# Mortgage bond funds dumped at 2020-fastest rates just as yields peak — buy the rebound in mortgage REITs

_AI-generated trading idea · BULLISH · AGNC, NLY_

> Canonical page: https://commonquant.ai/research/for-you/mortgage-bond-funds-dumped-at-2020-fastest-rates-just-as-yie--87a725da-eaf1-566a-8aba-5bbb1f8d0eba

Mortgage bonds are the most crowded exit on the board — investors have pulled out at the fastest rate since 2020, meaning selling pressure is nearly exhausted. JPMorgan's bond desk calls the final leg of the yield surge overdone, and the dovish jobs report gives the Fed room to ease up on hikes, the key driver of falling yields. When yields stop rising, the spread between what mortgage REITs earn on their holdings and what they pay to fund them widens sharply, and these stocks historically snap back hard from deeply discounted book value.

## Idea

Mortgage bonds are the most crowded exit on the board — investors have pulled out at the fastest rate since 2020, meaning selling pressure is nearly exhausted. JPMorgan's bond desk calls the final leg of the yield surge overdone, and the dovish jobs report gives the Fed room to ease up on hikes, the key driver of falling yields. When yields stop rising, the spread between what mortgage REITs earn on their holdings and what they pay to fund them widens sharply, and these stocks historically snap back hard from deeply discounted book value.

## Advanced Analysis

### Verdict: Wait — the fade rules and the bounce thesis haven't reconciled

The macro setup is genuinely interesting: mortgage bond ETF outflows were the fastest since 2020 (Bloomberg, October 1, 2026) and JPMorgan's Misra called the last leg of the selloff 'overdone' a day later, and the Q2 2026 fundamentals back the turn — AGNC swung from a $148M net loss to $654M net income while NLY's net income jumped 191% to $822.7M with operating cash flow flipping from -$1.4B to +$1.6B. The strongest point against is that this is a short-side fade setup whose bullish trigger (RSI back above 35 from 8.4 on AGNC and 6.8 on NLY) is the same event that would invalidate the fade, and the parameter search could not be completed so no robust setup was established. Insider filings lean the wrong way too — about $639,000 of net open-market selling at AGNC for the final June 30, 2026 period, and roughly $1.5M at NLY that is still provisional from a single reporter, so treat the NLY figure as dated and incomplete rather than current. Both names keep diluting (AGNC shares up 1.4% to 1.16B, NLY up 2.8% to 750.6M) and the pair's 0.85 correlation makes this one leveraged rate bet expressed twice. The verdict flips if the next full NLY ownership filing shows insiders reversing into net buying, or if a Fed pivot drives a sustained yield decline that confirms the spread-widening thesis.

#### Conviction score breakdown

Composite score computed by the server from the applicable evidence-tier dimensions.

| Measure | Value |
| --- | ---: |
| Thesis support | 58/100 |
| Trade readiness | 30/100 |
| Risk quality | 45/100 |
| Trigger proximity | 25/100 |
| Fundamentals trend | 62/100 |
| Score | 44/100 |
| Composite Score | 44/100 |
| Evidence Tier | rules\_not\_triggered |

#### Decision scenarios

Bull, base, and bear cases synthesized from the cited evidence tier. Likelihoods are rounded evidence-weighted judgments, not statistically calibrated forecasts.

| Measure | Value |
| --- | ---: |
| Evidence Tier | rules\_not\_triggered |

### Trade now

Nothing to do today — this is a watch-list setup, not an active signal. The compiled entries are short-side fade rules on AGNC (last close $8.79) and NLY (last close $18.50), and on real daily bars they have not opened an entry in the 187 evaluated days over the last nine months. That is the setup waiting for its conditions, not a reason to distrust it. Note the market has run well past the entry zone on the downside: AGNC's 14-day RSI sits at 8.4 and NLY's at 6.8, far below the 35 floor the rules require, so the bounce-side entry condition is the one blocking everything.

