And, as, I guess, predicted by Larry Summers in 2023, inflation is on its way back. The US has started a war with Iran, seemingly with the strategic goal of increasing inflation by creating a second 1970s-style oil shock. As a longstanding advocate for “Team Transitory” during the pandemic inflation, I have to admit, I did not see that one coming. Although to be fair, Paul Krugman didn’t either.

But an oil shock is also a great opportunity to explain the idea of “systemically significant prices”, and show how the mathematical machinery used to identify those prices can produce benchmark forecasts for inflation when supply shocks happen. The Consumer Price Index (CPI) data released last Friday show headline inflation of 0.87 percent month-over-month (11 percent annualized!), of which 0.67 percentage points — more than three quarters — came from direct energy price increases alone. The method we work through here would lead us to expect a further price increase on the order of 0.85 percentage points over the next few months as indirect effects push price increases through the supply side.

By now, we all know the basics. On February 28, the US and Israel attacked Iran, and Iran retaliated by closing the Strait of Hormuz to oil tankers. As we are drafting this newsletter, it is not clear how many directions this conflict will take, how and when it will be resolved, or how stable any resolution will prove. Even stockpickers are putting boots on the ground to monitor the situation at the Strait. Last week, the plan seemed to be for Iran to start charging tankers a buck a barrel to cross, but by the time you read this, it will likely be something else entirely.

The qualitative shape of an oil shock is familiar, but a shock of this magnitude is not.

The oil shocks that defined the 1970s and produced widespread shortages removed, at their worst, about five million barrels per day from global supply. The Ukraine War shock that pulled inflation up through 2022 and 2023 removed about three million barrels per day. The Hormuz closure, if the strait remains closed, takes out roughly 20 million barrels per day — about a fifth of total global petroleum consumption, something like seven times the Ukraine disruption.

A Tool for Thinking about Inflation

In response to the 2022 inflation episode, economists and policymakers went looking for tools to better understand how specific narrow price shocks can produce broad-based inflation.

In particular, Isabella Weber, Jesús Lara Jauregui, Lucas Teixeira, and Luiza Nassif Pires published a paper using the same input-output tables we used to look at tariffs last month to show how these price shocks evolve. For a given industry, they simulate a price shock to see how it propagates through the network of who buys what from whom and show how much the resulting cost cascade would increase the price index if the shock were passed along completely at every step.

The method creates a distinction between two kinds of inflation. The first is direct inflation: the price of a thing consumers buy goes up — housing costs rise, and the housing component of the CPI rises with them. The second is indirect inflation: the price of an input rises, and the price of something consumers actually buy rises with it — aluminum gets more expensive, and the cost of a can of soda goes up a few cents.

“Systemically significant prices” are the set of prices which, when shocked, produce the largest combined increase in direct and indirect inflation.

But, if you think about it, indirect effects for the consumer are just a sequence of direct effects between different industries which all end as direct effects on the consumer. A shock to crude oil raises the price of refined petroleum products, which raises the cost of running a trucking fleet, which raises the cost of delivering groceries to the store, which raises the price you pay at the register.

We can trace the whole sequence through the input-output tables and track a single upstream shock all the way out to the consumer economy as a whole.

From Crude Oil to Everything Else

Since we are facing another oil shock, let’s see how its direct effects turn into its indirect effects.

The first round of an oil shock is simple: the price shock is mostly paid by oil refineries. In the second round, that price shock fans out through refined products to different sectors in different roles. The final products of some sectors are basically made of the stuff — asphalt is not far from solidified oil — while other sectors rely on other refined products like fuel in transportation, or as feedstock for plastics, pharmaceuticals and fertilizers.

The third and fourth rounds move from those sectors to the rest of the economy. By the fourth round of tracing impacts, around 85 percent of the total indirect impact from the price shock has been passed on to consumer prices, through over hundreds of different industries. The remaining 15% of the indirect impact is spread across an indefinite echoing tail of longer paths.

However, it is important to remember that the price model used in this model computes these price shocks as if they propagate through all rounds instantaneously. In reality, firms across the affected industries have to decide to raise their prices sequentially, over time, to actually pass these price increases through to consumers. In the case of “Seller’s Inflation”, firms might even use the upstream price shock to raise their prices by more than their input costs increased.

So now that we understand how both direct and indirect inflation work, which prices are actually systemically significant according to this model?

Back-of-the-Envelope

So, what does all of this mean for inflation? The table above gives us some tools to investigate.

Let’s start with oil extraction by itself. In Weber’s model, oil and gas extraction has almost zero direct CPI weight — consumers do not buy crude oil — so its entire impact on inflation is indirect. Every cent of its CPI effect has to propagate through the production network before it reaches consumer prices. In the table below, we have extended Weber’s model to speak in terms of CPI rather than PCE inflation, and to break out the direct and indirect contributions separately.

