Arbitrage Funds: Low Risk, Tax Efficient…Too Good to Be True?

We’ve already covered arbitrage funds in this blog  and this one. Both these articles show arbitrage funds in a fairly positive light. That wasn’t the intent, but for once, what is advertised does seem to hold true. 

I was more curious about how these funds work, and if there’s any hidden risks when holding these kinds of funds. So I dug deeper. Typically, the kind of cash-future carry arbitrage strategies these funds deploy are negative skew in nature. So my first bit of investigation was in that direction.

Negative skew

Short primer

When an asset has a higher chance of a large down move than a large up move, then the asset’s returns are negatively skewed. Imagine making 0.02% return everyday for a year and then losing 10% in a single day - painful, right?

More often than not, if a trading strategy is making steady money and most trades are winners, there’s a good chance you’re in negative skew territory and you haven’t seen the big loss yet.

And the worst part is that the regular daily wins are small enough that folks end up taking leverage to trade these strategies. A negative skew strategy often requires leverage to achieve decent absolute returns when the going is good, but the same leverage will destroy you when things turn sour. Check out these articles on leverage that we've written before: Leverage in Trading: Long Options and Leverage in Trading: 2.25 Crores in 8 minutes.​

Negative skew in Arbitrage Funds

Now, lets take a look at a Sundaram arbitrage fund to really drive home the point of negative skewness:

Let’s understand this with an example, check out this website.

You see a nice 45 degree graph, looks like the arbitrage fund has done well since inception here.

But, the devil is in the details. Let’s change the allocation strategy from monthly SIP to lumpsum. Now, we see a dip in 2018 which was hidden since the SIP smoothened the volatility out:

You zoom in on this period and you’ll see that your portfolio fell from a peak of 11,323 to 10,727 from 7th September to 25th September. That’s a crazy annualised return during this period for a low risk fund.

Simple return = (Final Value − Initial Value) / Initial Value

              = (10 717 − 11 323) / 11 323

              = −0.05352 →  −5.352 %

Annualised return = (1 + simple return)^(365 / 19) − 1

                  = (1 − 0.05352)^(365 / 19) − 1

                  ≈ −0.6523   →  −65.23 % per year

It took it close to 18 months to recover its NAV back - you're in an 18 month time drawdown! Typical negative skew behaviour. 

What happened? During the IL&FS crisis, they had invested in DHFL bonds. Here’s the link to their disclosure.​

(I guess they haven’t mentioned the arbitrage fund in this pdf, maybe because the pdf was dated before they took a hit on the arb fund? Who knows.)

An arbitrage fund manager only needs 65% in equity, they’re free to invest the remaining 35% elsewhere, which is why this can happen if you’re invested in some bond that defaults.

You can confirm this on investHQ by backtesting a “buy and hold” strategy for Sundaram Arbitrage Fund: 

You see a max drawdown of -5.28% over a 10 year backtest, which is much higher than any other mutual fund in this category. For example, the Tata Arbitrage Mutual Fund has a max drawdown of -0.57%, but it wasn’t around for the 2018 period so maybe it's not a fair comparison.

So, I created this table below, and we can see 18 Arbitrage Funds that existed before this drawdown date, and this is their drawdown.

We’ve run a “buy and hold” backtest of every fund in the Arbitrage fund subcategory and created this table.

It seems that Sundaram is the black sheep here. So I guess the SEBI risk category for this fund category is “Low risk”, which is then by default being applied to each fund in the subcategory as well. 

Actually, let’s talk about SEBI’s riskometer next.

Riskometer

I'm sure we're all aware of the SEBI prescribed Riskometer that shows the relative riskiness of each mutual fund. 

We’ll do something different here - we’ll try to establish a risk profile by only looking at the historical NAV data, and ignoring the riskometer. It’s also easier to reason about risk of an asset when looking at it relative to the risk of other assets, so let’s compare HDFC MM with Sundaram Arbitrage. (Why HDFC MM? It’s my default cash parking option based on a legacy recommendation from a friend - from an era when I cared very little about my asset allocation and just needed something better than an FD).

