Latest Beige Book Reveals a Split U.S. Economy: Growth Continues as Households and Margins Come Under Pressure

Conclusion: U.S. growth has not stopped, but it is not uniform

In one sentence, the latest Beige Book says that the United States is growing modestly across a broad geographic area, but the spending that creates that growth and the people or businesses that bear its cost are increasingly separated. Because ten of twelve Districts reported slight-to-moderate expansion, the geographic breadth alone does not resemble a clear recession entry point. Yet the strength is concentrated in defense, data centers, related manufacturing, nonresidential construction and high-end travel or services. Residential construction, autos, price-sensitive mass-market consumption, crop agriculture and low-margin consumer businesses are experiencing a different economy at the same time.[1]

This distinction matters because the same aggregate number can lead to very different outcomes. As long as businesses can absorb higher input costs through margins, consumer prices may accelerate less than the cost shock would imply and layoffs may remain limited. When that capacity is exhausted, however, firms have to move the adjustment somewhere else: selective price increases, hiring freezes, reduced product variety, shorter opening hours, smaller inventories or delayed capital spending. In the opposite direction, if energy, transportation, insurance and materials pressures ease while strong investment demand spreads into surrounding industries, margins and employment could recover and the expansion could become broader. The economy is approaching that branch point, not yet past it.

The report therefore does not support simple claims such as “the Beige Book ordered a September rate increase,” “recession risk is over,” or “inflation has definitively reaccelerated.” On July 29, 2026, the Federal Open Market Committee maintained the federal funds target range at 3.50–3.75 percent, but three members preferred a 25-basis-point increase. Continued growth preserves an argument for restraint, while weak payroll data and nearly flat real consumption complicate a demand-overheating narrative. The Beige Book is most useful here as a map of that tension across regions and industries—not as a substitute for the policy decision.[5][6][8]

What is the Beige Book, and how much weight should it carry?

The Beige Book is the Federal Reserve System’s qualitative report on regional economic conditions, published eight times a year. The twelve Federal Reserve Banks gather information through interviews, questionnaires and discussions with businesses, community groups, economists, market participants and other contacts, then summarize what has changed since the previous reporting period. The August 2026 edition was prepared by the Federal Reserve Bank of Minneapolis, incorporates information collected through August 24, and was released on September 2. The contact base is not a random national sample, so the report is not a statistical estimate comparable with GDP or the employment report.[1][2]

That characteristic is both a limitation and the reason the document matters. Its limitation is that words such as “slight,” “moderate” and “strong” cannot be translated directly into a GDP growth rate or a payroll count. The mix of contacts can vary by District and over time, and memorable examples can attract more attention than their aggregate weight deserves. Its value is speed and texture: firms often describe price resistance, contract renewals, hiring standards, inventory adjustments, financing conditions, customer trade-down and skill shortages before those mechanisms are visible in official national statistics. The Federal Reserve’s own description says that qualitative information complements data by identifying developments that may not yet be apparent in available statistics.[2]

The best way to use the Beige Book is therefore as an early-warning network rather than a forecast. If national aggregates remain firm while several Districts independently report inventory compression, stalled housing, deteriorating consumer credit and rejected price increases, the cluster deserves attention as a possible precursor to slower activity. If weak monthly data arrive while orders, hiring plans, loan demand and nonresidential construction broaden, temporary noise becomes a plausible alternative explanation. The report should be added to quantitative evidence, not used in place of it.

The basic process for combining macro sources is set out in SG Group’s complete guide to connecting COT data, interest rates, real yields and EIA releases. The same discipline applies here: preserve the information cutoff, publication date, confirmed observations, alternative explanations and the next statistic that can update the hypothesis. A report published on September 2 should not be mistaken for a real-time measurement of conditions on September 2; most of its evidence predates August 24.

What the latest report confirms: activity, employment and prices

The national summary says economic activity increased modestly from early July. Ten of the twelve Districts reported growth in the slight-to-moderate range, while two reported no change. Consumer spending increased slightly on balance, but the accounts combined pronounced price sensitivity with solid high-end purchases. Auto sales were generally subdued because of weak confidence, high fuel prices and rising financing costs. Tourism increased, and airlines reported strong demand despite higher fares.[1]

On the production side, manufacturing improved in many Districts, with defense and data-center orders appearing repeatedly. Services grew slightly to modestly. Lending volumes were steady or higher in many areas, and financial conditions improved slightly. Construction showed a sharp internal contrast: residential building declined, while nonresidential construction rose overall and was unusually concentrated in data-center projects. Describing construction as simply “strong” would erase the most economically important part of the report’s composition.[1]

