The Siphoning Effect of Capital: One Year Later, Is Your Job Still Safe?

Special Topic: The Siphoning Effect of Capital One Year Later, Is Your Job Still Safe?

2026-10-18 · Macro Analysis

> ⚠️ This article is macro analysis, not a prediction, and does not constitute investment advice. Data sources and full notes are at the end.

---

Two chains converging in the same place

The October 9 special analysis "The Seventh Month of the Oil Shock: Stagflation Risks Hidden by the AI Rally" traced the channel through which the oil shock transmits via prices: energy prices push up overall inflation, the Federal Reserve places inflation ahead of employment, and the tightening costs affect employment.

This piece continues that analysis. In the same year, capital is concentrating toward a few industries, and the concentration itself pushes up borrowing costs for other firms. This channel is on top of price costs, and is another direct factor affecting corporate cash flow.

The two chains ultimately converge in the same place: employment.

One number illustrates what this article is about. Over the past year, the group of highest-rated firms saw their actual bond issuance cost rise from 4.80% to 5.86%, an increase of only 1.06 percentage points. The lowest-rated group rose from 11.68% to 17.50%, an increase of 5.82 percentage points. The same rise in interest rates was felt 5.5 times more by the latter than by the former.

---

1. The past year's gains came from three places

Over the past year (as of mid-2026-10), the S&P 500 as a whole rose 16.0%, but that gain was mainly driven by a few sectors, with substantial divergence across sectors (including dividends):

The drivers of these leading sectors are: energy from oil prices, with crude up 134% over the past year; semiconductors and technology from AI-related capital expenditure; healthcare from repricing and M&A around biosimilar patent expirations.

Note the direction of concentration. Splitting the year into two halves:

In the first half, ordinary stocks outperformed the index; in the second half the opposite occurred, with gains concentrating in a few sectors. The full-year difference is only 0.8 percentage points because the two halves offset each other. The breadth of the rally has clearly narrowed in the most recent six months.

---

2. In the same year, funding costs split into two worlds

Actual corporate borrowing costs, broken down by credit rating level:

Breaking this down, the source of the gap is clear. The risk-free real rate (the return on lending to the U.S. government after subtracting inflation expectations) rose from 1.77% to 2.78%, an increase of about 101 bps. Investment-grade firms bore exactly that 101 bps—their credit spreads widened by only 9 bps. The CCC tier bore that 101 bps plus an additional 480 bps of credit spread widening.

The same rise in rates was passed through to the top level as-is, and amplified more than fivefold at the bottom.

The absolute level of 17.50% needs clarification: at that price, firms typically stop borrowing and instead turn to cutting expenditures. There are limited ways to cut costs, and labor is one of the major items.

This tier's pressure is already visible in the equity market. The Consumer Discretionary sector fell 3.4% over the past year, and Small-cap lagged the S&P 500 by about 6.9 percentage points—small businesses are the main borrowers in this tier and also employ the most people.

---

3. Rate hikes cannot solve these two issues

The previous piece covered the first issue: this inflation is sourced in oil supply, and interest rates cannot fix chokepoints in straits, refineries, and tankers. Overall inflation is now 4.02%, versus 2.9% a year ago; the Producer Price Index for all commodities is up 12.8% year-on-year, versus 3.7% a year ago. Core inflation remains at 2.8%, which is the only defensive line not yet breached.

The second issue relates to long-term rates. Decomposing the change in the 10-year U.S. Treasury yield over the past year:

The rise almost entirely comes from the real rate, while inflation compensation slightly declined over the year. According to estimates from Fed staff models, about half of the real rate rise comes from the term premium—the market demanding more compensation to lend to the government for twenty years.

Sources of the term premium typically include the scale of fiscal issuance, central bank balance-sheet runoff, the participation of foreign buyers, and inflation uncertainty. These factors do not automatically disappear as the economy slows.

This implies one thing: the Federal Reserve can influence short-term rates, but its effect on long-term rates is much smaller. Even if policy turns dovish, the long end may not follow. In past cycles the channel "economic weakening → rate cuts → lower financing costs" worked end-to-end; this time it has only worked halfway.

---

4. Employment data hasn’t deteriorated, but the market is stagnant

On the surface, employment metrics are fairly steady: unemployment rate 4.2%, one year ago 4.3%; broad unemployment (U-6) 7.6%, one year ago 8.1%; initial jobless claims 199,000, one year ago 222,000. All three numbers have improved.

Looking one layer deeper, the picture is different:

Nonfarm payrolls added an average of 51,000 people per month over the past three months, and an average of 41,000 per month over the past twelve months. Hiring is essentially stalled, yet the unemployment rate has not risen.

Putting these numbers together describes the same state: firms are neither hiring nor laying off. Layoff rates are at historic lows, so the unemployment rate remains stable; hiring rates are also very low, so when a worker does lose a job it takes on average 24.8 weeks to find the next one—nearly half a year. Long-term unemployed increased by one quarter over two years. The labor force participation rate fell 0.9 percentage points—some people who cannot find work exit the labor force and are therefore not counted as unemployed. (Part of the participation decline is also due to population aging and cannot be entirely attributed to this dynamic.)