Brent crude is currently running at about 1.7 standard deviations above its pre-war baseline, which implies a minimum of 0.62 percentage points of indirect CPI inflation working its way through supply-side channels. Because these are all indirect effects — cost increases that have to pass through refineries, then through chemicals and transport, then through the industries that buy from them — they will not arrive in the measurement system all at once. This tells us the oil shock will be with us for a while, accumulating as it propagates.

But crude prices are only half the picture. We are also seeing strikes on refinery infrastructure in Iran and across the Gulf, and the resulting supply crunch in refined products has pushed downstream prices harder than upstream ones. Gasoline, diesel, and jet fuel have all moved more aggressively than the crude oil they are made from, as rationing dynamics take hold in the refining sector. To understand the full inflation picture, we need to look at the entire oil and refining complex.

The shock to refined products has so far been larger than the shock to crude — petroleum and coal products PPI is currently running at about 1.9 standard deviations, compared to 1.7 for extraction. And unlike extraction, refining has a substantial direct channel to CPI: consumers buy gasoline, heating oil, and propane directly. At 1.9 sigmas, the model shows 0.67 percentage points of direct CPI impact plus another 0.23 percentage points of indirect impact, for a total of 0.90 percentage points from the refining shock alone.

We can already see this playing out in the March CPI data released last Friday. Gasoline prices rose 21.2 percent in a single month — the largest monthly increase since the series began in 1967. Fuel oil was up 30.7 percent. The energy index as a whole jumped 10.9 percent, its biggest move since September 2005. That is the direct channel from the refinery shocks arriving. But core CPI — everything except food and energy — came in at just 0.2 percent, right on its recent trend. The indirect effects, the ones that have to work through petrochemicals and transport and packaging before they reach the register, have not shown up yet. They will.

Now, we would expect the direct component to show up relatively quickly as companies pass through higher input costs. But the CPI and PPI measurement windows do not perfectly align, which means the unmeasured present is always running a step ahead of the measured past. Even after that record gasoline print, there is still significant direct and indirect inflation in the pipeline from closure that has already happened.

The longer the strait stays closed, the worse these numbers will get.

Conclusion

The price of oil is systemically significant because it propagates.

The tables above quantify the direct hit, but the cascade we traced from crude to everything takes time to work through contracts, inventories, and pricing cycles. March CPI showed the fastest month-over-month increase in gasoline prices in the index’s history, and that is only the first round — the direct effect. The indirect effects, the ones that move through petrochemicals into plastics, through diesel into trucking, through trucking into groceries, will take months to fully pass through.

We should already be preparing for the inflationary consequences of the strait closure to last through the midterm elections at minimum. At the same time, negotiations seem unlikely to resolve the underlying problem given the counterparty reliability issues involved.

Economic policy has become increasingly entangled with foreign policy, and the Trump administration had been benefiting from a quiet tailwind: falling energy prices were canceling out some of the aggregate price impact of its tariff regime. That tailwind will now become a headwind.

It is also worth noting that there is an obvious temptation to use the Strategic Petroleum Reserve to paper over the economic consequences of a foreign policy disaster here. The problem is, unlike the disruption during the Ukraine war, the SPR is not up to this task. It can release 4.4 million barrels per day running flat out — roughly a quarter of the twenty million lost through closure of the strait — but even at Hormuz flow rates, the entire reserve holds less than three weeks of replacement supply. The SPR is a price buffer, not a substitute for global markets. Fixing inflation requires a solution to the geopolitical problem; the SPR cannot do it alone.

This Month’s Data

Employment

Total employment growth continues to decelerate towards zero in the March employment data.

Usually employment growth sliding towards zero would be a warning sign for the labor market, but according to a FEDS Note published last week, there are deeper problems. The Trump administration’s approach to immigration policy has slowed population growth so sharply that 2026 is on track for the weakest year since 1951. This is, fundamentally, a negative labor supply shock.

What this means, somewhat grimly, is that the breakeven pace of monthly job creation — the number needed to keep the unemployment rate from rising as the labor force grows — could fall to “nearly zero, requiring less than 10,000 new jobs per month in 2026”, down from about 85,000 in 2025. If breakeven is essentially zero, we could see the unemployment rate hold steady even with the aggregate jobs number flat or modestly negative. Another “year with no jobs, but no recession”, as Claudia Sahm has termed it.

We can see the same picture in the sectoral breakout as in the headline: goods-producing industries are shedding workers, tech remains in correction as pandemic-era surge hiring continues to unwind, and health care sector employment is growing four times as fast as total employment.

Industrial Production and Capacity

Industrial production is up 1.6 percent over 2025 levels and 2.6 percent over 2017 levels. Manufacturing has fallen below its 2017 level. Capacity utilization is 76.3 percent, within a tenth of a percentage point of its 2017 average.

I keep harping on these 2017 numbers because Oren Cass has a new piece in the FT citing rising industrial production as evidence that the Trump tariff regime is working.

