5D Library › Macro-Liquidity Regime Switching, Extreme Tail Dependence and Barbell Asset Allocation · 中文

Chapter 6 Sovereign Buffers and Liquidity Regimes

Reserves as self-insurance

The first five chapters dealt with investment portfolios. This chapter moves the same questions to the level of a country. An emerging economy borrows short-term funds abroad in normal times; when global liquidity tightens and capital suddenly stops flowing in, or flows out, what does it use to pay maturing external debt and imports? Calvo set out the simple economics of such “sudden stops” clearly: when capital inflows stop abruptly, the current account must adjust sharply within a short time, and the adjustment can be extremely costly (Calvo 1998). Foreign-exchange reserves are the buffer kept for such moments: in normal times holding them has an opportunity cost (reserves are usually invested in low-yielding safe assets); in a crisis they buy time.

The question resembles the barbell of Chapter 5: how large should the buffer be, what does it cost, and under what assumptions can one say it is “enough”?

Rules of thumb for adequacy

Three kinds of simple measure are traditional. Import cover: how many months of imports the reserves can pay for, with a common rule-of-thumb threshold of 3 months, suited to economies with fairly closed capital accounts. Reserves to broad money: a gauge of the potential scale of residents switching local-currency deposits into foreign currency (capital flight). Reserves to short-term external debt: the Greenspan–Guidotti rule.

Evidence

In a 1999 speech to a World Bank conference on reserve management, Greenspan described a proposal by Pablo Guidotti, Argentina’s deputy finance minister: countries should be able to live without new foreign borrowing for up to a year, that is, usable foreign-exchange reserves should exceed scheduled amortisations of foreign-currency debt over the following year (Greenspan 1999). Greenspan suggested two additions: the average maturity of external liabilities should exceed a threshold (such as three years); and a “liquidity-at-risk” standard, under which a country could avoid new borrowing for a year with a given ex ante probability (such as 95 %).

The author’s framework asks for the IMF’s ARA metric to be checked. The result is as follows.

Evidence

The IMF’s 2011 policy paper Assessing Reserve Adequacy introduced a risk-weighted metric for emerging markets (the ARA EM metric), with weights calibrated on past emerging-market capital outflows and differentiated by exchange-rate regime (International Monetary Fund 2011). The table gives the current weights as summarised in the IMF’s 2016 guidance note (International Monetary Fund 2016):

Component Fixed rate Floating rate
Short-term external debt (by remaining maturity) 30 % 30 %
Other portfolio liabilities 20 % 15 %
Broad money 10 % 5 %
Exports 10 % 5 %

Reserves of 100 % to 150 % of the metric are suggested. The weight on other portfolio liabilities is not the 2011 value: the 2011 paper proposed 15 % (fixed) and 10 % (floating); the 2013 paper discussed raising it, and according to the 2016 guidance note it was raised by 5 percentage points in the 2015 paper in the light of post-2008 experience. The 2016 guidance note also provides that, for countries with capital-flow management measures on residents, the weight on broad money may be halved (from 10 % and 5 % to 5 % and 2.5 %), with both the unadjusted and adjusted metrics reported; that the metric is not adjusted for commodity exporters, for which a separate buffer against price shocks is reported instead; and that credit-constrained economies have a separate, complementary approach rather than an adjustment to the EM metric.

The program below computes these measures for a fictional emerging economy.

