How to Forecast HDFC Bank Earnings: A Framework

To forecast a large bank’s earnings you build it from the balance sheet up, not from a single guess about profit: project how fast loans and deposits grow, model the margin the bank earns on those assets, layer on fee income, then subtract operating costs, credit costs and tax. Do that with a disciplined set of assumptions and the earnings number falls out the bottom, which is why a forecast is really a framework of choices rather than a prediction.

This piece walks through that framework using HDFC Bank as a worked example. The aim is to teach the method. It is not a view on the stock, a price target, or a specific earnings number to act on.

Start with the balance sheet, not the profit line

For a bank, earnings are downstream of the balance sheet. Most of the income is interest earned on loans and investments, funded by deposits and borrowings. So the first question is not “what will profit be” but “how big does the loan book get, and how is it funded”.

Two lines anchor everything:

• Loan growth. How fast does the advances book expand, and in which segments (retail, corporate, mortgages)? Mix matters because different loans carry different yields and risk.
• Deposit growth. Loans have to be funded. For a deposit-led bank, the pace and cost of deposit gathering caps how fast, and how profitably, it can lend.

For HDFC Bank specifically, deposit mobilisation is the item to watch most closely. After the 2023 merger with parent HDFC Ltd, the bank absorbed a large mortgage book that was historically funded with borrowings rather than deposits. Growing deposits fast enough to fund the enlarged book, while bringing the loan-to-deposit ratio back down, is the central operating story. So in the build, deposit growth is not a footnote assumption. It is a lead driver.

Model the margin: NIM and the CASA rebuild

Net interest margin, or NIM, is the spread between what the bank earns on assets and what it pays for funding, expressed against interest-earning assets. Multiply a NIM assumption by the average asset base and you get net interest income, usually the largest single revenue line.

The hard part is the funding cost side, and here the CASA ratio does a lot of work. CASA is the share of deposits held in current and savings accounts, which are low-cost or effectively free money for the bank. A higher CASA share means cheaper funding and a wider margin.

After the merger, HDFC Bank’s CASA ratio stepped down to roughly the high-30s percent range, from around the low-40s previously. These are approximate, publicly discussed levels, not precise figures. The direction is what matters: a lower CASA share pressures funding costs and, all else equal, weighs on NIM.

So when you model NIM for this bank, you are really modelling two things at once:

1. The CASA rebuild. How quickly does the low-cost deposit share recover toward prior levels? A faster rebuild supports margin. A slow one keeps funding costs elevated.
2. The rate cycle. Bank margins move with the interest rate environment. When policy rates fall, asset yields on floating-rate loans often reprice faster than deposit costs, which can compress margin near-term before deposits reprice too. When rates rise, the reverse can hold. You do not need to predict the central bank. You do need to state clearly what rate path your NIM assumption implies.

A sensible approach is to model a range for NIM rather than a point, and to tie the range explicitly to how optimistic you are on the CASA rebuild and where you think rates go. That honesty about uncertainty is the whole point of the exercise.

Add fee income, then apply the cost ratio

Banks earn more than interest. Fee and other income covers payments, cards, distribution of third-party products, forex and similar services. It tends to grow broadly with the customer base and transaction volumes, so a common approach is to grow it in line with, or a little differently from, the balance sheet, and to sanity-check the ratio of fees to assets against history.

Add net interest income and fee income together and you have operating revenue. From there, operating costs are usually modelled through the cost-to-income ratio: the share of operating revenue eaten by staff, technology, branches and other expenses. A large, established bank often runs an efficiency ratio that is relatively stable, and management commentary frequently flags whether it is investing (ratio drifts up) or harvesting scale (ratio drifts down). Pick a cost-to-income assumption, apply it to revenue, and you get pre-provision operating profit.

Subtract credit costs and tax

Pre-provision profit is not the end. Two more lines stand between it and the bottom.

• Credit costs (provisions). This is the expense set aside for loans that may go bad, usually expressed as a percentage of the loan book (the “credit cost” ratio). It is the single most volatile line in a bank model, because in a stress event provisions can spike and erase a lot of operating profit. HDFC Bank has historically maintained strong asset quality, with low non-performing loans relative to many peers, which is a genuine, well-established feature of the franchise. But history is context, not a guarantee, so a good model still stress-tests the credit cost assumption rather than hard-coding benign numbers forever.
• Tax. Apply the effective tax rate to pre-tax profit to arrive at net profit.

Net profit divided by the share count gives earnings per share. That is the mechanical end of the build.

Put the build in one place

Here is the whole framework as a single build, with the driver that matters most for HDFC Bank flagged on each line. The point of the table is the structure and the ordering, not any particular value.

Step	Line	What you assume	Why it matters here
1	Loan growth	Advances growth and mix	Sets the size of the earning-asset base
2	Deposit growth	Deposit growth and cost	Lead driver: funds the merged book, drives loan-to-deposit ratio
3	NIM	Margin, shaped by CASA and rates	Lead driver: CASA rebuild from high-30s is the swing factor
4	Net interest income	NIM x average assets	Largest revenue line
5	Fee income	Grows with franchise	Diversifies revenue
6	Cost-to-income	Efficiency ratio	Converts revenue to pre-provision profit
7	Credit costs	Provisions as percent of loans	Lead driver: most volatile line; historically low here
8	Tax	Effective tax rate	Gets to net profit

Change the assumptions in steps 2, 3 and 7 and the earnings answer moves the most. That is not an accident. It is telling you where to spend your research time.

Stress the assumptions, do not trust the point

Because the output is only as good as the inputs, the discipline is to run the build across a range, not a single scenario. A few habits help:

• Anchor to disclosure. Start every assumption from what the bank actually reports and what management says on its calls, then adjust deliberately. Do not invent precise numbers.
• Vary the swing drivers together. Model a slower and a faster CASA rebuild, a softer and firmer rate path, and a benign versus stressed credit cost. The spread between those outcomes is your honest uncertainty.
• Sanity-check the ratios. If your assumptions imply a NIM, cost-to-income or credit cost far outside the bank’s own recent history, ask why before you accept it.
• Separate the durable from the cyclical. Asset quality and franchise strength are structural. Near-term margin around a rate cycle is cyclical. Treat them differently.

None of this produces certainty, and it is not meant to. A forecast is a transparent set of assumptions you can defend and revise, not a number to bank on.

The levers that matter most

For a large deposit-led bank like HDFC Bank today, three assumptions do most of the work in any earnings build: deposit growth, the NIM trajectory as the low-cost CASA base rebuilds from the high-30s percent range after the 2023 merger, and credit costs, where the franchise has historically been strong but the line remains the most volatile in the model. Get honest ranges on those three, wire them through the build above, and you have a framework. Treat that framework as a set of assumptions to test, not a prediction to act on.

This article is educational and describes a general method. It is not investment advice, a recommendation, or a forecast of any specific figure. Altys Labs is not a registered research analyst or investment adviser.