Where each entry condition stands for both tickers: price is already below the 50-day average (AGNC $8.79 vs $10.21; NLY $18.50 vs $21.70 — the fresh-cross requirement is not met), RSI is below the 55 cap (met), the 20-day average is below the 50-day (met, $9.74 vs $10.21 on AGNC; $20.73 vs $21.70 on NLY), and the trend-strength filter is comfortably above 20 (AGNC 42.8, NLY 49.2 — met). The two live blockers are a fresh downward cross of price through the 50-day average and RSI moving back above 35 — which, at current oversold levels, is the same event. In plain terms: wait for the stock to bounce hard enough to lift RSI above 35 while staying under its 50-day average; only then does a fade entry arm.

If an entry triggers, the risk frame is mechanical: a 4.0% take-profit against a 2.0% stop-loss — roughly 2:1 reward to risk — plus a hard invalidation on a close above the second resistance level ($9.00 for AGNC, $21.00 for NLY), a close back above the 50-day average, and a 10-trading-day time stop. Position sizing is fixed-risk at 2% per trade with a 25% maximum position. The one scope limitation that matters for your expectations: the parameter search could not be completed because market data coverage for the strategy's dependencies was incomplete, so no robust parameter setup was established — you are watching the published rules exactly as written.

I'd flag one tension with the idea's own thesis, which argues mortgage REITs snap back hard from discounted book as yields stabilize. A violent snap-back is precisely what would push price back above the 50-day average and invalidate the fade before it profits. If you believe the bullish mean-reversion case fully, waiting is the right posture here either way — the rules and the thesis currently point in opposite directions.

#### AGNC price and trigger map

Uses the idea timeframe and keeps price levels on the price axis.

| Measure | Value |
| --- | ---: |
| Ticker | AGNC |
| Timeframe | 1d |

#### NLY price and trigger map

Uses the idea timeframe and keeps price levels on the price axis.

| Measure | Value |
| --- | ---: |
| Ticker | NLY |
| Timeframe | 1d |

### Why the bull case still has support

The core of the idea — that the marginal seller of mortgage bonds has largely been flushed — gets direct support from the cited tape: per Bloomberg (October 1, 2026), investors shed mortgage bond ETFs at the fastest rate since 2020, and JPMorgan's Misra called the last leg of the bond selloff 'overdone' (October 2, 2026). A dovish jobs read, characterized by BlackRock's Rosenberg as giving the Fed room to ease (October 2, 2026), targets the actual driver of mortgage REIT earnings: the funding spread. When yields stop rising, both names' economics improve mechanically rather than through heroic assumptions. The latest fundamentals already show that mechanism flipping on. AGNC swung from a -$148 million net loss in Q1 2026 to $654 million of net income in Q2 2026, with quarterly return on equity moving from -1.2% to +5.2%. NLY did the same thing bigger: net income jumped 191% quarter over quarter to $822.7 million, operating cash flow swung from -$1.4 billion to +$1.6 billion, and ROE rose 3.1 points to 4.9%. These are exactly the spread-wide quarters the thesis predicts once the rate path stabilizes — the Q2 2026 prints (period ended June 30, 2026) are the bull case in motion, not a promise. The income story supports patience while the thesis plays out. NLY pays monthly and has actually raised its payout within 2026, moving from $0.65…

#### AGNC Return on equity

Return on equity trend from CommonQuant fundamentals/XBRL data; -46.0% from first to latest point.

| Measure | Value |
| --- | ---: |
| 2009-12-31 | \0.21683912248628884% |
| 2010-12-31 | \0.18308963763509217% |
| 2010-12-31 | \0.08778130959949142% |
| 2011-03-31 | \0.04006576765861335% |
| 2011-06-30 | \0.019373905429071803% |
| 2011-09-30 | \0.050807083681493434% |
| 2011-12-31 | \0.1239536381197682% |
| 2011-12-31 | \0.033483580167417905% |
| 2012-03-31 | \0.07352603808212893% |
| 2012-06-30 | \-0.02856830122591944% |
| 2012-09-30 | \0.00762884768916881% |
| 2012-12-31 | \0.11719897209985317% |
| Latest Value | \0.11719897209985317% |
| Change Pct | \-45.95118687253317% |
| Ticker | AGNC |
| Timeframe | reported periods |