Calculation

# (1) Reserve-adequacy metrics for a FICTIONAL emerging economy (USD bn): Greenspan-
# Guidotti, import cover, reserves/broad money and the IMF ARA EM metric (weights for
# fixed / floating regimes: short-term debt 30/30 %, other portfolio liabilities 20/15 %,
# broad money 10/5 %, exports 10/5 %). (2) Reserves as a buffer stock under a two-regime
# global-liquidity cycle (ILLUSTRATIVE, monthly, 10 years, in units of the ARA metric).
import random, statistics

A = dict(reserves=120, std=70, opl=150, m2=400, exports=200, imports=180)

def ara(a, fixed):
    w = (0.30, 0.20, 0.10, 0.10) if fixed else (0.30, 0.15, 0.05, 0.05)
    return w[0] * a["std"] + w[1] * a["opl"] + w[2] * a["m2"] + w[3] * a["exports"]

def ten_years(k, shocks):
    r, crisis = k, False
    for tight, e, stop in shocks:
        if tight:
            flow = -0.025 + 0.04 * e - (0.15 if stop else 0.0)
        else:
            flow = 0.005 + 0.02 * e
        if not tight and r < k:
            flow += 0.01                                # rebuild towards the target
        r += flow; crisis = crisis or r < 0.4
    return crisis

def draw(rng, months=120):
    tight, out = False, []
    for _ in range(months):
        tight = (rng.random() < 0.90) if tight else (rng.random() < 0.02)
        out.append((tight, rng.gauss(0, 1), tight and rng.random() < 0.05))
    return out

if __name__ == "__main__":
    a = A
    print("fictional economy (USD bn): reserves 120, short-term debt 70, other portfolio")
    print("liabilities 150, broad money 400, exports 200 and imports 180 a year")
    print(f"   Greenspan-Guidotti: reserves / short-term debt ="
          f" {a['reserves'] / a['std']:.2f} (rule: >= 1)")
    print(f"   import cover = {a['reserves'] / (a['imports'] / 12):.1f} months"
          f" (rule of thumb: >= 3)")
    print(f"   reserves / broad money = {a['reserves'] / a['m2']:.0%}")
    for fixed in (True, False):
        m = ara(a, fixed)
        print(f"   ARA EM metric, {'fixed' if fixed else 'floating'} rate: {m:.1f}"
              f" -> reserves = {a['reserves'] / m:.0%} of it (suggested 100-150 %)")
    rng = random.Random(1999); paths = [draw(rng) for _ in range(10_000)]
    tight_share = statistics.fmean(sum(t for t, e, s in p) / len(p) for p in paths)
    print(f"buffer model: share of tight-liquidity months {tight_share:.1%};"
          f" crisis = reserves below 0.4 x metric")
    ks = (0.5, 0.75, 1.0, 1.25, 1.5, 2.0, 2.5)
    pc = {k: sum(ten_years(k, p) for p in paths) / len(paths) for k in ks}
    print(" target   P(crisis in   carry cost   expected 10-year cost (carry + P x loss)")
    print(" x metric  10 years)    10 years      loss = 1      loss = 3      loss = 10")
    best = {}
    for k in ks:
        carry = 0.02 * k * 10
        tot = [carry + pc[k] * L for L in (1, 3, 10)]
        for L, t in zip((1, 3, 10), tot):
            best[L] = min(best.get(L, (9e9, 0)), (t, k))
        cols = "".join(f"{t:7.3f}       " for t in tot).rstrip()
        print(f"  {k:4.2f}     {pc[k]:6.1%}       {carry:5.2f}      " + cols)
    print("   cheapest target on this grid: " + ", ".join(f"loss {L}: {best[L][1]:.2f}"
                                                         for L in (1, 3, 10)))
# fictional economy (USD bn): reserves 120, short-term debt 70, other portfolio
# liabilities 150, broad money 400, exports 200 and imports 180 a year
#    Greenspan-Guidotti: reserves / short-term debt = 1.71 (rule: >= 1)
#    import cover = 8.0 months (rule of thumb: >= 3)
#    reserves / broad money = 30%
#    ARA EM metric, fixed rate: 111.0 -> reserves = 108% of it (suggested 100-150 %)
#    ARA EM metric, floating rate: 73.5 -> reserves = 163% of it (suggested 100-150 %)
# buffer model: share of tight-liquidity months 15.6%; crisis = reserves below 0.4 x metric
#  target   P(crisis in   carry cost   expected 10-year cost (carry + P x loss)
#  x metric  10 years)    10 years      loss = 1      loss = 3      loss = 10
#   0.50      57.7%        0.10        0.677         1.830         5.868
#   0.75      35.9%        0.15        0.509         1.227         3.741
#   1.00      20.9%        0.20        0.409         0.828         2.294
#   1.25      12.1%        0.25        0.371         0.614         1.463
#   1.50       6.7%        0.30        0.367         0.501         0.970
#   2.00       1.8%        0.40        0.418         0.455         0.582
#   2.50       0.4%        0.50        0.504         0.513         0.544
#    cheapest target on this grid: loss 1: 1.50, loss 3: 2.00, loss 10: 2.50

