Global Investment Outlook 2010 — The Divergent Recovery

2010 opened on a reasonable assumption: a common shock had been met with a common policy response, so the recovery would be broadly common too.

It was not. By the end of the year three distinct trajectories were visible, and the differences were structural rather than a matter of timing.

The United States recovered fastest among developed economies, having recapitalised its banks early and forcefully, and having a financial system that transmits policy through capital markets rather than principally through banks.

Europe did not, for reasons the Europe Investment Report 2010 covers in detail: bank repair was slower and less complete, transmission runs through those same impaired banks, and a sovereign crisis began that made the repair harder still.

Emerging economies recovered fastest of all — and then faced a problem they had not asked for, as capital displaced by zero rates in the developed world arrived in size.

The organising insight is that recovery from a balance sheet recession is a domestic process. Policy can support it, but the work — households reducing debt, banks rebuilding capital, firms repairing finances — happens inside each economy at a pace set by how much damage there was and how quickly losses were recognised. A common shock does not produce a common recovery, because the pre-existing conditions were not common.

The year's most consequential development was structural. A sovereign debt crisis began in the euro area that exposed something long known to specialists and never priced: a currency union without a banking union or meaningful fiscal risk-sharing has a specific failure mode, and 2010 was its first full demonstration.

Why recoveries diverge

The divergence was not a policy failure. It follows from what a balance sheet recession is.

An ordinary recession is a fall in demand. Policy lowers rates, borrowing becomes attractive, demand returns. The mechanism is fast because it works through willingness to borrow.

A balance sheet recession is different. Households, firms and banks emerge from a credit boom holding more debt than they want against assets worth less than they paid. Their priority is not to borrow but to repair — to pay down debt and rebuild capital.

Lower rates do not fix this, and may not even help much:

• A household repairing its balance sheet does not borrow at any rate. Cheaper credit is irrelevant to someone whose objective is less debt.
• A bank rebuilding capital lends less, not more. Cheaper funding improves its margin and therefore its capital accumulation, which is useful, but it does not increase its appetite for new risk.
• The repair takes as long as it takes. It is arithmetic — a stock of excess debt divided by the rate at which income can be diverted to reducing it.

So the recovery speed is determined by three things, all domestic and all pre-existing:

1. How much repair is needed — the size of the debt overhang entering the crisis.
2. How quickly losses are recognised. An economy that forces recognition early repairs faster than one that allows losses to sit unacknowledged, because unrecognised losses freeze the institutions carrying them.
3. Whether the financial system transmits policy — which, per the Europe Investment Report 2015, differs fundamentally between bank-based and market-based systems.

A shared shock and a shared policy response do not produce a shared recovery. The recovery happens inside each balance sheet, and those were never shared.

The forecasting implication is durable. Through 2010 and for years afterward, growth forecasts for the slower economies were repeatedly revised down, because the models assumed policy transmission that a repairing system does not provide. The error was not in the policy but in the expectation of how fast it could work.

The doom loop opens

The euro area's sovereign crisis, beginning in 2010, is best understood through a single mechanism.

The loop:

1. A government's fiscal position deteriorates, so its bonds fall in value.
2. Domestic banks hold large quantities of their own government's bonds — for regulatory, liquidity and relationship reasons, this is normal everywhere.
3. Those holdings fall, damaging bank capital.
4. Weaker banks lend less, weakening the economy, worsening the fiscal position.
5. And weaker banks may need government support — which the government must fund by borrowing, worsening its position further.
6. Return to step 1.

Each link is individually reasonable and the loop is self-reinforcing. No participant is behaving irrationally.

Two features make it particularly severe in a currency union:

• The government cannot print the currency its debt is denominated in. A sovereign borrowing in its own currency has a central bank that can always meet nominal obligations. A euro-area member does not. That converts a liquidity problem into a solvency question, which is a categorically different market.
• Deposits are guaranteed nationally. Without common deposit insurance, the safety of a deposit depends on the fiscal capacity of the state where the bank sits — so a weak sovereign makes its banks' deposits less credibly guaranteed, inviting exactly the funding withdrawal the Global Investment Outlook 2008 describes.

The investment consequence, which the Europe Investment Report 2016 develops: euro-area bank exposure and euro-area sovereign exposure are not independent positions. A portfolio holding both in the same member state is far more concentrated than the line items suggest, and the concentration is largest where both assets look cheapest.

Half a union

The crisis exposed an incompleteness that had been discussed by economists since before the currency launched and priced by markets not at all.

What a durable monetary union requires, on the standard analysis:

• A single monetary authority. Present from the start.
• A banking union — common supervision, common resolution, and common deposit insurance — so bank health is not tied to sovereign health. Absent entirely in 2010.
• Fiscal risk-sharing sufficient to absorb a shock that hits one member and not others. Largely absent.
• Labour mobility sufficient to adjust to asymmetric shocks. Present in law, limited by language and qualification recognition.

Why the gaps did not matter for a decade. Before 2008, spreads between member sovereigns had compressed almost to nothing. Markets priced the debt of very different economies as near-equivalent, which meant the missing architecture was never tested. The compression was itself the mispricing — it implied a mutualisation that did not legally exist.