But if we look at which sectors are actually driving the acceleration, the data doesn’t quite match the story.

The only sector in the top five of this table with any connection to tariffs is primary metals. Oil and gas extraction has no tariff story, and rising computer and electronic products production is a CHIPS Act and onshoring story whose trajectory was set before Trump returned to office.

The semiconductor timeline makes this disconnect concrete. The CHIPS Act was signed at a local peak in capacity utilization, and measured capacity begins inflecting upwards as federal dollars translated into investment in the sector. Two years later, inflation-adjusted investment in manufacturing-related structures hit its own peak — right when output began inflecting upwards while utilization rates held steady. That is the shape of the CHIPS Act working its way through the data. The tariffs had nothing to do with it.

Prices

Producer prices on final demand have been running steady around three percent year-over-year for several months.

Breaking out producer prices for goods-producing sectors, we see prices rising over ten percent annually for primary metals, mining ex oil and gas, utilities, and computer and electronic products. We also see acceleration over the last six months of readings, rather than deceleration, suggesting these price rises will continue.

As it stands, the only goods-producing sectors holding the headline down — petroleum and coal products and oil and gas extraction — will soon be some of the biggest sources of inflationary pressure. When that happens, the offset that has been keeping headline goods PPI near three percent disappears, and the rest of the matrix is already running well above that pace. The headline will have to move with them.

The services heatmap shows rising prices largely concentrated in wholesale trade and the retail margin sectors that sit immediately downstream of it. The only significant price decreases come from broadcasting, continuing a long deflationary streak.

Trade

In the January data released in March, we see imports continuing to shrink, but at a decelerating rate. Export growth is rising at the same time, which cashes out to a modest improvement in the trade balance.

The same sectors whose output we see lagging in the industrial production data continue to see declining exports. Interestingly, many of them are also seeing declining imports, suggesting a demand problem as much as an issue of competitiveness. In contrast, computer and electronic products are moving in both directions at speed, with exports up 35.5 percent year-over-year and imports up 53.2 percent. This is what a globally-integrated semiconductor supply chain looks like in motion.

Yet, outside of extraction and tech, we see imports and domestic production are both down. Whatever is replacing the imports is not US manufacturing.

What It All Means

What this month’s data captures is a system that has been holding remarkably still in the moment before a major oil shock. Industrial production hovers around 2017 levels, capacity utilization has been roughly unchanged for a decade while employment is decelerating towards a structurally lower breakeven growth rate. Producer prices are running near three percent — things have calmed down on the supply side despite the ongoing uncertainty.

Now, inflation is coming.

But tariffs as a policy are more important to the president’s agenda than any specific legal authorization is to the tariffs, so they will likely be back soon, under different and likely just-as-challengeable authorities. There is not much sense in trying to predict which authorities, or which tariffs exactly, since the point of the tariffs is to produce a reaction, not so much to be understood.

Over the seven months the “Liberation Day” tariffs were in force, reactions were more or less constant, but only rarely impacted long-term business decisions. The chaos of implementation and the randomness of incidence meant that the economic impact was closer to the supply-side price rationing of the pandemic era, where specific inputs suddenly and significantly changed prices.

We can seesome evidence of this in the responses to Fed manufacturing surveys while the tariffs were in force: higher uncertainty about the policy environment led firms to delay or cancel investments in new capacity while waiting for the demand-side impacts to become clear. In effect, the scattershot Trump II tariffs produced a kind of mini-pandemic-supply-chain-crisis — one that is now over, and which may even be rebated if the administration can figure out the required administrative infrastructure.

Yet there is one group of tariffs that survived the Supreme Court’s ruling, which have in fact been in force since the first Trump administration: the Section 232 tariffs on steel and aluminum. These tariffs were put in place nearly eight years ago, and the rates have only gone up since then — from 25 percent on steel and 10 percent on aluminum in March 2018, to 25 percent on both when country exceptions were eliminated last March, to 50 percent on both when the rate doubled in June.

Tracking the Wily Tariff

So, instead of memorializing tariffs lost, let’s take a closer look at the tariffs that have stuck around and think through where to look for their impact. Tariffs work by changing the cost of imported goods through an extra tax, for households and businesses alike. In the consumer case — say, ordering LPs online from the UK to the US — this is relatively easy to understand: the consumer pays an extra tax on the goods they ordered.

But the Section 232 tariffs are taxes on intermediate goods used by industry. I don’t know about you, but I do not order a lot of raw steel from China to the house these days. This means that the tariffs will have some kind of impact on the relative cost structure across different industries: different sectors buy steel and aluminum, to use for different things, and that will affect profitability, competitiveness, and on a long enough timeline maybe even capital structure for the downstream industries most exposed. But which industries? And how exposed are they? To answer these questions, we have to turn to the input-output tables.

