Supplementary memorandum submitted by
HM Treasury (PAC00-01/107)
THE UK GOLD
Further to the Public Accounts Committee hearing
on the sale of part of the UK gold reserves on 5 February 2001,
please find enclosed the notes requested by the PAC Chairman,
1. Note on the determination of the composition
of the UK's net foreign currency reserves.
2. Note on the supposed gameplaying in the
run up to auctions of part of the UK's gold holdings.
3. Value at Risk: A Worked Example.
Note on the determination of the composition
of the UK's net foreign currency reserves
At the recent PAC hearing on the sale of part
of the UK gold reserves, the PAC Chairman, David Davies, asked
for a note on the determination of the 40/40/20 benchmark and
an explanation of the impact a slowdown in the US economy might
have on the composition of the benchmark. The following addresses
The net foreign currency reserves currently
stand at US$13.7 billion. The main components of the net reserves
are dollars, euros, yen, Gold and SDRs (Special Drawing Rights
make up about 13 per cent of the net reserves). The non gold and
SDR components of the net reserves (dollar, euro and yen) are
held in the proportions 40 per cent, 40 per cent and 20 per cent
The primary reasons for holding reserves are
for the purpose of intervention or to sustain trade in times of
emergency or financial market dislocation. As such the currencies
held need to be liquid, usable for intervention and, for the purposes
of trade, acceptable as payment by the UK's major trading partners.
In broad terms, this dictates that the currencies to be held are
those of the larger economic blocs, ie the US$, euro and yen.
Financial markets have changed significantly over
the last 10-20 years and, as a result, the methods of managing
foreign currency reserves have also changed. In the early 1990s
the key determinant of the weights in the benchmark for the currency
component of the net reserves was the pattern of UK trade. The
rationale being that in times of sustained financial crisis, the
UK may not have access to the capital markets and the foreign
currency reserves would have to be drawn on to sustain trade.
UK imports of goods and services (by value)
by exporting region, for 1992 and 1999, were:
|1992||51 per cent||13 per cent
||5 per cent|
|1999||50 per cent||14 per cent
||4 per cent|
In terms of invoicing currency (excluding sterling invoicing)
the shares in 1999 were:
|EMU country currencies
|1999||38 per cent||50 per cent
||<1 per cent|
The International Monetary Fund uses a similar approach in
determining the weights of currencies that make up the SDR basket.
The weight of each of the currencies broadly reflects the relative
importance of each currency in international trade and reserves.
If sterling is excluded, the weights are 45 per cent dollar, 36
per cent euro and 19 per cent yen; very similar to those used
for the UK net reserves.
Although trade is still one aspect considered when setting
the weights of the currencies in the benchmark, the importance
of portfolio management is in the ascendance. This is because
it is felt that the probability that the financial markets would
be closed to the UK for a sustained period of time has fallen
and new financial management techniques now allow portfolios to
be run more efficiently (minimising risk for a given return).
As stated above, the proportions of the net reserves held
in each currency were originally determined by the share of UK
imports from the US, Europe and Japan. However, the focus now
is more on optimal portfolio theory.
Optimal portfolio theory is based on the trade-off between
risk and return. Investors rarely concentrate their entire wealth
in a single asset, preferring instead to hold a diversified portfolio
of assets. This is because, although investors seek high returns,
they generally seek to minimise the high risks (the volatility
of returns) usually associated with high returns.
Therefore, as well as choosing between individual assets
of differing returns and risks, investors can also form portfolios
containing a number of different assets. The risk and return characteristics
of the overall portfolio will depend not only on those of the
individual assets contained in it, but also on the correlation
between the returns on the different assets. The degree of diversification
benefit that can be achieved depends on whether assets are perfectly
positively correlated (no diversification benefit possible), perfectly
negatively correlated (maximum diversification benefit) or somewhere
in between; and also on the number of assets held in the portfolio
(the benefits of diversification increase as more assets with
uncorrelated returns are included in a portfolio).
The consequence of this is that, within the universe of all
possible portfolios made up of the assets that the manager is
willing/able to hold, there is a set of efficient portfolios.
This set of efficient portfolios dominate all other portfolios
in that they have minimum risk for a given level of return. The
efficient set lies on the frontier A to B in figure 1. And it
is a portfolio on this frontier that portfolio managers seek to
hold. Thus, for the UK net reserves, it is possible to determine
the set of portfolios that yield maximum return for a given level
of risk from within the universe of portfolios made up of the
assets that can be acceptably held, eg the analysis considers
the shares to be held in dollar, euro and yen but doesn't include
assets, such as equities and property, that are not suitable for
the purposes that reserves are held.
The following section summarises the factors that are considered
when the benchmark is set. However, as outlined above, the purposes
for holding reserves determine the assets to be held and the benefits
of diversification (minimising risk for a given return) that are
of prime importance.
The main determinants of the benchmark are:
(a) One of the purposes of holding reserves is to be able
to intervene in the foreign exchange markets. This suggests that
the reserves should be invested in relatively liquid assets, which
in turn points to holding the currencies of the US, Europe and
(b) Of these currencies, the dollar and euro are the most
readily usable for intervention and the most likely currencies
to intervene in, which points to holding larger proportions of
(ii) Macroeconomic factors:
(a) Reserves may be required to finance outflows of trade
and capital account for a sustained period without selling sterling
in the event of a period of financial instability in which sterling
becomes fundamentally undervalued and liquidity in the currency
(b) The foreign currency balance sheet is just one component
of the government's portfolio of financial assets. And if the
government were risk averse, then its optimal strategy would be
to structure its entire balance sheet to provide protection against
shocks to its macroeconomic policy objectives. This strengthens
the case for having a net exposure to the yen, because bad times
for the UK economy would be less likely to be correlated with
weakness in the yen than with weakness in other currencies.