Employment increased only “very slightly” overall. Three Districts reported modest growth, four slight growth and five no change. Manufacturing, construction and selected services retained healthy labor demand, while retail and leisure-and-hospitality demand weakened. Skilled trades and technical workers were difficult to find. Artificial intelligence was reported as increasing labor demand in some settings and reducing it in others. Wages rose slightly to moderately in most Districts, with larger increases often tied to demand for skilled construction and manufacturing workers.[1]

Prices rose moderately in eight Districts, modestly in two, slightly in one and robustly in one. Compared with the previous report, the pace was unchanged in eight Districts, slower in three and faster in one, which is not a nationwide synchronous acceleration. Input costs nevertheless remained high in manufacturing and construction. Contacts repeatedly cited energy, transportation, metals, petrochemicals, tariffs, health care and insurance. Some consumer-facing firms said highly price-sensitive customers prevented them from passing through all of those costs.[1]

The current quantitative backdrop is mixed as well. July nonfarm payrolls fell by 23,000, the unemployment rate was 4.1 percent, and average monthly job growth over the previous twelve months was 34,000. July CPI rose 0.1 percent month over month and 3.4 percent year over year; core CPI rose 0.2 percent and 2.5 percent. July personal income rose 0.4 percent, nominal PCE rose 0.2 percent and real PCE increased by less than 0.1 percent, while the saving rate was 3.0 percent. The second estimate placed second-quarter real GDP growth at a 1.5 percent annual rate, but real final sales to private domestic purchasers increased 4.2 percent. No single one of those figures resolves the qualitative split in the Beige Book.[6][7][8][9]

What changed from the previous two Beige Books

Reading a single edition in isolation makes it easy to exaggerate direction. Placing the May, July and August 2026 reports side by side shows that activity remained broadly positive while the sources of strength became more concentrated. The May report had ten growing Districts, one unchanged and one declining slightly. July had eleven growing and one unchanged. The latest report has ten growing and two unchanged. By District count alone, the breadth narrowed slightly from July, so the document does not support a claim that growth accelerated everywhere.[3][4][1]

Employment offers a limited improvement in breadth rather than a hiring boom. In July, five Districts reported modest, moderate or solid gains and seven reported little or no change. In the latest edition, three reported modest growth, four slight growth and five no change, so seven Districts reported some increase. Yet the national description remained “very slight.” Hiring improved across more geography, but not with much depth, and skill-specific shortages persisted.[3][1]

Prices are more subtle. The July report described moderate increases in nine Districts, robust increases in two and slight increases in one; every District reported the same or a slower pace than before. In August, eight were moderate, two modest, one slight and one robust; eight were unchanged in pace, three slower and one faster. The simplest defensible conclusion is that a broad reacceleration in final prices was not confirmed. That does not mean business costs stabilized: energy, transportation, metals, petrochemicals and insurance remained prominent. Stable final prices and stable input costs are different propositions.[3][1]

This distinction also matters for regime analysis. SG Group’s guide to point-in-time macro regime classification explains why information available on a given date should be preserved rather than rewritten using later revisions. Treat each Beige Book as a fixed information set. Do not retroactively change what could have been known on September 2 after future payroll, GDP or inflation revisions arrive.

Low-hire, low-fire—and the “skill islands” hidden by labor-market averages

The labor picture is neither a collapse nor a renewed boom. Employment rose very slightly overall and was unchanged in five Districts. Distribution matters more than the aggregate. Manufacturing, construction and selected professional services still need workers, and skilled trades and technical staff are hard to find. Retail and leisure-and-hospitality demand has weakened. In some administrative and entry-level roles, AI deployment and business redesign are reducing recruitment. Firms are often reluctant to dismiss existing employees but cautious about adding new ones: a low-hire, low-fire labor market.[1]

July 2026 nonfarm payrolls fell by 23,000, while average monthly growth over the prior twelve months was 34,000. The unemployment rate remained 4.1 percent, but labor-force participation was 61.4 percent and the employment-population ratio 58.9 percent—down 0.7 and 0.5 percentage point respectively since January. A stable unemployment rate can conceal cooling if fewer people are participating. The payroll decline itself may contain monthly noise and future revisions, so neither the survey nor the Beige Book should be used alone to make a recession call.[6]

The result feels very different to an incumbent employee and to a person trying to enter or re-enter work. Existing workers may not see mass layoffs, while graduates, career switchers, unemployed workers and people unable to relocate encounter scarce openings. Employers can simultaneously report that labor is not generally scarce and that the exact skills they need are unavailable. Some wages can therefore accelerate even as aggregate employment weakens; skill-specific wage growth is not automatic evidence of economy-wide excess demand.