The IO tables are one of the less well-known products of the national accounts. The Bureau of Economic Analysis (BEA) produces detailed benchmarks every five years — most recently for 2017 — whichcover the linkages between 402 individual industries, mapping how many dollars flow from one industry (the seller) to another (the buyer). The diagram below shows this whole system: every arc represents a transaction between two sectors.

This is not the clearest chart in the world, but it does give a real sense of the complexity involved. The IO tables are a bit like a map of the supply side. So for our purposes, we can use the fact that the economy is coordinated through a network of purchase relationships to track how a tariff on one product propagates through the whole network to every industry that buys it, and then onwards to every industry that buys from them, onwards.

Upstream and Downstream

The Section 232 tariffs cover commodities primarily sold by the Primary Metals Production industry — NAICS 331, which includes iron and steel mills, aluminum smelting, and related subsectors. Who buys from 331? The single largest downstream customer is fabricated metals, NAICS 332: the 49,000 firms that turn steel plate, aluminum sheet, and metal bar stock into structural beams, pipe fittings, springs, machine parts, and hardware. Here is that flow as a Sankey diagram, roughly sized by dollar volume.

The IO tables give us a “direct requirements coefficient” for each buyer-seller pair — the fraction of each dollar of output that an industry spends on a particular input. For fabricated pipe and fittings (NAICS 332996), the coefficient for steel inputs is 0.197: about twenty cents of every dollar of output goes to buying steel. For plumbing trim (332913), it is half a cent, as trim is more often made from copper than steel. Steel intensity varies enormously across the 20 subsectors of fabricated metals.

To take a naive estimate of the tariff burden on a given sector, we can multiply three numbers: the IO coefficient for steel and aluminum inputs, the share of those inputs that are imported (and therefore subject to the tariff), and the tariff rate itself. This gives us an effective tariff cost per dollar of output — what we can call the tariff burden, measured in basis points.

These estimates are admittedly rough. The IO coefficients are from the 2017 benchmark, so nearly a decade old and measured before the tariffs were implemented. Import shares do shift and firms do substitute – especially as the Inflation Reduction Act drove investments into clean steel production lines – so, take these calculations with a grain of salt. But the order of magnitude is still an informativeway to connect macro and micro to see behind the headline numbers.

On the upstream side: ten primary metals subsectors, their Section 232 coverage, and how much of their own output they source from within the 331 Primary Metals sector.

Five of these ten subsectors are directly covered by Section 232: iron and steel mills, iron and steel forging, steel wire drawing, aluminum production, and aluminum foundries are all covered at the full 50 percent rate. The remaining five — copper, nonferrous foundries, other rolling and drawing — are not covered, but do purchase from the covered subsectors, so they still likely face some amount of tariff indirectly through higher input costs. The “own-sector coefficient” column shows how tightly linked these industries are to each other: copper rolling and drawing sources 67 percent of its output value from within primary metals, though nearly all of that is untariffed copper.

And on the downstream side: the twenty fabricated metals subsectors, ranked by their estimated effective tariff burden.

The range is substantial, from 4 cents on the dollar, to four-hundredths of a cent on the dollar. Metal containers are the most exposed, driven almost entirely by aluminum content. Fabricated pipe fittings bear 356 basis points, mostly steel. At the bottom, plumbing trim bears only 4 basis points — the fact plumbing trim is largely copper means the tariff is essentially invisible in its cost structure. Between these extremes, every subsector has a specific, calculable exposure that depends on three things: how much steel and aluminum it uses per dollar of output, how much of that steel and aluminum is imported, and which of its primary metals purchases are actually covered by Section 232.

Did It Work?

None of this information can tell you whether these tariffs were a good idea. That depends on what you think they were supposed to accomplish, which is a longer conversation. What it does tell you is that “the steel tariff” is not a simple cost like a consumer tariff, nor one cost borne equally across sectors. It is a whole manifold of different costs, ranging from negligible to substantial, distributed across the economy’s production network in a pattern that the IO tables make visible and that would otherwise be invisible.

These are the kinds of cross-industry impacts that need to inform policy, especially as we confront the challenge of decarbonization. In essentially every case, the real work of decarbonization is to bring down the carbon output of a sector by swapping in different equipment. How this work is sequenced and flows through the existing productive structure is critical to its short and long term success.

This Month’s Data

This last month saw new GDP data, in addition to the usual data for employment, prices, production, and trade. But first, before the charts, the key takeaways.

Key Takeaways

• 2025Q4 GDP has come in soft due to the government shutdown, but the continued growth in consumer spending, especially on services, means the macroeconomy in 2025 did not run into any of the “sudden stops” many commentators had expected.

• Instead, there was a simple and steady degradation of the growth outlook, as investment narrowed to a small number of “winner” sectors while the labor market continued to downshift on hiring, with hires rates approaching early-2010s Great Recession lows.

