The Reserve Bank is responsible for managing South Africa’s foreign reserves. Among other things, South Africa invests in liquid foreign reserves to provide resources for dealing with external shocks.
As the amount of reserves South Africa holds has risen over time, so has the implied holding cost of our reserves. Since our inflation tends to be higher than in our trading partners, the rand tends to depreciate over time and our interest rates have tended to be higher than in advanced economies. This has meant that the implied cost of carrying FX reserves has been relatively high in South Africa. Rand depreciation over time has also seen an accumulation of profits on South Africa’s foreign reserve holdings accrue to the SARB’s Gold and Foreign Exchange Contingency Reserve Account (GFECRA).
The implied cost of holding foreign reserves will depend on how reserves have been funded and the central bank’s reserve management strategy. Reserves can be funded through currency in circulation, an increase in commercial bank reserves, or selling government assets and buying foreign currency assets. The last option requires sterilisation transactions to limit the impact of such transactions on domestic money supply. The cost of funding reserves through debt issuances depends on whether these are in local or foreign currency, and would come at a premium based on the relative term premium driving the slope of South Africa’s yield curve and the extent of sovereign credit risk.
The returns on reserve assets, on the other hand, are typically lower than these costs. This is because reserves are usually held in liquid and relatively short maturity instruments, and foreign treasuries have yielded lower rates than the South African policy rate or domestic Treasury instruments.
In the chart below, I estimate the annual carrying cost of reserve assets using the differential between the South African policy rate and the US 5-year Treasury rate and the financial year end value of SARB’s gross gold and other foreign reserves. Valuation changes are taken directly from the annual profit or loss on SARB’s GFECRA account. These estimates suggest that since 2005, the median valuation gain (which will include the impact of currency changes) has been around 1% of GDP against a median carry cost of 0.5% of GDP. These estimates suggest that carrying costs are time-varying and likely to be smaller on average (and especially during normal times) than valuation effects. One expects these carry estimates to be slightly lower than for other ‘carry currency’ countries such as Turkey, Brazil or Russia, where interest differentials with major advanced economies are higher and profits and losses on reserves have been larger than in South Africa.
The above chart also gives a sense of how events such as the global financial crisis and the COVID-pandemic (see the valuation losses in 2010/2021) create vulnerabilities for the South African economy to external shocks, and the potential for foreign reserve holdings to act as a counter cyclical buffer. In yesterday’s post, I also suggested that selling reserves to realise the gains that have accumulated in GFECRA would likely be counter-productive if South Africa’s fundamentals, including our fiscal position, do not improve.
Nevertheless, it is hard to know whether spending 0.5% of GDP on reserves as an insurance policy offers value for money or not, especially in our current macroeconomic context. The South African government spends upwards of 6% of GDP on servicing its debt, and this can be expected to rise further. A key policy question remains as to what the appropriate role of SARB’s foreign reserve management policy is in the context of the nation’s deteriorating sovereign financial position. For example, should South Africa accumulate more reserves as the IMF suggests, or pay off its debt more quickly, given how steep South Africa’s yield curve is?
It is important to note that these carrying cost estimates may be overstated compared to an approach based on detailed estimation using component level ZAR-to-foreign currency spreads. Some forms of potential funding, such as notes and coins or IMF Special Drawing Rights, might also be cheaper than assumed implicitly in these calculations. Unfortunately, it is difficult to estimate carry costs accurately given the limited public information around SARB’s seigniorage revenues and the composition of its reserves.
If one instead measures the cost of outright foreign exchange purchases using the policy rate differential with the US, carrying costs would be slightly higher.
The monetary policy implementation framework reform (MPIF) has likely changed the relative cost of using the monetary base to expand reserves and it would be interesting for someone to assess how the actual carry cost of South Africa’s reserves have changes over time.