Time to reform the Ghana Reference Rate: From Treasury Bill Yields to Banks’ Deposit Costs

Authors: Williams Kwasi PEPRAH, PhD, Professor of Finance, Andrews University, email: peprah@andrews.edu (L) and Fred Bortey Kyeremeh ACCA, PhD(c), Adventist University of the Philippine, Email: Fbkyeremeh@gmail.com (R) For businesses borrowing to buy equipment, restock a shop or meet payroll, the price of credit affects everyday decisions. The Ghana Reference Rate (GRR) provides a common starting point for that price. Its usefulness depends on whether the formula reflects financing conditions and whether borrowers can understand how it becomes the interest rate they pay. The central weakness is specific. The GRR blends a policy signal, interbank borrowing and government investment returns without directly measuring banks’ deposit costs. Its reserve adjustment also gives Treasury bill movements more influence than the stated 40% weight initially suggests. The evidence supports a review of the benchmark’s weights and a test of a simple formula that directly includes banks’ deposit interest costs. The main questions for this article are: Whose Cost Is Ghana’s Reference Rate Measuring? Why Should Bank Deposit Costs Replace Government Borrowing Rates in the GRR Formula? Here, T₉₁ is the annualised 91-day Treasury bill rate. CRR is the cash reserve ratio, the share of eligible deposits banks must hold as reserves. CIV is the cash-in-vault ratio used in the adjustment. The denominator represents the share remaining after these amounts are deducted. The three stated weights sum to 100%. [3, 4] At a 20% CRR and an illustrative 2% CIV, the denominator is 0.78. A 5.38% Treasury bill rate therefore enters the formula as 6.90%, before the 40% weight is applied. The adjustment recognises a cost from funds held in cash and reserves, but its design and size deserve scrutiny. [2, 5] The Treasury bill component has more leverage than it appears With CRR at 20% and CIV at 2%, a one percentage-point rise in the raw Treasury bill rate increases the formula-based GRR by 0.513 percentage points. The same rise in the policy rate increases it by 0.400 points, and an interbank increase adds 0.200 points. These comparisons hold the other inputs constant. Sources: Bank of Ghana interest-rate series and reserve information [2, 5]; 40/20/40 formula [3]. Sensitivities use CRR 20% and CIV 2%. One basis point is 0.01 percentage points. The bars show separate responses, not shares that must total 100%. For example, a Treasury bill increase from 5% to 10%, with the other inputs unchanged, would add approximately 2.56 percentage points to the GRR under these assumptions. A business with a floating-rate loan linked to the benchmark could face that increase even without a change in the policy rate or its own creditworthiness, depending on its contract. The formula’s 40% Treasury bill weight is applied to an adjusted rate. Relative to the unadjusted rate that most readers see in the news, the sensitivity is about 51.3%, compared with 40% for the policy rate. This is not an arithmetic error. It is a feature that should be disclosed prominently whenever the benchmark’s weights are explained. The reserve setting changes that sensitivity. At a 10% CRR and 2% CIV, the Treasury bill coefficient is approximately 0.455. At 20% CRR and 2% CIV, it becomes 0.513. Thus, raising reserve requirements changes the GRR’s response to future Treasury bill movements as well as increasing its current level. The required improvement is straightforward: every monthly GRR release should show both the stated weights and the effective sensitivities to the three headline rates. Borrowers should not have to infer this distinction from the formula. Treasury bills dominate the historical movement An analysis of 102 monthly observations from April 2018 to September 2026 shows how significant this design becomes over time. The Treasury bill component, including its reserve adjustment, accounts for 78.8% of the variance in monthly changes in the formula-based GRR. The policy component accounts for 12.6%, and the interbank component for 8.6%. [1, 2] The 78.8% figure concerns variation in the formula-based benchmark, not the percentage of every loan rate attributable to Treasury bills. Nevertheless, it shows that a component with a stated 40% weight has dominated the benchmark’s historical movement. A Treasury bill is an investment available to banks. Its yield is therefore relevant when a bank compares lending with holding government paper. However, an investment return is different from the cost of obtaining deposits. Treasury bill yields can change because of government financing needs, auction supply and investor demand. The formula can transmit those conditions into the starting price of private credit. Using Treasury bill yields in the GRR passes the government’s