The first half of the output shows that different measures can give different verdicts. This economy’s reserves are 1.71 times its short-term external debt and cover 8 months of imports, both well above the rules of thumb. By the ARA metric, under a fixed exchange rate its reserves are only 108 % of the metric, just inside the 100–150 % range; under a floating rate they are 163 %, above the upper end. The same reserves are “just enough” under a fixed rate and “on the high side” under a floating rate, because a floating rate itself absorbs part of a shock and the metric asks for a smaller buffer. All these measures are rules of thumb, not “right answers” derived from an optimisation problem.

Buffer dynamics: the author’s model

The demand for reserves can be seen as a buffer-stock problem. Frenkel and Jovanovic modelled net international payments as a stochastic process and derived optimal reserves that depend on the opportunity cost of holding them and on the variance of the net flows (Frenkel and Jovanovic 1981); Jeanne and Rancière treated reserves as insurance against sudden stops, with an optimal level that depends on the probability and cost of a sudden stop, risk aversion and the cost of holding reserves (Jeanne and Rancière 2011). In commodity markets, Deaton and Laroque’s competitive storage model shows the consequences of the constraint that “inventories cannot be negative”: when stocks run out, prices spike (Deaton and Laroque 1992). Chapter 7 will show that the ever-normal granary faced the same constraint.

Global liquidity itself has regimes. Miranda-Agrippino and Rey found that US monetary policy drives a global financial cycle that moves global risky asset prices, capital flows and the leverage of financial institutions (Miranda-Agrippino and Rey 2020). This book uses no real-time data and makes no forecast for any country; it uses a synthetic two-regime model only to illustrate buffer dynamics.

The author’s model (illustrative parameters, monthly, ten years): reserves are measured in units of the ARA metric. Global liquidity switches between an “ample” and a “tight” regime; the ample regime persists with probability 0.98 a month (about 50 months on average), the tight regime with 0.90 (10 months). In ample months there is a net inflow of 0.5 % plus noise, and when reserves are below target the central bank accumulates an extra 1 % a month; in tight months there is a net outflow of 2.5 % plus noise, and each month a sudden stop drains 15 % with probability 5 %. Reserves falling below 40 % of the metric count as a “crisis”. Holding reserves costs 2 % a year (in metric units). The second half of the program computes, for different initial targets, the probability of a crisis within ten years, the ten-year holding cost, and the expected total cost when a crisis costs 1, 3 or 10 (in the same units).

In this model about 15.6 % of the ten years’ months are tight. The crisis probability falls as the target rises, but ever more slowly: 57.7 % at a target of 50 % of the metric, 20.9 % at 100 %, 6.7 % at 150 %, 1.8 % at 200 % and 0.4 % at 250 %. Each extra half-unit of reserves adds 0.10 to the holding cost linearly while removing less and less crisis probability: the diminishing marginal protection of a buffer. The target with the lowest expected total cost depends on how expensive a crisis is: 1.5 when a crisis costs 1, 2.0 when it costs 3 and 2.5 when it costs 10. Figure 5 plots these curves.