2010 was the repricing of that assumption, and it is the same mechanism the archive documents repeatedly: an assumption held so long it stops being modelled, then tested.

What the member states faced, absent the missing architecture:

• No devaluation. The traditional adjustment for an uncompetitive economy is a weaker currency. Inside a union that is unavailable.
• Internal devaluation instead — falling wages and prices to restore competitiveness. This is slow, socially costly, and raises the real burden of existing debt while it happens.
• No lender of last resort for the sovereign, initially, which is what made the solvency question live.

The eventual resolutions — support mechanisms, banking supervision, and ultimately the commitment described in the Global Investment Outlook 2012 — were the union building the missing parts under pressure. The Europe Investment Report 2020 describes the fiscal leg finally being approached a decade later.

Quantitative easing has a foreign policy

An underappreciated feature of 2010 is that a domestic policy in one economy became a problem for others, through a channel nobody had to intend.

The mechanism. Zero rates and asset purchases in a large developed economy push its investors toward higher-yielding assets, per the portfolio balance channel the Global Investment Outlook 2009 describes. Some of those assets are abroad, in economies growing faster and offering higher rates.

What the recipient economies experienced:

• Currency appreciation, damaging export competitiveness.
• Asset price inflation in property and equities, driven by inflows rather than domestic conditions.
• Credit growth as cheap foreign capital funded domestic lending.
• A policy bind. Raising rates to control the resulting inflation attracts more capital, worsening the appreciation. Cutting rates to deter inflows fuels the credit growth. Neither instrument works, because the problem is not domestic.

This is the policy trilemma the Asia-Pacific Investment Report 2016 sets out, encountered from the receiving end: with free capital movement, an economy cannot simultaneously control its exchange rate and run monetary policy for domestic conditions. The inflows forced the choice.

The responses available were all imperfect: accept the appreciation, accumulate reserves to resist it at a cost, or restrict the flows through capital controls — which moved from heterodox to broadly accepted over this period, a genuine shift in economic orthodoxy.

The durable lesson is about spillover. A monetary policy set for domestic conditions in a large economy is a shock to smaller ones with no vote in it. The dollar-debt burden the Global Investment Outlook 2015 describes is the same channel running in the opposite direction, five years later, when the flows reversed.

When the market structure is the risk

A brief episode in May 2010 demonstrated something about modern markets that had no precedent, and it belongs in this archive because the mechanism recurs.

What happened structurally. A large automated sell order met a market where most liquidity was provided by automated systems. Those systems, detecting abnormal conditions, withdrew. With liquidity gone, prices moved violently within minutes before recovering almost as fast.

The mechanism, stated generally:

• Displayed liquidity is not committed liquidity. A quote can be withdrawn instantly, and the parties displaying it have no obligation to trade.
• Automated liquidity provision is conditional by design. A system that detects conditions outside its parameters stops quoting — which is prudent individually and removes the market collectively.
• The withdrawal is correlated, because the systems observe similar signals and respond similarly.
• It is faster than any human response. The episode was substantially over before anyone could intervene.

Liquidity that disappears when it is needed is not liquidity. It is a service available in conditions where you do not require it.

This is the same finding as the Global Investment Outlook 2008's counterparty freeze and the 2020 March liquidity event, arriving through a completely different route. In all three, the liquidity was conditional and the condition was stress. That an automated microstructure could produce it in minutes rather than weeks was the new information.

The practical consequence is that execution risk became a genuine portfolio consideration rather than an operational detail, and that any liquidity assumption should be stated as a claim about conditions.

What an allocator could act on

Model recovery speed from balance sheet damage, not from policy support. The repair is domestic and arithmetic — a debt overhang divided by the rate of repayment. Forecasts built on policy transmission were repeatedly wrong for years, in the same direction.

Check whether sovereign and bank exposures are independent. In a currency union without common deposit insurance they are not. Holding both in one member state is a single levered position on that state's fiscal capacity, and the ECB publishes bank holdings of domestic sovereign debt free.

Price the absence of a devaluation option. An uncompetitive economy inside a currency union must adjust through wages and prices, which is slow and raises the real value of its debt while it happens. That path is materially worse for asset holders than a currency adjustment.

Watch where displaced capital goes. Easing in a large economy pushes capital abroad, creating appreciation and credit growth in recipients that their own policy cannot address. The flows are traceable and the bind is predictable.

Treat displayed liquidity as conditional. Quotes can be withdrawn instantly and the withdrawal is correlated across providers. Any position whose exit assumes normal-market depth has an unstated condition attached.

Note that an assumption compressed to zero is still an assumption. Euro-area sovereign spreads had converged almost completely before 2008, implying a mutualisation that did not legally exist. A price of zero for a risk is a claim, not an absence of the risk — and it is the most expensive kind to hold unexamined.

What 2010 established

• Balance sheet repair is domestic, so a shared shock and a shared response still produce divergent recoveries.
• The sovereign-bank doom loop was demonstrated, tying two exposures that portfolios treated as independent.
• A monetary union without a banking or fiscal union was shown to have a specific failure mode, previously theoretical.
• Monetary policy exports itself through capital flows, creating binds in recipient economies with no domestic instrument to address them.
• Market structure became a risk factor, with automated liquidity shown to be conditional, correlated and faster than any human response.