Woodford's natural ally. Its macro unit has spent a decade arguing that Britain's problem is under-investment and weak dynamism, and it has been broadly supportive of the current government's fiscal stance. Which is why an outlook that sets out to test whether the productivity pickup is real, and finds that it is, deserves to be read alongside him rather than instead of him.

Theme 1: The Pickup Is Real

The Summary: Using its own administrative employment series, built from PAYE Real Time Information for employees and self-assessment returns for the self-employed, the Foundation finds output per hour grew 1.1% a year over the two years to Q2 2026, against an average FALL of 0.7% a year in the two years before and 0.7% a year growth in the late 2010s. On the LFS measure the same two years show a fall of 0.2% a year. The difference is entirely in the labour input: in 2025 the LFS recorded employment growth of 1.7% while the administrative data recorded a fall of 0.1%, "an implausible gap that is more likely to reflect survey improvements rather than genuine employment growth". The ONS Workforce Jobs series gives the same 1.1%. Productivity nevertheless remains around 5% below the level implied by the OBR's last pre-Covid forecast, around £150bn a year of lost output or £4,500 per worker, and the OBR's November downgrade to around 1% has brought it "in line with reality arguably for the first time ever".

The Propenomix Perspective: Same numbers, different house, different politics, same conclusion. That is the definition of robust, or as close as macroeconomics gets, and it is why this paper is here rather than leaving Woodford to make his own case. The Foundation's 1.1% over two years sits below Woodford's LSE-cited 1.6% and Bloomberg's 2%, and above the ONS's 0.7% for the latest year, which is the spread you would expect from four windows on the same data; I would take the middle. What strikes me most is the OBR line. Woodford says the downgrade to 1% was a mistake made on bad data. The Foundation says the downgrade to 1% was the OBR accidentally arriving at the right answer after fifteen years of overshooting. Both can be true: the OBR downgraded for the wrong reason and landed on the right number. If so, the fiscal argument is weaker than Woodford makes it, because a Chancellor rebuilding the forecast on the corrected series would get, on this reading, medium-term growth of about 1%, which is what she was given. The monetary argument survives intact. The hawks were reading a series that said productivity was falling when it was rising at 1.1%; that is a two-point error in the supply side, and it is the MPC's to own, not the Chancellor's - even if the MPC are the ones relying on these numbers, not the ones producing them. 

Theme 2: Flat Hours, Not Faster Output

The Summary: In accounting terms, stronger productivity growth is "down to flatlining hours, not faster output growth". Over the two years to Q2 2026 GDP grew 1.3% a year while hours grew 0.2% a year; in the two years before, GDP grew 0.9% and hours 1.6%. Three-quarters (77%) of the pickup comes from slower hours growth. Against the late 2010s, when GDP grew 2.0% a year and hours 1.3%, the higher recent productivity growth is entirely due to weaker hours. The Foundation asks whether this is simply the cyclical pattern of a cooling labour market and concludes not: in the UK, unlike the US, productivity growth has historically been positively correlated with employment growth, and the 1.4 point slowdown in hours would be expected to add only 0.3 points, "explaining only a sixth of the 1.9 percentage point improvement".

The Propenomix Perspective: This is the paragraph to read before anyone gets excited, and it reconciles the productivity story with the labour market section I write every month. Britain is not producing much more. It is producing slightly more with the same or fewer hours, in an economy where payrolls have fallen by tens of thousands a year, vacancies are at 707k, services headcount has fallen for twenty-three months and manufacturing has only just started hiring. Productivity up because hours are flat is a real improvement in the sense that matters for the Bank: the same output with less labour is less wage pressure per unit of output. It is a much less exciting improvement for anyone hoping it shows up as growth, receipts or rents, because 1.3% GDP growth is 1.3% whichever way you divide it. Where I part company with the sanguine tone is on what flat hours mean for the people not working them. Employment rate 75.1%, inactivity 20.9%, and a productivity miracle built on not hiring is, from where most of my readers sit, a market with fewer tenants in work. Two things true at once, again.