• Services continue to drive economic growth, in both low-margin sectors like Health Care, and high-margin sectors like Software. It would be interesting to dig in on recent rule changes that have increased the share of software work that can be counted as capital expenditures, rather than operational expenditures, to see how much that is changing the measured composition of the boom in spending on large language models and other AI-related software.

• The Iran war is already showing up as an inflationary energy shock, as oil, gasoline, and especially diesel prices have ramped up in recent days. The question now is how long the disruption will last for, and what the knock on effects will be.

• For context, a geostrategically significant amount of global oil and gas flows through the strait — roughly 20 percent of daily global supply — and in order for markets to balance if the strait remains closed, we will probably have to see prices go above $200 a barrel to see durable demand destruction at the scale of 20 million barrels per day. Oil hovered around $120/barrel in the late 2000s on strong global commodity price growth. It briefly touched the $130s during the 2022 shock, which only represented a disruption of 1-2 million barrels per day. Already, since I have started drafting this newsletter edition, oil has moved from around $60/barrel to almost ping $120/barrel spot in overnight trading. This has started pulling up the oil futures market as well as the market begins reckoning with how long the war will last.

• If they don’t sort this one out, there may not be much else to talk about besides oil in next month’s issue.

GDP

Headline GDP growth fell to 1.4 percent annualized in Q4, down sharply from 4.4 percent in Q3. The single largest drag was the federal government, which subtracted 1.15 percentage points from growth as the shutdown drove a 16.6 percent annualized plunge in federal government consumption and investment for the quarter. Absent the shutdown, growth would land around 2.5 percent — not spectacular, not weak, but a step down from stronger mid-year growth.

Consumer spending growth held steady at 2.4 percent, contributing 1.58 percentage points to headline growth. Almost all of it came from growth in services (+1.59pp); goods spending contributed essentially nothing (-0.01pp). This is in line with what we have been seeing for most of 2025: the consumer economy runs on services, not stuff. Net exports were roughly flat (+0.08pp). Inventories added 0.2 percentage points. Full year 2025 growth came in at 2.2 percent, a step down from 2.8 percent the year before. Investment in intellectual property products grew 7.4 percent, the strongest component by far, driven by AI-related software and R&D, while equipment investment grew a more modest 3.2 percent. Structures contracted for the eighth straight quarter at -2.4 percent, and residential investment fell 1.5 percent, declining in all four quarters of 2025. The investment mix continues to narrow in recent quarters: intellectual property products and equipment carry the total, while construction — factories and housing alike — continue to contract. This narrowing is broadly in line with the slow slide in investment we flagged as likely last month.

Industrial Production

Industrial production continues to grow in the most recent data, though capacity utilization still lags, and has only just begun growing in the last six months, after years of contraction.

Capacity utilization shifts were mostly positive: mining ex oil and gas accelerated by 8.3 percentage points, beverages and tobacco by 7.0. Apparel was the notable decline, losing 2.8 percentage points of output year-over-year.

The charts below rank every 3-digit manufacturing sector by year-over-year output change and pull out the top and bottom two. Among durables, the winners are the high-tech, heavy-duty end: computers and electronics up 7.7 percent transportation equipment up 5.9 percent. The losers are the lower-tech end: wood products down 4.6 percent, furniture down 2.8 percent.

Nondurables tell a bleaker version of the same story. The aggregate is barely positive at +0.6 percent. Paper is down 4.9 percent, printing down 2.7 percent. Only beverage and tobacco (+2.4%) and food (+1.7%) managed to stay above the zero line. The production recovery, such as it is, belongs entirely to durables.

Employment

The employment picture is largely the same story as last month. Ranking all 2-digit services sectors by year-over-year employment change, health care continues to lead at +3.2%, with other services growing at a more subdued +0.8 percent. The services aggregate grew 0.5 percent. At the bottom, information shed 1.5 percent and administrative and waste services lost 1.1 percent of its employees — both sectors that have been contracting for months.

On the goods side, manufacturing employment is essentially flat year-over-year, with construction leading at +1.2 percent. Mining and logging continue to contract.

Overall, no major surprises in the January data.

Producer Prices

Producer prices show two things worth watching. Ranking 3-digit sectors by year-over-year PPI change, the services side is led by securities at +9.2 percent and merchandise wholesale at +8.1 percent. The aggregate for all services sits at +2.9%. At the bottom, broadcasting and internet continue its deflationary streak at -7.9 percent.

The same ranking on the goods side puts primary metals at the top: PPI up 16 percent year-over-year — directly relevant to the tariff analysis above, and consistent with what we would expect from anow markedly higher 50 percent Section 232 rate. Mining excluding oil and gas is close behind at +14.9 percent. At the bottom, oil and gas extraction prices are down 10.7 percent, with petroleum and coal products down 5.3 percent. Total manufacturing PPI sits at +2.4 percent.

Energy prices have been softening for months but no matter how this Iran conflict goes, that trend is now unreliable as a forecast.

