Balanced advantage funds: estimated net equity (route A)

Pre-registered specification and results log: docs/research/baf_equity_spec.md. Code: pipeline/tipsheet/compute/baf.py and publish/baf.py. Tests: pipeline/tests/test_baf.py.

What is measured

Balanced advantage and dynamic asset allocation funds (BAFs) move between equity and debt, and they hedge part of their equity with futures. What matters to an investor is the net equity: shares held, minus the part hedged away. Portfolio files show this once a month, a few weeks late, and not for every AMC. This page estimates it every day, for every fund, from the NAV alone.

The estimate.

• Each day, the last 60 sessions of a fund’s daily NAV returns are regressed on the Nifty 50 total-return index and the NSE 5-year G-sec index.
• The equity slope is the estimated net equity. A slope of 0.55 means the NAV has been moving like a portfolio with 55% in the Nifty 50.
• Stock futures hedges against the same stocks cancel out of the slope, which is exactly why the method works.

One plan per scheme. The direct growth plan is used, and the regular growth plan before direct plans existed (January 2013). Both hold the same portfolio, so the slope is the same.

Precision. The typical standard error is 2.5 points, and the regression explains 87% of daily NAV variance (median R²). Each scheme’s standard error is published with its estimate.

Sources

Universe and start dates

• Schemes. 43 schemes in AMFI’s “Balanced Advantage / Dynamic Asset Allocation” category, of which 37 are live (Rs 3.2 lakh crore of AUM). 48 arbitrage funds are run through the same code as a control.
• Per-scheme estimates start 60 sessions after a scheme’s first NAV.
• The industry series starts in June 2018, when SEBI’s scheme categories took effect.
  • Before that, AMFI’s category labels are applied retrospectively to schemes that exist today.
  • Funds that closed earlier are missing.
  • Some of today’s BAFs were different funds then. HDFC Balanced Advantage was HDFC Prudence, a balanced fund.

Industry series

• AUM-weighted mean (headline). Each scheme is weighted by its latest quarterly average AUM that had been published for at least 45 days, so no future information enters.
• Equal-weighted median, published alongside.
• Robustness versions: a 120-session window, and the Nifty 500 as the equity factor.

Checks

Arbitrage funds (control).

• Their equity is fully hedged, so the slope should be about zero. It is: a median of −0.014 across 46 schemes since 2018, and 98.4% of scheme-days within ±0.05.
• The control has crept up to about +0.07 in Aug-Sep 2026, its highest reading, during a falling market. It is being watched.

Validation against disclosed net equity.

• Primary sample: 26 BAFs whose files show their hedges, 1,156 scheme-months from 2019 to 2026.

• Bias. The estimate runs about 3 points below disclosed net on average.
• It tracks net, not gross. The comparison with gross equity shows the estimate is seeing the hedges.
• Arbitrage funds: a median absolute error of 1.2 points.
• The disclosure screen. Three AMCs’ parsed files (ABSL, HDFC, quant) do not net their futures. Their own arbitrage funds show 38–145% “net equity”, which a hedged fund cannot have. Their BAF rows are left out of the primary numbers above. With them included, the median absolute error is 5.4 and the RMSE 11.6.

Known biases

1. The slope measures sensitivity to the Nifty 50, not a holdings percentage.
   • Mid and small caps move more than the Nifty, which pushes the slope up. Defensive large caps push it down.
   • Foreign stocks barely move with the Nifty on the same day.
   • The Nifty 500 version reduces the first problem, not the second.
2. NAV timing. Indian holdings are priced at the same close as the index. Foreign holdings are not, which pulls their contribution down.
3. Options. The slope counts options at their effective delta. The disclosed net figure leaves options out. Funds that use options show a gap: Edelweiss is about 15 points below its disclosed figure.
4. Lag. A 60-session window averages roughly the last three months, so a sharp cut shows up over weeks. A centred window, used as a diagnostic only, lowers the validation error from 4.9 to 4.4.
5. Bond factor. Rate risk in the debt sleeve is assigned to the 5-year G-sec. Credit-spread moves are not modelled.

What the data says (2026-09-30)

• Industry net equity, AUM-weighted:
  • 40% in June 2018
  • about 70–74% through late 2019 and 2020
  • a low of about 40–45% from late 2021 to 2023
  • 59% now (equal-weighted median 57%)
• The largest funds today: HDFC about 64%, ICICI Prudential 62%, SBI 51%, Kotak 57%.
• Valuation-driven? Only weakly, fund by fund, and not at all for the industry.
  • 26 of 32 schemes with three years of history hold less equity when the Nifty PE is higher, but the typical correlation is small (median −0.22).
  • The industry aggregate shows no consistent link. Over 100 months, the AUM-weighted series correlates +0.03 with PE and −0.50 with CAPE. The equal-weighted median correlates −0.46 with PE and +0.44 with CAPE.
  • Three-month changes, the cleaner test, are near zero (−0.04 to −0.23, n = 33).
  • PE and CAPE also disagreed over 2021–2026: PE fell as earnings grew, while CAPE rose. That disagreement is part of why the signs flip.

Bundles

Derived tables in .cache/derived:

• baf_schemes.parquet
• baf_scheme_daily.parquet (every scheme from launch, all variants)
• baf_industry_daily.parquet
• baf_validation.parquet
• baf_checks.json