Let’s look at their existing fund pages, published on their own websites. We’ve already seen the max drawdown of Sundaram Arbitrage Fund above. Now, let’s look at the max drawdown of HDFC MM, let’s do a 10 year backtest here:

It has a much lower max drawdown → 

See, it lost a bit of money during the covid period but it gained it back in the next few days. A much lower max drawdown and a much lower max time drawdown as well. With a much higher XIRR, the return of HDFC MM is much higher on a risk-adjusted basis as well.

What’s the SEBI risk category for HDFC Market Money?

Again, this is a data-only approach that I’m taking here, data taken out of context could be misconstrued. SEBI has taken a “what’s in it” approach to these funds - they take a bottom up approach where they look at the fund’s mandate, whether it follows SEBI’s mandate for that particular subcategory, and then what’s actually in the mutual fund? 

Our top down, data-only approach shows that the relative risk profile is different from the SEBI riskometer between these two funds.

So in the image above, we can see that all funds in the money-market category  except one have a lower max DD than the sundaram arb fund. This isn’t an attack on the Sundaram arb fund - they weren’t the only ones who got caught holding DHFL NCDs (they were rated AAA; don't get me started on AAA ratings), but the point I’m trying to make is that the devil is in the details; nobody really considers “Low risk” to mean negative skew until the tail event arrives.

Since we’re seeing an at-most 35% allocation in non-arbitrage strategies as well, I’m curious to understand how much of the overall return is actually attributed to arbitrage, vs debt. Lets explore that next.

Arbitrage Fund: How much is actually Arbitrage?

It’s interesting that this fund is called an arbitrage fund, even though most funds have 65%-70% of their assets involved in arbitrage strategies, and the remaining assets are in different bonds and commercial papers. Most of these funds have generated returns in the range of 6%-8%. I’m curious to understand what proportion of these returns are generated using arbitrage and what proportion of returns are generated using the bonds. 

Here are two scenarios which I think are possible:

Example 1:

A higher chunk of return is actually coming from bonds, so this shouldn’t really be labelled as an arbitrage fund. 

Example 2:

A higher chunk of returns are being generated from actual arbitrage.

It would be interesting to know if we can get this data from anywhere to actually quantify the source of returns.

Let’s look at TATA, which is generating a 6.48% CAGR since inception. Going to its official website and trying to download the current holdings (ignoring the fact that the holdings PDF is dated 30 Sept, 2023, which is ~2 years ago), I see that this is the breakdown of the holdings:

Arbitrage Component

The arbitrage portion includes equity stocks paired with equivalent short positions in futures, which hedge out market risk. The fund shows:

• Hedge Positions: 68.52% long in stocks, matched by -68.97% in futures

• These net to near-zero exposure but represent capital deployed in arbitrage trades

• Total arbitrage exposure: approximately 68.52% of NAV

Debt & Money Market Instruments

• Government securities: 0.65%

• Non-convertible debentures/bonds: 5.51%

• Money market instruments (e.g., CPs): 4.31%

• Mutual fund units (Tata Money Market & Treasury Advantage): 17.80%

• TREPS and Repo: 5.72% + 2.05% = 7.77%

Total debt & related instruments:
= 0.65% + 5.51% + 4.31% + 17.80% + 7.77%
= 36.04% of NAV

So, approximately:

• ~68.5% of the fund is deployed in arbitrage strategies

• ~36% is parked in debt/money market instruments (some of this is likely used for margin funding as well)

So it appears that 68% of the fund is cash-future arbitrage. Lets actually simulate a cash-future trade to see what kind of return we can generate from this trade.