borrowing pressures on to businesses and households through bank lending rates, rather than directly reflecting banks’ own funding costs (Deposit cost). The policy rate and interbank rate also reflect related monetary conditions. Treating their 40% and 20% weights as independent economic information overstates the diversity of the formula’s inputs. Their relationship does not make the blend invalid, but the allocation needs evidence. BoG and the Ghana Association of Banks should specify whether the GRR measures funding costs, the opportunity cost of lending or monetary policy conditions. If it is primarily a funding benchmark, the present 40% Treasury bill allocation needs to be tested against a measure that includes actual deposit funding costs. A common benchmark does not tell the borrower’s whole story Published rates show that the GRR has a substantial relationship with bank lending. Across 92 matched monthly observations, the statistical fit between the GRR and average lending-rate levels is approximately R² = 0.95. The difficulty is therefore not an absence of connection. It is that the benchmark cannot explain the entire price of a loan or establish whether that price is fair. [1, 2] The average lending-rate gap above the GRR was 7.34 percentage points in 2019, compared with 1.37 points across the available observations in 2024. By August 2026, the GRR was 10.61% and the average lending rate was 15.94%, leaving a gap of 5.33 points. A changing gap matters to borrowers because a lower benchmark does not guarantee an equally large reduction in the final loan rate. [1, 2] These aggregate gaps are not measures of bank profit or default risk alone. They can reflect funding differences, operating costs, capital charges, risk, loan composition and margins. Fees can raise the annual percentage rate further. BoG’s APR framework distinguishes these additional charges from the reference rate itself. [4] The underlying design problem remains; the GRR contains no direct deposit-cost input, although banks can fund loans with current accounts, savings, term deposits and wholesale borrowing. It also combines overnight and 91-day rates while many business loans run for years. A funding benchmark should acknowledge these differences, with product-specific credit and maturity costs disclosed separately. How the MPR and GRR can move apart The MPR signals monetary policy; the GRR blends that signal with market rates. The MPR is neither a required floor nor a ceiling for the GRR. Comparing 102 published GRRs with their prior-month MPR inputs gives 44 months below, 58 above and none exactly equal. Using prior-month inputs aligns the comparison with the formula’s timing. [1, 2] For the formula-based GRR, let A represent the reserve-adjusted Treasury bill rate. Subtracting the MPR from both sides gives: When the MPR is greater than the GRR In September 2026, MPR was 14.00% and GRR was 10.18%, a 3.82-point gap. Lower interbank and adjusted Treasury bill inputs pulled the formula below the policy signal. This can weaken the MPR’s influence on the benchmark when other rates remain low. It does not guarantee cheap loans: banks still add contractual margins and charges. When the GRR is greater than the MPR In February 2023, GRR was 33.25% against MPR of 28.00%, a 5.25-point gap. The adjusted Treasury bill input was approximately 41.42%, while interbank was 25.24%. The higher Treasury bill component outweighed the lower interbank component. Government financing conditions can therefore lift the lending benchmark above the policy rate, adding pressure to GRR-linked borrowing costs. When the two rates are equal Exact equality would mean the weighted differences cancel, not that the model is accurate. Hypothetical inputs of MPR 14%, IBR 10% and adjusted Treasury bills 16% produce GRR of 14%. August 2021 was close: GRR was 13.51% and MPR 13.50%. Equality should not be forced through a floor, cap or target. The mismatch also affects changes in the rates. Holding interbank and reserves constant, a one-point MPR cut lowers GRR by 0.40 points, but a simultaneous one-point Treasury bill rise adds 0.513 points at CRR 20% and CIV 2%. The net result is a 0.113-point GRR increase despite the policy cut. The critique is the formula’s mixed signals and indirect funding proxy. The deposit-cost proposal below addresses that proxy; it would not guarantee equality with the MPR. The monthly calculation should be reproducible Using August 2026 month-end inputs of a 14.00% policy rate, 10.20% interbank rate and 5.38% Treasury bill rate, together with CRR of 20% and CIV of 2%, the formula produces 10.40% for September. The published September GRR was 10.18%. The difference is approximately 0.22 percentage points. [1, 2, 5] Under this calculation, the reserve adjustment alone adds about 0.61 percentage points, lifting the