The Beige Book reported both positive and negative AI effects. That finding rules out slogans such as “AI created jobs” or “AI eliminated jobs” as a complete national conclusion. Some firms use AI to raise productivity, expand demand and hire more technical staff. Others automate screening, entry-level analysis, routine administration or customer service and simply do not create a vacancy. Outcomes also depend on whether existing workers move to higher-value tasks, outsourced work is brought in-house, or new capacity creates additional demand.

General reporting often overweights the stable unemployment rate and anecdotes about worker shortages. Neither proves that the entire labor market is strong. It often underweights the delayed effect of weaker hiring. Layoffs reduce income immediately; fewer hires gradually lengthen the time it takes graduates, job changers and displaced workers to find work, weakening wage bargaining and consumption after several months rather than on the announcement date.

The September FOMC decision matrix: neither a rate-hike order nor permission to ease

The FOMC kept the federal funds target range at 3.50–3.75 percent on July 29, 2026. The statement described economic activity as solid and inflation as above the 2 percent objective. The vote was 9–3; the three dissenters preferred a 25-basis-point increase. The next meeting is scheduled for September 15–16 and is due to include updated economic projections.[5][12]

The latest Beige Book contains evidence for restraint and evidence for caution at the same time. The restraint case includes growth in ten Districts, defense and data-center investment, skilled-trade wages and continued energy, tariff, insurance and input pressure. The caution case includes very slight employment growth, weak housing and auto demand, broad price sensitivity, limited pass-through, the July payroll decline and near-stagnant real PCE.[1][6][8]

Rates analysis should distinguish the policy rate from the two-year and ten-year Treasury yields, real yields, inflation compensation and term premium. As SG Group’s guide to Treasury yields and the yield curve explains, a long-rate increase has different consequences depending on whether it reflects expected policy, real growth, inflation expectations, fiscal risk or term premium.

Before the meeting, the August employment report is scheduled for September 4 and August CPI for September 11, both at 8:30 a.m. U.S. Eastern Time. Because the Beige Book’s information cutoff was August 24, those releases can update conditions after much of its contact evidence. Fixing a policy conclusion immediately after the Beige Book is less useful than comparing payrolls, wages, participation, core services, energy and prior-month revisions.[10][11]

Five common misreadings—and the strongest alternative hypothesis

Misreading 1: Ten growing Districts mean the U.S. economy is reaccelerating

The District count measures geographic breadth, not the magnitude of growth or contribution to national output. Eleven Districts grew in July, compared with ten in the latest report. Several Districts also share the same data-center or defense engine. Geographic breadth and diversity of demand sources are different. A genuine broad reacceleration would need to reach general services, housing, real consumption volumes, non-data-center capital spending and hours worked.

Misreading 2: Moderate final prices mean the inflation problem is almost over

When input costs are high but final prices are restrained, margins may be absorbing the gap. That can suppress consumer inflation for a time, but it is not costless or indefinitely sustainable. Lower margins can push adjustment into prices, investment, hiring or product scope. If the original costs decline, however, future increases are not inevitable. The relevant evidence is the gap between input and selling prices, unit volumes, gross margins and investment plans.

Misreading 3: A stable 4.1 percent unemployment rate proves labor is strong

Unemployment is important, but participation, the employment-population ratio, hires, hours, long-term unemployment and entry-level openings complete the picture. A low-hire, low-fire market can protect incumbents while making job search difficult. Skilled shortages are more likely to reflect occupational and regional mismatch than a universal shortage of labor.

Misreading 4: AI and data centers will automatically spread growth

Capital expenditure creates near-term demand, but its long-run effects on productivity, profitability, employment and electricity costs remain uncertain. Post-construction staffing, utilization, grid upgrades, financing costs and technological obsolescence matter. The Beige Book reports positive and negative employment effects from AI and does not quantify a net national effect.

Misreading 5: The Beige Book tells us the FOMC decision in advance

The Beige Book is one input into policy deliberations, not a preview of the vote. Much of the information stops at August 24, and employment, CPI, financial conditions and international developments update afterward. The document is not an official statement of Fed officials’ views. A policy judgment requires data, expectations, forecasts, risk management and the distribution of votes.[2]

That objection deserves weight. Strong private domestic final demand, affluent consumption, airline traffic, loan volumes and manufacturing orders show that substantial underlying demand remains. July payrolls may also be revised, and one negative month cannot establish a downturn. Lower energy prices could restore real income and margins relatively quickly.