Theme 3: What It Is Not

The Summary: The Foundation tests and rejects three explanations. It is not reallocation from low- to high-productivity sectors: across 19 industries the reallocation term "is a rounding error" outside the pandemic years, and hospitality's share of employee jobs (6.6%) is the same as in 2016-19. It is not a shake-out of low-paid jobs within sectors: across around 400 occupations, jobs in below-median-paid roles grew 3.4% in the year to April 2025 against 1.4% in above-median roles, and essentially all of 2025's 2.3% real pay growth came from higher pay within occupations. And it is not capital deepening: investment has been flat at 19% of GDP since 2023 and public net investment moved only from 2.4% to 2.7% of GDP. The pickup is broad, with 12 of 19 sectors contributing; information and communication is the largest contributor at 0.5 points a year, but it contributed 0.4 points in 2016-19 too, "well before AI existed in any commercial form"; financial services moved from minus 0.7 to zero, wholesale and retail from minus 0.2 to plus 0.2.

The Propenomix Perspective: Here the Foundation and Woodford part company on the cause while agreeing on the fact, and the disagreement is instructive. Woodford's explanation is a labour-market one: the 2025 rise in employer NICs and the minimum wage made people dearer, firms substituted capital for labour, productivity rose. The Foundation looks for exactly that and cannot find it: low-paid jobs grew FASTER than high-paid ones in the year the NICs rise landed, hospitality did not shrink, investment did not rise. The chief issue is that the occupational data stops in April 2025, the month of the change (making their summary around hospitality jobs basically pointless - bearing in mind in November 2025 the ONS measured 170k fewer people on payrolls, with hospitality accounting for over half of that - but also in October 2025 they mentioned it being 4% of the jobs in the sector that had gone) and the services PMI respondents are telling S&P Global every month that they are freezing hiring and automating, which is a substitution story told by the people doing the substituting. So my hedge: Woodford's mechanism was not visible by April 2025 and may be visible by April 2026, and we will not know until next year's ASHE (a counter would be that when the announcement came in what was effectively November 2024, hospitality stopped hiring and started restructuring immediately or after Xmas - an argument which would have some validity). What the report establishes is that the improvement is within sectors, with the same workers, in the same jobs, which is the least sinister version of a productivity pickup and the one most likely to persist. Not too cynical, I hope.

Theme 4: The Synthesis: The Mystery, and What the MPC Does With It

The Summary: With reallocation and capital deepening ruled out, the Foundation concludes the pickup looks like total factor productivity growth, "the secret sauce of economic growth that all but vanished after the financial crisis", measured, as always, as a residual. It calls attributing it to AI premature: 46% of businesses with ten or more employees use AI in some form, up from 23% three years ago, but only around 4.6% of all firms use it extensively, and the Bank's Decision Maker Panel has 62% of firms expecting AI to lift revenue per employee over the next three years against 17% saying it already has. "For now, the revival of productivity looks as mysterious as its fall. But it looks genuine." The report draws no explicit monetary policy conclusion.

The Propenomix Perspective: So where does this leave the argument I set out to test? Woodford said productivity was understated by half; the ONS agrees on the growth rate and is still working on the level. Woodford said it changes the Bank's arithmetic; the ONS's own table says the supply side is two points bigger than the LFS showed, and the Resolution Foundation says the improvement is real, broad and not an artefact. Woodford said inflation lands nearer 3% than 4%; the Treasury's panel has it at 3.3% and the fours are two June forecasts and this week's gas price. Where he overreaches, on my reading, is in the cause (the Foundation cannot find his mechanism), in the fiscal conclusion (the OBR's 1% may be right by accident) and in leaving out the Prime Minister's thumb on the CPI scale. Where he is simply right is the bit that matters for the 17th: a committee with three members voting for a hike last time out on a slack estimate built from a survey the statistics office no longer recommends. If the corrected series had been in front of them in April, I doubt Pill would have voted as he did back then. That is the practical consequence and it is not small (he started the fire here, after all). Will the MPC say so on the 17th? I'm not convinced, with the curve where it is and gas where it is, and a central bank that has just watched its American counterpart rediscover the word "mystery" is unlikely to want it in a minute. But the ground under the hawks moved this month, and it was the ONS that moved it. Onwards, to the diary.