unadjusted blend from 9.79% to 10.40%. The question is whether increasing the return on government paper is the best way to represent the cost of deposits held in reserves. An alternative is to calculate that allowance directly from the cost of the funding subject to the reserve requirement. Across the 102-month period, formula-based estimates fall within 0.15 percentage points of published GRRs in 68 months. In 11 months, the difference exceeds 0.50 points. Different observation dates, reserve ratios or conventions may explain these gaps. The comparison cannot identify the cause without the precise inputs used for each announcement. Under this calculation, the reserve adjustment alone adds about 0.61 percentage points, lifting the unadjusted blend from 9.79% to 10.40%. The question is whether increasing the return on government paper is the best way to represent the cost of deposits held in reserves. An alternative is to calculate that allowance directly from the cost of the funding subject to the reserve requirement. Across the 102-month period, formula-based estimates fall within 0.15 percentage points of published GRRs in 68 months. In 11 months, the difference exceeds 0.50 points. Different observation dates, reserve ratios or conventions may explain these gaps. The comparison cannot identify the cause without the precise inputs used for each announcement. Historical reserve treatment also matters. During the earlier tiered regime, banks faced reserve ratios of 15%, 20% or 25%. A common benchmark should state which ratio was applied and why. BoG subsequently moved to a uniform 20% CRR on 4 June 2026. Any industry averaging and the vault-cash allowance should be explicit. [5, 6] Publication of the calculation-date inputs would settle these questions more effectively than asking the public to approximate the announced rate from separate statistical releases. The 40% policy-rate weight preserves the monetary policy signal. The 20% interbank weight preserves a traded bank funding indicator. The remaining 40% introduces the direct deposit interest cost. Keeping the weights initially unchanged makes the effect of replacing the Treasury bill input easier to assess. How to measure the deposit cost BoG should derive D from total interest accrued on eligible cedi deposits during the month, divided by their average balances, and annualise the result. Pooling the amounts across banks gives an industry rate weighted by deposit balances. Include eligible non-interest-bearing deposits at a zero interest rate; using only advertised fixed-deposit rates would overstate the direct cost of the whole deposit base. Consider an illustrative GH¢100 deposit base: GH¢50 pays no interest, GH¢30 pays 10% annually and GH¢20 pays 15%. Annual interest cost is GH¢3 plus GH¢3, or GH¢6. The direct deposit cost D is therefore 6%, rather than either quoted rate of 10% or 15%. If CRR is 20% and the applicable CIV allowance is 2%, GH¢78 remains after the adjustment. Dividing the GH¢6 annual interest cost by GH¢78 gives an adjusted deposit cost of 7.69%. Its 40% contribution to the proposed GRR is 3.08 percentage points. The deposit balances and reserve ratios must cover the same eligible funding. The reserve and vault-cash treatment should be confirmed and published. Direct deposit cost measures interest expense; operating costs, borrower risk and other loan charges remain separate. This is still a hybrid reference rate, not a complete measure of the cost of lending. How the change would affect the GRR The same August 2026 market inputs used above allow a direct comparison of the two formulas. Hold MPR at 14.00%, IBR at 10.20%, CRR at 20% and CIV at 2%. Only the Treasury bill input is replaced with the illustrative deposit costs below. The existing formula estimate remains 10.40%; it is distinct from the published September GRR of 10.18%. With a hypothetical direct deposit cost of 6%, the proposed formula adds 5.60 percentage points from the policy rate, 2.04 points from interbank funding and 3.08 points from adjusted deposits. The result is 10.72%, approximately 0.32 percentage points above the existing formula’s estimate. Market input sources: Bank of Ghana [2]. All deposit costs shown are hypothetical, not observed industry estimates. MPR 14%, IBR 10.20%, CRR 20% and CIV 2% are held constant. The dashed line is the existing formula estimate using a 5.38% Treasury bill rate, not the published September GRR. If the direct deposit cost were 4%, the proposed GRR would be 9.69%, approximately 0.71 percentage points lower than the existing formula estimate. At 8%, it would be 11.74%, approximately 1.34 points higher. At D equal to the Treasury bill rate of 5.38%, the substitution would leave the formula result unchanged. Three effects follow. First, the benchmark would respond directly to deposit interest costs: a one