Trade

In trade, we see some large month-to-month swings concentrated in a few sectors. Mining imports surged 25 percentage points in their year-over-year growth rate between November and December 2025 — a sharp acceleration. Oil and gas imports also picked up (+13.3pp), as did computer and electronics imports (+11.7pp). Forestry imports fell 33 percentage points.

Exports show a mirror pattern for some of these: forestry exports surged (+29.6pp) against falling imports while computer and electronics exports accelerated (+20.1pp) on rising imports.

Our work at Forces of Production is to provide monthly tracking on the supply side data of the US economy.

Our goal is to contribute a public image of the supply side to support and sharpen ongoing conversations about industrial policy, public ownership, and new approaches to holding down the cost of living.

For more, check out our launch essay.

Subscribe now

In this, our first issue, we will start by introducing some key tools necessary to understanding the supply side and then, use them to tell a familiar story — the shift from goods to services on the supply side. After that, we will look at new data published last month to see if the business cycle is pointing to a crisis on the supply side.

Before setting off to see what last month’s data shows, it’s worth demonstrating how the tools we will use can explain changes on the supply side over the short and long term.

Our Methods

When we talk about the supply side, we are talking about production. Everything that is supplied to the market has to be produced by someone or something, somewhere.

On one level, this is obvious, but on another level, the supply side remains one of the great unsolved problems in economics: how to accurately model the intricate web of production capacity.

Let’s start with some definitions.

Production takes place in firms, which employ labor to use capital equipment to produce output, output which sometimes requires extra materials, services, and energy.

In the Bureau of Labor Statistics documentation, this breakdown of the production process into its component parts — capital, labor, energy, materials, and services — is reported in the KLEMS accounts. The “output” of each firm’s production process is the “supply” that we are interested in.

The biggest challenge in studying the supply side is the sheer variety. Different production processes use different inputs to producedifferent outputs. To handle this, we at Forces of Production use the North American Industrial Classification System, or NAICS Codes.In it, firms with similar production processes are grouped together by sector and industry and organized by hierarchical code: firms in the same group can more easily sub in for one another with less need to adapt their production processes than firms in different groups.

In fact, every transaction between these different sectors — each given by a different NAICS code — falls into a KLEMS category, whether capital, energy, materials or services. The rest goes to regular consumers like you and me.

Let’s see how we can use these tools to tell a familiar story with added detail — to show how these tools work before diving in to analyze the monthly supply side data.

The shift towards services, and the changing composition of those services, is the biggest story of the 21^st century for the US economy. But one of the major limitations in analyzing the supply side is the lack of output data for the service sector. Goods are more closely tracked than services in the economic data, with multiple measures of price and output. For services, data on the price of services sold to other businesses is published, but no measures of output. There is one indicator that provides data on both goods and services though: employment. Given that labor is the original force of production, employment is a great example to work from.

Employment is tricky, though, because it is so tied to the business cycle. In times of crisis, the supply side loses jobs across goods and services alike, and those lost jobs have to be rebuilt from scratch. This is what makes recessions so damaging for the supply side: existing capacity disappears, and new capacity has to be constructed from the ground up.

This dynamic has been given many names over the years, from the crisis of reproduction, to Milton Friedman’s “plucking model”, named for the fact that charts of unemployment across recession and recovery look a bit like a plucked string.

To get a sense of where we are now, let’s compare how different sectors across goods and services bounced back following the last two major crises: the pandemic and the Great Financial Crisis.

Over both swings through the macro business cycle, we saw the total employment level grow, but that growth favored different sectors.Some parts of the supply side have returned stronger than before, while others have steadily weakened over the last twenty years.

The chart above shows employment growth over both recession recoveries, across both goods and services, broken down by 3-digit NAICS code. While a few sectors saw growth across only one recovery, the majority either grew through both recoveries or lost jobs across both recoveries.

Looking at the sectors that have gained and lost employees over the past two decades will help us understand how what people do for work has changed.

Most of these industries in which employment levels grew back stronger over both crises are in the service sector, with a split between high-wage/high-margin work like Professional Services, Insurance, and Data Hosting and low-wage/low-margin work like Social Assistance, Warehousing, and Food Manufacturing.

Much of the strongest employment growth has been in care sectors, as the supply side adapts to an aging population in need of care. Warehousing and Courier employment has also grown as shopping has been increasingly rerouted from storefronts to online.

In some of these sectors — as in fabricated metal product manufacturing, or oil and gas extraction — the Industrial Production data shows output has largely been maintained despite falling employment. Yet many of the sectors showing outright declines over both recoveries have been stories of structural decline rather than productivity gain. Clothing retail employment has declined by double digits across both periods while trade-exposed industries like furniture, printing, and paper manufacturing have all seen falling employment.

Last Month’s Data

Many have been worried about the supply side, with rising geopolitical tensions and daily-changing tariffs creating an environment of incredible uncertainty.

