A rupee bond index looks like an aggregate index and behaves like three different markets stapled together. Central government securities are liquid, benchmarked, and concentrated in a handful of on-the-run issues. State development loans are issued by more than two dozen states across many maturities, in small individual sizes, and trade thinly once they leave the auction window. Corporate bonds are largely bought to hold to maturity, so secondary volume is a fraction of outstanding.
What do state loans do to a bond tracker?
The consequence for a tracker is structural, and Worked Example 7 quantifies it on a synthetic composite index weighted 52% central government, 28% state loans and 20% corporate. If the tracker holds state loans only from the six largest issuing states, which carry 61% of the outstanding stock, it can source 17.08% of the index directly and must substitute for the other 10.92%. Filling that with central government bonds of matched duration leaves the spread between state loans and central government unhedged, and at a spread duration of 5.40 and a spread volatility of 22 basis points a year that substitution alone contributes 12.972960 basis points of tracking error.
Advertisement
The maturity dimension bites harder than the issuer dimension. Long state loans, twelve years and beyond, are the least liquid segment of the least liquid sleeve, and a tracker that simply drops that bucket and redistributes its weight into what trades ends up 2.412000 years short of the index inside the sleeve, with a key-rate mismatch worth 18.368606 basis points of tracking error. Filling the long bucket with long central government bonds instead brings the key-rate mismatch to zero and raises the substitution exposure to 15.360800% of the index, for a combined substitution term of 17.355108 basis points. Adding an assumed 18 basis points from the corporate sleeve's own sampling, the two designs come out at 28.804572 and 25.003995 basis points respectively.
The second design is better, and the reason is worth generalising. Neither design removes the risk. One carries it as a curve-shape mismatch and the other as a spread basis, and the spread basis is smaller because the state-to-central spread is less volatile than the curve is. A tracker builder's real choice is which residual risk to hold, and the arithmetic above is how you pick.
The same logic reappears wherever a market is segmented rather than deep. Substitution buys curve accuracy and sells basis accuracy, and the exchange rate between them is the ratio of the two volatilities.
How much do state loans cost to trade?
Transaction costs run the same way. At an illustrative 0.75 basis points round-trip on central government paper and 6.50 on off-the-run state loans, with 35% annual index turnover, the state sleeve carries 0.343735 basis points a year of extra drag and the whole index costs 0.606235 basis points a year to track. That is small next to the 25 basis points of tracking error, which is the point: in Indian fixed income the sampling error dominates the trading cost, and a manager who optimises the trading cost while ignoring the cell structure has optimised the smaller number. (All Indian liquidity, spread and turnover figures here are stated illustrative parameters for a synthetic index. Check the current state-loan issuance shares against the RBI's weekly auction results and current traded volumes against CCIL before using any of them; checked 2026-09-08.)