#### NLY Return on equity

Return on equity trend from CommonQuant fundamentals/XBRL data; -31.6% from first to latest point.

| Measure | Value |
| --- | ---: |
| 2007-12-31 | \0.0796136284428364% |
| 2008-12-31 | \0.048192522850311115% |
| 2009-06-30 | \0.06888283984838378% |
| 2009-09-30 | \0.03064194558502587% |
| 2009-12-31 | \0.205294488648507% |
| 2009-12-31 | \0.07633477929495712% |
| 2010-03-31 | \0.02933448250597565% |
| 2010-06-30 | \-0.0226546989899963% |
| 2010-09-30 | \-0.001466510266719208% |
| 2010-12-31 | \0.1284635424586159% |
| 2010-12-31 | \0.12351914363044736% |
| 2011-03-31 | \0.05445375198679034% |
| Latest Value | \0.05445375198679034% |
| Change Pct | \-31.602474285054317% |
| Ticker | NLY |
| Timeframe | reported periods |

### Scores

- **Conviction score breakdown:** 44
- **Thesis support:** 58
- **Trade readiness:** 30
- **Risk quality:** 45
- **Trigger proximity:** 25
- **Fundamentals trend:** 62

### Watch items

- **AGNC — RSI (14)**
- **AGNC — Price vs EMA (50)**
- **NLY — RSI (14)**
- **NLY — Price vs EMA (50)**
- **AGNC — Close above second resistance level**
- **NLY — Close above second resistance level**
- **AGNC — Next dividend ex-date ($0.12/share monthly)**
- **NLY — Next dividend ex-date ($0.75/share quarterly)**
- **AGNC — Insider open-market net selling**
- **NLY — Insider open-market net selling (provisional filing)**
- **AGNC — Quarterly net income trend**

## Key details

- Symbols: AGNC, NLY
- Timeframes: 1d
- Tags: \#canonical-demand, \#cluster-version:1, \#direction:bullish, \#entity-kind:instrument, \#entity:AGNC, \#entity:NLY, \#horizon:unspecified, \#intent:research, \#symbol:AGNC, \#symbol:NLY

## Community

- Upvotes: 0
- Views: 0
- Copies: 0
- Cosigns: 0

## News sources

- [Investors Shed Mortgage Bond ETFs At Fastest Rate Since 2020](https://www.bloomberg.com/news/articles/2026-10-01/investors-shed-mortgage-bond-etfs-at-fastest-rate-since-2020) — Bloomberg
- [Last Leg of Bond Selloff 'Overdone,' Says JPM's Misra](https://www.bloomberg.com/news/videos/2026-10-02/last-leg-of-bond-selloff-overdone-says-jpm-s-misra-video) — Bloomberg
- [US Jobs Data a 'Dovish' Report, Says BlackRock's Rosenberg](https://www.bloomberg.com/news/videos/2026-10-02/us-jobs-data-a-dovish-report-blackrock-s-rosenberg-video) — Bloomberg

## Related

- [AGNC trade ideas](https://commonquant.ai/markets/agnc)
- [NLY trade ideas](https://commonquant.ai/markets/nly)
- [Latest market news](https://commonquant.ai/news)

## About CommonQuant Research

Ideas and Advanced Analysis are generated with fresh, LLM-selected news headlines and grounded in SEC XBRL fundamentals and peer percentiles. Strategies built from these ideas are automatically backtested (over a timeframe-dependent historical window with walk-forward tuning) and stress-tested before they can go live, then monitored against the news hourly while they run.