It is easy to see this uncertainty as the source of any crisis if one arises, but the supply side of the US economy is both very large and very inertial.

Looking across last month’s data, we see a broad continuation of previous trends. Employment growth continues to slow whileIndustrial Production continues to gradually rise, led now by Durable Goods manufacturing.

Durable Goods manufacturing is also driving recent increases in Capacity Utilization as well. Producer Prices continue their gradual re-acceleration, with the greatest seller’s inflation showing up earlier in the supply chain, from Mining and Primary Metals production. Price increases for upstream materials like these can have an outsized influence on the cost of production across a wide range of industriesthat use them.

Employment

The service sector continues to drive employment gains, with Goods-Producing employment levels down slightly over the past year.

The biggest gains have all been in the service sector, with Health Care the clear standout at 3.1% employment growth since December 2024.

Interestingly given the rising producer prices in the Mining sector, we see the largest employment losses in this sector over the past year, at a 2.2% decline.

Production and Capacity

Industrial Production has begun to grow again in recent months. However, the total level of production has increased less than 2% from the index year in 2017.

At the aggregate level, manufacturing remains a story of industries shedding workers while adapting production processes to keep output roughly constant.

Interestingly, Durable Manufacturing has taken over as the source of output growth, following declines through much of 2023 and 2024.

The biggest output gains over the past year have come from industries related to Biden-era industrial policy as well as Oil & Gas Extraction, suggesting that although many industrial policy programs may have been cancelled, the supply side still benefits from the momentum they created.

The Wood Products, Paper, and Printing complex continues to show notable production declines.

Overall, Capacity Utilization continues to track the same dynamics as production, with Durable Goods manufacturing leading.

Producer Prices

The Producer Price Index — inflation statistics for goods and services sold to firms rather than consumers — has begun to drift upward again in recent months.

We see the biggest gains in the Producer Price Index at upstream points in the supply chain.

Mining and Primary Metals are significant inputs to a wide range of production processes, and rising costs for Building Materials will also hurt affordability for housing construction. Rising Utilities costs show that increased electricity demand is affecting businesses alongside households.

Analysis

Overall, we have seen surprising inertia across the supply side, despite the extreme shifts in trade policy and consumer confidence. Tariffs are up, and consumer confidence is down, but the supply side of the US economy is like a massive ship. Steering it in a different direction takes a long time, and blowing holes in the side of the hull won’t make it turn much faster.

If anything, what’s happening seems like a very old business cycle dynamic. Back in the days of Keynes and Jevons, the market economy was much smaller and cycles were driven more by investment than consumption. Early macroeconomists noticed this and identified a pattern where a slowdown in investment spending would lead to a slowdown in hiring, and in output growth. This could become self-reinforcing, because the decision to invest less was validated by the lower consumer demand from lower hiring. Since today’s economy is much more insulated by consumption spending, which remains high, this dynamic can still produce a slowdown, but it will likely be muted, protracted, and mild on its own.

Despite this, the uncertainty created by a more volatile global geopolitical situation and nonsensical domestic economic policy does seem to be having an effect on investment, at least outside of AI. Manufacturers have been saying as much in Federal Reserve surveys all year.

There is also evidence that this slowdown is feeding through to employment, with a lower hiring rate than any point in over a decade in 2025.

Should you build that new factory domestically or abroad, given the new tariffs? Should industrial policy target innovation or stability? Should you really pay that much for a basket of groceries? Answering feels like steering, but these decisions are made atop currents no one fully controls — transactional oceans that move you even as you try to move yourself.

Yet your navigational decisions — which you must make constantly — are responsible for keeping your ship intact and on course into the uncharted waters of an uncertain future. The economy can be more treacherous than thesea: new stars that don’t match old charts seem to appear and disappear; the productive forces shape the waves and tilt the skies.

Old maps of development showed a one-way path to manufacturing and modernization, yet the service sector increasingly drives developed economies. Boosters promise unlimited productivity gains from AI while the electrical system promises ever-growing lead times for the new transformers and turbines which will power the data centers behind those “unlimited gains”. Consumer spending hits new highs while consumer sentiment shows epochal pessimism, and the world adapts to Trump’s tariffs, whose effects remain far from clear.

All of us now are traveling into the open ocean of an economy whose activities are actively changing the climate that has so far made them possible.

In conditions like these, sailors have long distinguished two ways to navigate. Any sailor who can, navigates by Position Fixing. When you can see the stars, pick up the beacon of a lighthouse or satellite, or find a shoreline landmark, you can triangulate your precise position with little more than simple geometry. If the angles of the landmarks can be found, then you can fix your position as the third point of a triangle. By keeping track of these position fixes, it becomes possible to plot a trajectory, to map forces in the winds and in the seas, and to begin to find your way.