(iii) Risk and return:
(a) Within the parameters set, portfolio optimisation
suggests that (on investment grounds alone) the market risk of
holding net reserves is minimised when the proportions of the
liquid currencies selected are 30 per cent to 50 per cent in dollars,
40 per cent to 70 per cent euros and 0 per cent to 10 per cent
yen. As described above, this is determined by calculating the
shares (based on variances and covariances) of each asset to be
held in the portfolio to minimise risk.
(b) Economic theory suggests that the returns on currencies,
net of short-term interest rate differentials, should be zero.
In other words, taking on higher risk cannot increase returns.
However, analysis of recent data suggests that this has not been
the case and that, of these three currencies, the dollar has been
the highest yielding. Providing that the past repeats itself this
would suggest that the return on holding the net reserves can
be increased by increasing the proportion held in dollars.
Given the number of different factors involved, there is
clearly an element of judgement as to the exact proportions of
each of these currencies that should be held in the benchmark,
but all of the above factors are taken into account when making
The benchmark for the currency holdings of the net reserves
is reviewed periodically to ensure that it is still appropriate
given changes in macroeconomic conditions and currency volatilities
One issue raised at the PAC hearing related to a slowdown
in the US economy and its impact on the net reserves benchmark.
As can be seen from the list of determinants above, the benchmark
is set on the basis of long run factors and it is unlikely that
the benchmark would be adjusted on the basis of a cyclical downturn
in one economy. Changes to the benchmark might come about if,
for example, structural changes in one of the major economies
or in the foreign currency markets led to a currency becoming
more volatile. Alternatively, if a currency from a fast growing
emerging economy became more liquid, then this could lead to an
alteration of the benchmark.
However, as part of normal portfolio management, as currencies
strengthen (weaken) against each other the benchmark portfolio
is adjusted by selling (purchasing) currencies to get back to
40/40/20. If, for example, the US dollar were to weaken against
the euro and the yen then, in the absence of corrective action,
this would lead to less than 40 per cent of the portfolio being
held in dollars and more than 40 per cent and 20 per cent being
held in euros and yen respectively. To correct this, the benchmark
portfolio would be adjusted through the sale of euros and yen,
and the purchase of dollars.
Note on supposed gameplaying in the run up to auctions
of part of the UK's gold holdings
At the PAC hearing into the sale of part of the UK gold reserves,
the PAC Chairman, David Davies, asked for a note on supposed gameplaying
in the run up to UK gold sales. The following note analyses the
price movements in the run up to the ten gold auctions held by
the UK Government to date.
Observations of bond prices in the run up to, and following,
bond auctions shows that bonds tend to cheapen (prices tend to
fall) into auctions and rally out of them. This is entirely consistent
with market makers selling at a small discount to the market price
and in effect this discount can be thought of as an "underwriting
fee" paid to market makers for helping to sell the auction
stock. A second reason why a dip in prices into an auction, and
a rally in prices following an auction, might be observed is if
large market participants are forcing the price down in the run
up to the auction in order to purchase the asset at what subsequently
looks to be cheap prices.
Tables 1 and 2 show the price movements (in US$ and percentage
terms) into each of the ten auctions at 15, 10, 5 and 1 day(s)
before the auction. The average price movements, with and without
the September 1999 auction,
are also shown. Charts 1 to 10 show the price movements around
the ten auctions graphically.
As can be seen from the data there is no consistent pattern
in the price movements in the run up to the auctions. Although,
on average, prices have tended to fall slightly into the auctions,
they have also tended to fall after the auctions. This is not
consistent with "gameplaying" by the market; rather
it is what would be expected during a period in which prices have
tended to trend down over time. Looking behind the averages, we
also see that prices have risen into some auctions but fallen
into the next auction. For example, prices fell into the first
auction but rose into the second. In fact, prices fell before
and rose after, only the sixth, eighth and ninth auctions.
The increases in gold prices following the second, fourth
and sixth auction can be explained as follows. The rise following
the second auction resulted from the announcement of fifteen European
central banks' agreement, which limited the amount of gold to
be sold by the central banks signing up to the agreement to 2,000
tonnes until 2004. The price rose following the fourth auction
as a result of a number of announcements from gold producers which
suggested that the levels of producer hedging would be less than
they had been in the past. And finally, the rise in price following
the sixth auction was a direct result of the weakening in the
US dollar, which boosted non-US demand for gold by lowering the
effective price of gold for investors dealing in currencies other
than the dollar.
In conclusion, the analysis shows that there has been no
consistent pattern in price movements in the run up to or following
auctions and that the overall pattern in price movements is not
consistent with gameplaying.
Table 1: Price movements in US$ terms: (-) indicates a
(ex Sep 99)
Table 2: Price movements in percentage terms: (-) indicates
a price fall
(ex Sep 99)
The average has been shown excluding the September 1999 auction
because of the one off sharp rise in prices resulting from the
announcement of the European central banks' gold agreement that
occurred shortly after the auction took place. Back