The Week Ahead. A lighter data week than the one just gone, and a heavier speech week, which in a Budget countdown is the more dangerous kind. Monday brings John Healey's first major speech as Chancellor: listen for the words "fiscal rules", count how many times, and note whether "headroom" is a number or an adjective, because the OBR is eight weeks into building the forecast his Budget will sit on and every thirty-year gilt print since PMQs has been adding to the debt interest line. Also, take a shot (possibly of Gaviscon) every time he says “wealth creators”. Monday is also Labor Day in the US, so the Treasury market is shut and the gilt market will be pricing Healey on its own, which is not always a comfortable thing to watch - the opposite of the Bank Holiday just gone. Tuesday: the BRC-NIQ retail sales monitor for August, the till receipts to set against the CBI's shocker of a fortnight ago, and the Lloyds house price index for August, which will either confirm Nationwide's 1.6% or give us a third number for the same month. Thursday: the RICS residential survey for August, where the tenant-demand line went negative in July for the first time since 2020 and I would like to know whether that was heat or trend; and US producer prices, the first half of the print that decides the Fed. Friday is the day: the ONS monthly GDP estimate for July at 7am, the first hard read on the third quarter, with construction output and trade alongside it. If July lands at or above consensus, the Resolution Foundation's line that the UK has led the G7 on growth this year gets another data point, and, more usefully for us, so does the productivity story, because a flat-hours economy needs a positive output number to keep the numerator moving. Friday afternoon brings the US CPI for August, which is the print the 15th-16th September Fed meeting will be decided on, and therefore the print that your five-year swap (and derived mortgages) will be decided on. Set your alerts.

Non-calendar items. The ONS methods article on the "component approach" to productivity is due this month, date unannounced; if it lands before the 17th it will be in the MPC's papers, and it is the single most interesting statistical release of the autumn for anyone who has read this Deep Dive. Oil and gas, obviously, with the January Ofgem window filling at $96 and €72 and the President "prepared to do another one any time we want". The Sunday papers are Budget-briefing documents again from here to 28th October, and the LHA rates, the property income surcharge and the pension IHT change are all live watch items on that front. And a fortnight out: the labour market report on Tuesday 15th, the August CPI on Wednesday 16th, the next HMT forecast comparison the same day, and the MPC and the QT envelope on Thursday 17th. Mark the 17th. Clear the diary for the afternoon, if you have a completion that week.

As we get towards the end for this week - I can't wait for our next workshop in Manchester on 1st October. JVs and M&A is the topic, and this week has made the argument for it better than I could: a five-year swap that moved fourteen basis points in a month (but more than that in a week), a mortgage market at its lowest approvals since January 2024, and a forecasting community that cannot agree within two points on where inflation will be at Christmas. In that world the deals that complete are the ones with more than one source of capital behind them and a structure that survives a rate that moves between offer and exchange, which is precisely what Rod and I will be working through, with the real-life case studies and the no-holds-barred Q+A as always. The VIP dinner returns too, in the smaller setting where the properly specific questions get answered, and there is very little of it left. Early Bird pricing, at more than 10% off, is in its final days. Book your tickets for Thursday 1st October, Manchester at: https://tinyurl.com/pbwoct26 - powered by Roma Finance, now covering all of your needs from bridging and development to term finance.

Above all - please remember to Keep Calm, ALWAYS listen to or read the Supplement, and Carry On. It has been a week about the difference between what is measured and what is true: a productivity series that was there all along, a house price index that rewrote its own July, a gilt yield that went to 2007 and came back in three days, and a forecasting profession whose consensus is nearer the truth than its critics say and further from it than its house price line admits. Galbraith's line was a joke, and like most good jokes it was only half one. In weeks like this the temptation is to trust the loudest number, which this week was 5.29%, and the loudest number was gone by Friday. The fundamentals have not changed: there remains a shortage of 2-3 bed terraces and semis, there was one fifteen years ago, and nothing in this week's data, not a housebuilding index at 37.6, not 56,000 approvals, not a productivity revival that turns out to be flat hours rather than more houses, does anything but deepen it. "Investing in property" is not a guaranteed win, and with the next rate move pointing up and the market now saying sooner rather than later, the discipline around leverage matters more this month than last. But buying below replacement cost, strong on the spreadsheet, strong on yield, in a market repricing slowly in real terms from sellers who are finally reading the reductions column, from a lender who has just sent you a repricing letter you should have anticipated, that remains as close to a guarantee as this game offers. The houses still aren't being built. The people still need somewhere to live. The forecasters, for once, may be less wrong than the market. KCCO!