Most importantly, you know where your ship really is, in a way you can tell someone else about — maybe someone who can alert you to an oncoming squall or an invisible reef. If, say, we can see employment slowdowns and falling capacity utilization in machinery production, but demand is rising for machinery while employment in metals mining and refinement grows, we might be able to pinpoint an economic bottleneck. If we see rising importsand rising import costs, for both machinery and the metals complex, we might be looking at a global pinch-point for capital goods used to produce machinery.

But if position fixing is impossible, you fall into a position that sailors have long dreaded: Dead Reckoning. Without stars or landmarks, the best you can do is feed old position fixes into the system. Dead Reckoning is nothing more than a guess at where you are, based on where you thought you were, and where it seemed you were heading then. Errors accumulate, and ships drift off to wreck without ever seeing where they really were. Auto manufacturers learned this over the pandemic — expecting a cataclysmic fall in demand, they cut orders for the parts they use in production, especially semiconductors. By the time they had updated to the government’s fiscal strategy, they were at the back of the line for those semiconductors. The result for the broader economy? An inflationary shortage from millions of missing vehicles.

Dead Reckoning can help, but if Dead Reckoning is your only strategy, your ship will need more help than your strategy can provide.

Worse still, the navigational problem we collectively face is bigger than any one ship. Your ship can only be one of hundreds of millions in the convoy of the US economy. However well-navigated, its trajectory will always be contained and shaped by the direction and drift of the convoy as a whole: everyone is heading somewhere. Wassily Leontief anticipated the automakers’ pandemic-era problems decades in advance:

The American economy is a gigantic, intricate machine. A serious malfunction of any one of its component parts affects, sometimes with a long delay, the workings of all the other parts. Decisions that are made and actions taken today will affect the life of everyone. Some of these changes might be felt only five, perhaps ten years, and even twenty years from now.

The real problem is that, despite all its gigantic, intricate complexity, the economic convoy has no leader. The government issues documents each month — employment reports, production figures, price indices, trade data — which share their measurements of the convoy’s boundaries and internal dynamics. Yet no one is responsible for fixing the convoy’s position, let alone tracking its trajectory. Without tracking, decisionmakers use simplistic heuristics — are inflation and unemployment too high, is growth too low, are exchange rates stable? — to make decisions about a tangled economy whose internal structure is never examined. Steering the convoy as a whole requires understanding how different sectors relate to one another domestically, and how each relates to the broader world of global trade.

The best record of the broader economy kept by most ships is Dead Reckoning — guesses about how things are built on earlier guesses about how things were. Have they and the people they know been doing better or worse? What do they hear on TV? How are their suppliers and customers doing? Have the prices they care about gone up? Individual ships react to the ships around them, and those reactions propagate across the convoy: price shocks during the pandemic cascaded from sector to sector as businesses and households adjusted to changes nearby. Now reports filter in suggesting danger ahead — climate crisis, inequality, loss of capacity — and that the convoy will need to change direction.

Most ships navigate by feeling, but even the best-navigated individual ships cannot steer a convoy that doesn’t know where it is.

The work of steering towards an affordable and democratic future can only begin with a shared understanding of where we actually are.

To avoid blunders and plan for a real future, someone must fix position for the convoy as a whole and begin keeping a record. That record can then be used to design the new maps necessary for directing renewed public investment — not by commanding every individual ship, but by charting a course the convoy can actually follow.

With this problem in view, Common Wealth US would like to announce the launch of Forces of Production: a monthly log of position fixes for the convoy, covering the supply side of the US economy.

Subscribe now

Each month we will explain what happened the month before based on the newest government data and some reactions from the press. We plan to survey the Producer Price Index, Employment Situation reports, Consumer Price Index, trade statistics, Industrial Production and Capacity Utilization figures, manufacturers’ surveys, and other hard and soft economic indicators to develop a holistic account of what has changed and where.

Every monthly entry will build on the one that came before. Our goal at Common Wealth is not only to keep a record of the economy, but to be able to show where our policy proposals fit into the economy that actually exists, and how better planning and coordination can be achieved. All attempts to change the economy — from industrial policy to macroeconomic management of the business cycle — must be based on an image of the actual forces, the actual capabilities and powers of the economy itself.

Forces of Production offers a record of where the economy has actually been, publishing on the second Monday of each month on data released in the previous month.

At Common Wealth US, we work to complete the climate transition while building toward a decommodified future and democratized. These are immense, existential goals: decarbonization will require us to restructure production at civilizational scale, decommodifying life’s essentials — housing, healthcare, childcare — will mean a fundamental reorganization of the service economy.

These transformations require coordination: the convoy as a whole must change direction in ways no market can organize and no individual actor can command. Though we cannot make a map of the future, we must make maps for the future.

Forces of Production is for those captains who would like the convoy to move beyond dead reckoning, to fix position and find its way in the growing dark. If we do not know where we are, coordination is impossible. But by fixing a trajectory and building a map, we can begin to understand the forces that move beneath us, to find which paths lead to safety and which to wreckage. At Forces of Production we work to start building the common reference point.