This state’s characteristic is that it can only be resolved in one direction. The layoff rate at 1.0% is already near historical lows, leaving little room to decline further; hiring is also low, meaning that if layoffs increase, there is not enough hiring to absorb them. Under these conditions, changes in the unemployment rate are more likely to be sudden jumps than slow climbs.

---

5. If the siphoning of capital continues, layoffs are one possible path

Connecting the previous four sections, the path is roughly:

Capital continues flowing into a few capital-intensive industries → aggregate demand for capital rises, pushing up real rates → the 101 bps is passed through at the top and amplified to 582 bps at the bottom → firms with borrowing costs near 17.5% essentially exit the credit market → these firms turn to cost cutting, with labor a primary target → the labor market, already in a "no hiring, no firing" balance, cannot absorb increased layoffs due to lack of hiring capacity → unemployment rises.

Each link in this chain is a premise and could fail: the three sectoral drivers must continue (oil prices, AI capex, patent expiry cycles); the credit-spread divergence must continue to widen rather than converge; firms’ spending-cutting order may not primarily hit labor first. So this is a path to watch, not a conclusion.

Notably, the capital demand on the AI side is changing itself. As of late September 2026, an observation showed that major buyers of AI infrastructure had capex equal to 90.3% of operating cash flow, versus 68.7% a year earlier. When capex reaches this share of operating cash flow, further expansion is more likely to rely on debt, and the price of that debt is precisely the rising real rate discussed earlier. The forces that push up funding costs will ultimately affect the parties that pushed them up.

---

6. There are few historical precedents

Compared with past episodes, the most unusual feature this time is the degree of divergence within the credit market.

Using Moody’s Baa versus Aaa corporate yield spread to measure the market’s differentiation by credit quality, this series goes back to 1986:

The current value is 0.49%, at the 0.8th percentile over forty years. In ten thousand trading days, fewer than one percent of days show the market less willing than now to differentiate credit quality.

At the same time, CCC and below credit spreads are at the highest level in the available data (three years).

In past episodes when low-rated firms ran into trouble—such as the 2015–2016 energy high-yield debt crisis and the first half of 2007—the divergence on the investment-grade side was above historical medians. In other words, those episodes featured stress across the whole credit market, with the low end being even tighter.

This time’s combination—top-tier unusually calm in forty years, bottom-tier at the high end of comparable data—has no counterpart in the available historical record.

No precedent means history cannot provide an outcome reference. That does not mean the result will necessarily be worse, nor that there is no cause for concern. It means this time we must rely on signposts rather than analogies.

---

7. Three questions ordinary people need to answer for themselves

Are wages keeping up with inflation?

This is the most direct question and there are public data for comparison.

(Hourly wages use the production and non-supervisory worker series, covering about 80% of employment and excluding senior management.)

A year ago real hourly wages were still growing; now they are negative. Over the same period the personal savings rate fell from 6.2% two years ago to 4.1% now, and revolving consumer credit (mainly credit cards) balances grew 3.37% year-on-year. When wages do not keep up with prices, the spending gap is first filled by savings and credit.

Compared with the 1970s, one clear difference exists. Then, inflation became self-fulfilling because wages automatically rose with prices in a spiral. Today, real hourly wages are negative, and 10-year inflation compensation moved only −3 bps over the year; both indicate that such a spiral has not formed. This reduces the likelihood of the kind of prolonged, large-rate hikes that occurred then, but also means the cost of rising prices is borne more by wage earners.

Will my job still be safe in a year?

There is no universal answer, but one judgment angle is: which tier does your employer borrow in?

Capital is flowing into three capital-intensive industries that employ relatively few people. Pressure is borne by firms financing in the high-yield and CCC tiers—small and medium-sized enterprises, consumer-facing companies, and firms with high shares of floating-rate debt. These happen to be the categories that employ the most people.

Note the timing lag: credit spreads reflect market pricing of risk, while defaults and layoffs typically lag. The regional banking sector rose 15.3% over the past year, indicating lenders have not yet taken large loss provisions. The pressure is visible in prices but not yet in outcomes.

Also compare your personal debt: fixed-rate portions of personal debt see their real burden fall with inflation; parts that reset with benchmark rates (credit cards, some consumer loans) move the other way. At the corporate level, the CCC move from 11.68% to 17.50% is the same mechanism playing out at a larger scale.

Where should investors be cautious?

Sorting the more than one hundred tracked ETFs by their real returns over the past year (nominal return minus 4.02% inflation) reveals a clear dividing line: fixed-coupon products all underperformed, while products whose coupons reset with rates held up.

First, the underperformers. Bond products with fixed coupons all underperformed inflation this year without exception; real returns were roughly between −18% and −2%. The reason can be deduced from earlier breakdowns: coupons are fixed at issuance and do not rise with prices, while the discounting real rate rose about 101 bps.

Two asset classes typically seen as inflation hedges also did not match this year’s risks. Gold products had real returns around −5%; gold has no cash flow and its price mainly corresponds to real rates, so it is pressured when real rates rise, irrespective of inflation. Inflation-protected bonds protect against changes in inflation expectations, and inflation expectations changed little this year; their losses also stem from rising real rates. These asset types are not flawed per se, but the risk they hedge is different from the risk that actually materialized this year.