The objection does not eliminate the two-layer interpretation, however. A strong aggregate can coexist with an unusually concentrated contribution. The test is whether private demand remains broad after removing the largest capital projects and the highest-income consumption, and whether the household-credit and margin buffers strengthen or erode. The article’s thesis is not that a downturn is certain; it is that resilience depends on buffers whose durability must be measured.

A methodological objection to the Beige Book itself

Contacts are not randomly selected, qualitative terms are not calibrated like a statistical index, and the prominence of a sector can reflect the current contact mix. These are valid limitations. The remedy is not to discard the report, but to demand independent confirmation: repeated language across unrelated Districts, consistency over successive editions, agreement with hard data, and a plausible transmission mechanism. A single anecdote is not a national trend. A repeated pattern appearing in several Districts can be an early signal worth testing.

The core question is how long the economy’s shock absorbers can hold

SG Group reads the latest edition neither as confirmation of recession nor as confirmation of a completed soft landing. It is a transition in which strength on the supply-and-investment side is balancing weakness in household purchasing power and corporate profitability. Three buffers currently support the system: spending capacity among higher-income households; large capital projects and defense demand; and the ability of firms to absorb higher inputs through margins and operating adjustments.

While all three work, aggregate demand does not collapse, final prices rise less than inputs and mass layoffs remain limited. The headline therefore looks like “modest growth.” But those buffers are not free. Dependence on affluent spending narrows the consumer base. Dependence on a small group of investment themes increases financing and project-economics risk. Margin absorption reduces the capacity to hire, train, maintain and invest later. An economy can look stable precisely because the burden has not yet reached the most visible headline statistic.

That is why behavior before a formal price increase deserves special attention. Firms can shorten inventory days, order less frequently, reduce product variety, raise hiring standards, delay replacement, switch suppliers or require a shorter payback period before approving capital expenditure. Those actions may not appear directly in CPI, yet they change future supply, quality, employment and competitiveness. The Beige Book’s distinctive value is its ability to reveal adjustments that have not yet become a statistical headline.

The second core issue is whether data-center investment should be treated as a cyclical engine or a structural transformation. If it stimulates power, semiconductors, construction and software broadly and raises productivity, today’s concentration can become tomorrow’s diffusion. If too much capital chases the same demand outlook while grid and financing constraints delay completion or utilization, the eventual retrenchment could hit several Districts at once. Construction spending alone cannot establish the long-run productivity return.

This view is explicitly falsifiable. If real household income and consumption volumes recover, housing bottoms, capital spending broadens beyond data centers, and corporate margins improve without renewed inflation, this article has placed too much weight on downside risk. If payroll weakness persists after revisions, consumer-credit quality deteriorates and nonresidential investment becomes even more dependent on one theme, it has placed too little weight on recession risk. The conclusion is a monitoring hypothesis, not a fixed story.

What it means for households, careers, businesses, markets and Japan

Households: the combination of fixed costs and financing matters more than the average inflation rate

For a household, the headline CPI rate is less important than how fuel, housing, insurance, health care, food and loan rates overlap in its own budget. A large year-over-year gasoline increase has different consequences depending on commuting distance, the need to own a car, mortgage terms and insurance renewal dates. Strong travel spending by high-income households supports the average but does not offset the cost of necessities for middle- and lower-income groups.

With a low saving rate, households can adjust through saving drawdowns, card use, postponed replacement, trade-down and lower discretionary spending. Those responses do not occur all at once. Optional spending may weaken first, durable purchases later, and delinquencies or restructuring only after buffers are exhausted. Nominal retail sales are therefore insufficient; volumes, credit-card, auto and mortgage delinquency, interest expense and real income provide the transmission path.

Work: the experience splits by occupation and point of entry

Skilled trades, data-center construction, electrical equipment, manufacturing technology and defense-related roles retain bargaining power. General administration, junior technical work, retail and hospitality face softer demand, automation, preference for experienced applicants and hiring freezes. The same labor market can feel stable to a person already employed and closed to a graduate or career switcher. Training helps only with a lag and remains constrained by location, licensing, mobility and the time required to acquire skill.

Business: the number of available adjustment tools matters more than abstract “pricing power”

Large firms are more likely to have long-term contracts, hedges, diversified suppliers, capital-market access and a broad product mix. Small businesses often buy in smaller quantities, devote a larger share of cost to rent, insurance or benefits, and lose customers more easily after a price increase. The same margin decline can have radically different survival implications depending on financing capacity and the timing of renewals. Management must coordinate price, volume, inventory, payment terms, hiring and capital spending rather than asking only whether it can raise prices.