percentage-point change in D would move the proposed GRR by about 0.513 points at these reserve settings. Second, Treasury bill movements would lose their direct mechanical effect on the formula, although they could still affect deposit pricing indirectly. Third, deposit costs would have to fall before this input could support a lower benchmark. The purpose is a more representative starting point for loan pricing. A lower GRR is a possible outcome, not a guaranteed one. Historical improvements in accuracy or stability cannot be established without a verified deposit-cost series, and borrowers benefit only to the extent that their contractual loan rates reflect the benchmark change. Three priorities for reform 1 Test the deposit cost formula BoG and the Ghana Association of Banks should run the proposed formula alongside the existing GRR for at least three monthly releases. Banks should submit deposit interest accruals and average eligible balances under one reporting definition. Check the submissions, include zero-interest deposits correctly, and publish the comparison before deciding whether to replace the Treasury bill input. Retain 40/20/40 initially to isolate the substitution, then review the weights against measured funding conditions. 2 Publish a calculation that everyone can reproduce Each GRR announcement should show the policy rate, interbank rate, Reserve-adjusted 91-day Treasury bill rate (if replaced by deposit cost as proposed), reserve ratio and CIV allowance, with their observation dates and weighted contributions. Publish the precise deposit-cost definition and explain how the reserve adjustment applies. The inputs should add up to the announced rate, and banks should not recover the same reserve cost again through an unexplained loan margin. Transparent inputs and periodic review are consistent with IOSCO benchmark principles. [7] 3 Show borrowers how the benchmark changes their loan Require each floating-rate loan statement to show the GRR, the contractual margin, fees, the annual percentage rate and the next reset date. BoG should check that rate adjustments follow the contract and that any permitted margin change has a documented basis. Fixed-rate contracts should retain their agreed treatment. This connects benchmark reform to the interest and payments borrowers actually face. [4] A deposit-cost input would give Ghana’s lending benchmark a clearer economic meaning. Its value would come from measuring an actual funding cost, explaining the calculation and making contractual loan pricing easier to verify. Sources and measurement notes Historical comparisons cover April 2018–September 2026. Lending comparisons use 92 matched observations without filling missing values. Formula-based estimates use prior month-end rates, a 2% CIV assumption, the lowest 15% tier during the tiered reserve period and approximate dates for some historical reserve changes. The estimates are distinct from official GRR calculations. Variance shares describe the formula-based series; R² describes statistical association. Exact historical replication requires calculation-date inputs and confirmed reserve conventions. The proposed deposit-cost formula retains the 40/20/40 weights as a transitional design for testing. The deposit-cost scenarios are hypothetical. A verified industry deposit-cost series has not been used to backtest the proposal, so the historical variance shares and lending-rate fit describe the earlier GRR analysis, not demonstrated performance of the new formula. References [1] Ghana Association of Banks. (n.d.). Ghana reference rate historical data, 2018–2026. Retrieved October 23, 2026. Read source [2] Bank of Ghana. (2018–2026). Summary of economic and financial data. Releases from July 2018 to September 2026; monetary policy, interbank, 91-day Treasury bill and average lending-rate series. Read source [3] The Ghana reference rate: A new era? (2018, May 3). Modern Ghana. Historical explanation of the 40/20/40 weights and reserve adjustment. Read source [4] Bank of Ghana. (2019). Annual percentage rates of banks as at September 30, 2019. Explains the GRR components, reserve adjustment and distinction from APR. Read source [5] Bank of Ghana. (2026). Bi-monthly publication of questions from the media July 2026. Confirms the transition to a uniform 20% CRR on June 4, 2026. Read source [6] Bank of Ghana. (2024, March). Monetary policy committee press release. Describes the tiered cash reserve requirements. Read source [7] International Organization of Securities Commissions. (2013, July). Principles for financial benchmarks. Provides standards for methodology transparency, data sufficiency, periodic review and accountability. Read source The recommendations above apply the principles of transparent inputs, a defined economic purpose and periodic methodological review to the GRR. [7]