The author’s reserve-buffer model (illustrative parameters, ten years). Left: probability of a crisis within ten years against the reserve target (in units of the ARA metric); the grey band is the IMF’s suggested 100–150 %. Right: holding cost plus expected crisis loss for three crisis costs; dots mark each curve’s minimum (the figure uses 4,000 simulated paths, so its minima may differ by one grid step from the table’s 10,000 paths).

The model illustrates structure, not a number: “the optimal level of reserves” depends on the frequency and size of sudden stops, the cost of a crisis, the holding cost and the duration of liquidity regimes, all of which are hard to estimate in reality. The IMF’s 100–150 % range is an empirically calibrated reference, not the solution of this optimisation problem.

Note

What the model leaves out. Changes of exchange-rate regime, currency mismatches, central-bank swap lines, IMF lending facilities, capital controls, the currency and liquidity composition of reserves, self-fulfilling crises (falling reserves themselves triggering a run), and the costs of excessive reserves (such as the fiscal cost of sterilised intervention). The crisis threshold of 0.4 is arbitrary.

Hypothesis · not established

The author’s hypothesis (not established): for a given ARA adequacy ratio, the probability of an external crisis in an emerging economy is significantly higher in tight global-liquidity regimes than in ample ones; an adequacy standard that adjusts with the liquidity regime would therefore separate high-risk from low-risk country-years better than the fixed 100–150 % range. What would show it wrong: with the country sample, crisis definition and real-time regime classification fixed in advance, a confidence interval for the interaction of adequacy ratio and regime that includes zero, or an out-of-sample discriminating power (for example, area under the ROC curve) of the regime-adjusted standard no higher than that of the fixed standard.

Chapter summary

Foreign-exchange reserves are self-insurance against sudden stops of capital. The Greenspan–Guidotti rule requires usable reserves to cover external debt falling due within a year; the IMF’s ARA metric applies regime-dependent weights (short-term external debt 30 %, other portfolio liabilities 20 % or 15 %, broad money and exports 10 % or 5 % each) and suggests reserves of 100–150 % of the metric. The same fictional economy scores 108 % under a fixed rate and 163 % under a floating one, and different measures can give different verdicts. The author’s two-regime buffer model shows that more reserves reduce the crisis probability ever more slowly while the holding cost rises linearly; the “optimal” level depends on the cost of a crisis and the nature of liquidity regimes, and rules of thumb are only references.

References

Calvo, Guillermo A. 1998. “Capital Flows and Capital-Market Crises: The Simple Economics of Sudden Stops.” Journal of Applied Economics 1 (1): 35–54.
Deaton, Angus, and Guy Laroque. 1992. “On the Behaviour of Commodity Prices.” The Review of Economic Studies 59 (1): 1–23. https://doi.org/10.2307/2297923.
Frenkel, Jacob A., and Boyan Jovanovic. 1981. “Optimal International Reserves: A Stochastic Framework.” The Economic Journal 91 (362): 507–14. https://doi.org/10.2307/2232599.
Greenspan, Alan. 1999. Currency Reserves and Debt.
International Monetary Fund. 2011. Assessing Reserve Adequacy. IMF Policy Paper. International Monetary Fund.
International Monetary Fund. 2016. Guidance Note on the Assessment of Reserve Adequacy and Related Considerations. IMF Policy Paper. International Monetary Fund.
Jeanne, Olivier, and Romain Rancière. 2011. “The Optimal Level of International Reserves for Emerging Market Countries: A New Formula and Some Applications.” The Economic Journal 121 (555): 905–30.
Miranda-Agrippino, Silvia, and Hélène Rey. 2020. “U.S. Monetary Policy and the Global Financial Cycle.” The Review of Economic Studies 87 (6): 2754–76. https://doi.org/10.1093/restud/rdaa019.

Fan, Y. (2026). Macro-Liquidity Regime Switching, Extreme Tail Dependence and Barbell Asset Allocation: Vine Copulas, Hidden Markov Models and Sovereign Buffer Dynamics [宏观流动性体制转换、极值尾部相依与杠铃资产配置] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23828049 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).