You can open up the AlgoTest simulator and try to mimic an arbitrage strategy by buying shares and selling the future. Let’s take this trade with RELIANCE, which is TATA’s largest holding in the arbitrage book (approx 4.5%), and let’s take this trade on Rollover day, i.e. monthly expiry. We’ll take 24th April, 2025 as the date. A cursory glance difference between the future price and the underlying price shows that the spread ranges from ~Rs. 5 to Rs 8.50 during the day. We’re only looking at LTP at 1-min intervals (this is how the simulator data shows it).

Reliance Cash-and-Futures Arbitrage (35-day expiry)

09:27 AM

• Spot: ₹1,293.0
• Futures: ₹1,301.5
• Spread captured: ₹8.5
• Position: 500 shares → gross P/L ₹4,250 (8.5 × 500)
• Capital tied up: ₹6,46,500 (1 293 × 500)
• Annualised return (before costs): (4,250 ÷ 6,46,500) × (365 ÷ 35) ≈ 6.8%

01:55 PM

• Spot: ₹1,299.4
• Futures: ₹1,304.7

• Spread captured: ₹5.3
• Position: 500 shares → gross P/L ₹2,650 (5.3 × 500)
• Capital tied up: ₹ 6,49,700 (1,299.4 × 500)
• Annualised return (before costs): (2,650 ÷ 6,49,700) × (365 ÷ 35) ≈ 4.2%

Locking in the morning spread delivers a pre-cost annualised yield of about 6.8%, whereas waiting until early afternoon trims it to roughly 4.2%. Figures exclude brokerage, taxes, and other statutory charges.

Expenses (without brokerage)¹ = 455.60/- 

After costs, the returns reduce, i.e. from 4.2%, it reduces to 3.52%, and from 6.8%, it reduces to 6.12%.

Note: 

We’re assuming that the cash required to short the future contract is 0, as you can potentially pledge your debt holdings and the underlying securities to get some sort of margin benefit. But you’ll need some cash margin in order to handle the daily mark-to-market settlement that comes from the futures.

Also there’s costs associated with actually putting these positions on, using some sort of arbitrage capturing hardware/software.

So, we would love the opportunity for an AMC or fund manager of some arbitrage fund to actually explain how the sausage is made here, and how much have we gotten correct here in our analysis.

TL;DR:

Spreads are tight, a large chunk of return is coming debt and the ability to pledge the debt to get lower cash margin requirements. So, if interest rates decline, we should see a decline in the future rate of return offered by these funds because of bond rates reducing as well as the arbitrage opportunity reducing due to the cost of carry reducing. 

(Cost of carry = futures price – spot price. It’s driven by interest rates. Higher the interest rate, higher the cost of carry, higher the potential return offered by these funds.)

Conclusion:

Arbitrage funds could be a  “low-risk, tax-efficient” shortcut, but you’re getting some “negative skew” sprinkled in it as well. Finally, before investing, you should look into:

Historical CAGR and volatility – check the Sharpe ratio, not just headline returns.

Maximum drawdown – even “market-neutral” products can suffer setbacks.

Non-arbitrage allocation – if only 65 % sits in true arb trades, where is the other 35 % parked?

And, it would be great if SEBI could ask the AMCs to post what proportion of the 6-8% XIRR is actually coming from arbitrage vs the bond portfolio?

Post credits:

Arbitrage Mutual Funds keep ~15–30 % in debt to supply margin and meet redemptions. Reaching for a few extra basis points can backfire:

• Credit events (already covered above): Sundaram Arbitrage Fund etc. held AAA-rated DHFL paper that was marked down 75 % when DHFL defaulted in June 2019, slicing ~5 % off the fund’s NAV overnight.
• Liquidity freezes: If spreads go negative and the fund faces large redemptions, managers may dump the debt sleeve at a discount just to raise cash, crystallising losses that equity convergence would otherwise have covered. This happened during covid. Check it out here.

Expenses Breakdown¹

* Clearing charges will only applicable, if buy/sell date of Cash is different (15*1.18 = 17.70).