Markets: the same Beige Book can support opposite price reactions

Resilient activity can support higher short-rate expectations, while weak households and employment can lower long-run growth expectations. Input pressure can raise inflation compensation, while margin compression can lower equity earnings expectations. Data-center investment may support related equities, credit and electricity demand, but also increases concentration and capital-cost risk. Prices react not only to facts but to their difference from expectations and to existing positioning.

The dollar cannot be inferred from U.S. rates alone. The policy, growth, inflation, risk appetite and capital flows of the other currency are part of every exchange rate. SG Group’s relative framework for interest rates, inflation, growth and FX flows is designed to prevent a one-country headline from being translated mechanically into a currency-pair forecast.

Japan: the first transmission may come through U.S. rates, the dollar and energy—not exports

Japan receives the shock through at least four channels. First, U.S. rates and dollar-yen affect imported costs, hedging and financing. Second, energy and freight affect households and corporate margins. Third, U.S. demand affects Japanese machinery, semiconductor equipment, components and other exports. Fourth, global capital costs change valuations and financing even for businesses without direct U.S. sales.

Strong U.S. data-center and defense investment can benefit selected Japanese suppliers, but higher U.S. rates or a stronger dollar can raise imported energy and hedge costs. A U.S. slowdown can reduce export demand yet also lower energy and yields. The net effect depends on the Bank of Japan, oil prices, corporate currency exposure and whether U.S. growth broadens beyond capital projects. “A stronger U.S. economy is good for Japan” is therefore too simple.

What remains unknown—and the releases that can change the assessment

The Beige Book cannot tell us the quantitative scale of each reported change. “Moderate price growth” and “strong orders” cannot be converted into national dollars. It cannot allocate input pressure precisely among energy, tariffs, insurance, wages and supply constraints. The permanent productivity and employment effects of AI and data-center investment remain unknown until projects are completed and used. Nor can it show how long households can defer the burden through credit and saving without income-group and delinquency data.

Remaining margin-absorption capacity is also unobservable in this document. Listed-company earnings overrepresent larger firms and do not fully describe private or small-business cash flow. A company can avoid a price increase and preserve customers while cutting training and equipment replacement, moving the burden into the future. If input pressure is temporary, the same company may instead rebuild profitability without losing volume. Both paths remain open.

The August employment and CPI releases include information after the Beige Book cutoff. If payroll weakness is revised away and inflation slows, the split economy moves closer to a sustainable soft landing. If jobs weaken further while core inflation or expectations remain high, the economy faces a difficult combination of softer demand and persistent supply cost. The judgment must be based on a contemporaneous combination, not a single headline.[10][11][12][13]

In the next Beige Book, count more than the number of expanding Districts. Track how many independently use language such as “price sensitivity,” “inventory compression,” “hiring freeze,” “delinquency,” “discounting” or “investment delay.” Watch whether orders spread beyond defense and data centers and whether the decline in housing stops. Repeated language across unrelated Districts is how a local anecdote begins to resemble a national pattern.

The key question is no longer whether growth exists, but where the burden is being placed

The latest Beige Book does not show that the U.S. economy stopped in the summer of 2026. Activity increased in most Districts, and parts of manufacturing, nonresidential construction, lending and travel were firm. Declaring an immediate recession would go beyond the evidence. Ignoring weak housing, autos, price-sensitive consumption, corporate margins and entry-level hiring to declare a completed soft landing would go beyond it in the other direction.

The central feature is that the burden from external costs and interest rates has not yet converged into one statistic. Firms are absorbing some through margins; households through saving, credit and postponed purchases; and concentrated investment is supporting the average. GDP, payrolls, CPI and earnings can therefore point in different directions. One is not necessarily correct and the others wrong—they measure different layers and different lags.

For the FOMC, continuing growth means inflation cannot be dismissed, while weak households and labor mean a supply shock cannot be answered mechanically with tighter policy. The branch point depends on whether input costs fade before reaching final prices and wages, whether margin compression reaches hiring and investment, and whether capital expenditure converts into productivity and broad income.

This conclusion is not a forecast of an asset price or a recommendation to buy or sell. A practical monitoring process should place the September 4 jobs report, September 11 CPI, September 15–16 FOMC meeting, September 30 PCE and GDP revision, and October 14 Beige Book in the same evidence table and update only the assumptions that change. Turning a news release into a falsifiable monitoring framework is the most useful way to read the Beige Book.