Two directions on the other side deserve attention.

The first is floating-rate credit products, such as bank-loan funds. These products hold corporate loans with interest rates reset every few months; when rates rise, the interest they receive rises too, so they do not suffer simultaneously from both rising rates and inflation. In the observed set, they are among the few non-equity categories with positive real returns after adjusting for inflation.

The risk must be seen together. The borrowers in these products are precisely those that experienced the 582 bps cost increase mentioned earlier—their true risk is credit, not rates. If the layoff path discussed in Section 5 plays out, these borrowers would be affected first. In other words, the advantage of these products in this environment comes from the rate side, while their vulnerability maps directly to the other major risk in this environment.

The second is shortening duration. Whether Treasuries, corporate bonds, or inflation-protected bonds, shorter-maturity products clearly outperformed longer-dated ones this year. For example, with inflation-protected bonds of the same type, just shortening duration yields differences in real return exceeding three percentage points. Duration is the cheapest variable here that investors can control themselves.

Both directions rest on the same premise: real rates remain at high levels. If real rates reverse and decline, fixed-coupon long-term bonds would switch from worst-performing to best-performing, and the advantages of the two directions above would disappear. When the premise changes, conclusions must be recomputed; the signposts in Section 8 are meant to judge whether the premise has changed.

Two more cautions:

First, cross-asset diversification provided limited protection this year. Bonds as a whole underperformed, while stock outcomes depended entirely on the theme: energy- and semiconductor-related products all beat inflation, China-related products all underperformed, with a range from +79% to −40%. In such an environment, the effectiveness of diversification depends more on whether themes are diversified than simply whether asset classes are diversified.

Second, the best-performing themes this past year were already the ones that rose 50%–60%. Using "it performed well last year" as a rationale is hard to distinguish operationally from chasing a trade that has largely priced in expectations. Each driver mentioned earlier—oil prices, AI capex, patent expiry cycles—depends on its own conditions continuing, and each condition has its own fragilities.

---

8. Signposts to watch

The following indicators reflect how far the path in Section 5 has progressed. None are currently breached, which helps explain the market’s overall calm, but their subsequent moves deserve continual monitoring.

---

Limitations of this article

First, the path analysis is based on public data and judgment from historical experience. The chain in Section 5 describes a possibility; ETF Radar’s quantitative system did not assign it a probability. Each link has a chance of not holding, as noted in the text.

Second, some data have definitional limits. The term premium uses Fed staff model estimates, not a directly observable market price; different models yield different estimates. The Baa–Aaa spread is a yield difference, not an option-adjusted spread, so the two are not directly comparable, and the composition and maturities of related indices have changed over decades. Public data for high-yield and CCC credit spreads currently go back only about three years, so references to "highest" for those items refer only to that period.

Third, the return metrics. Returns cited are total returns including dividends, with the window from mid-2025-10 to mid-2026-10. Real return is nominal return minus concurrent overall inflation of 4.02%, which is a simplification. Specific ETF names are not listed; the figures come from representative large funds in each category and may vary across funds.

Fourth, interpretation of employment data. The decline in labor force participation includes both demographic structure and job-search willingness; this article does not decompose the two.

---

Conclusion

The U.S. equity market’s gains over the past year mainly came from three directions: energy, AI-related semiconductors, and healthcare, and the breadth narrowed in the recent six months. In the same year, corporate borrowing costs split into two worlds: the best credit cohort saw only a 1.06 percentage-point increase, while the weakest cohort saw a 5.82 percentage-point rise. The source of the gap can be decomposed—the risk-free real rate rose about 101 bps, passed through at the top and amplified more than fivefold at the bottom.

Inflation shows no sign of easing: overall inflation 4.02%, producer prices up 12.8% year-on-year, core inflation 2.8% remains an unbroken line of defense. Rate hikes have limited effect on two issues: constraints in oil supply and long-term rates driven by term premium.

Employment data appear stable on the surface but are actually in a "no hiring, no firing" stall. Layoff rates are near historical lows, but average unemployment duration reached 24.8 weeks, participation fell, and long-term unemployed rose by one quarter over two years. Once this balance breaks, unemployment is more likely to jump than to creep up.

If capital concentration continues, the burden on many firms will increase and layoffs are one possible path. Every link in that path might fail, but there is little historical precedent to guide us—the top of the credit market is unusually calm in forty years while the bottom is high in comparable data; this combination lacks precedent in the record.

For wage earners, the most direct figure is that real hourly wages have turned negative. For asset allocators, the dividing line this year was whether coupon payments reset with interest rates: fixed-coupon products universally underperformed inflation, while floating-rate credit and shorter-duration bonds held up. Traditional inflation hedges did not necessarily correspond to this year’s risk, and diversification’s effectiveness depends more on theme diversification than asset-class diversification.

These judgments hinge on the signposts in Section 8. None have been breached yet, but their subsequent moves deserve continued attention—the progression of this path will be reflected in these numbers earlier than in stock indices.