Europe Investment Report 2012 — The Number That Was Wrong by a Factor of Three

Between 2010 and 2012 the growth forecasts for euro-area programme and adjustment countries were wrong in a specific and revealing way: they were too optimistic, every year, in every country, by increasing amounts.

A random forecast error goes in both directions. A systematic one in a single direction means a parameter in the model is wrong — and by 2012 the parameter had been identified.

The fiscal multiplier measures how much output changes when government spending or taxation changes. A multiplier of 0.5 means a one-unit spending cut reduces output by half a unit. A multiplier of 1.5 means it reduces output by one and a half units.

The forecasts had used values around 0.5, drawn from empirical work conducted largely in normal economic conditions. The evidence that emerged suggested the actual value in these circumstances was substantially higher — plausibly two to three times higher.

Why the multiplier is larger in these specific conditions:

At the zero bound, monetary policy cannot offset the fiscal contraction. In normal conditions a central bank responds to fiscal tightening by cutting rates, which offsets much of the effect — and the multiplier estimated from those periods measures the net effect after that offset. With rates already at zero, no offset is available, so the full effect appears.

In a downturn, more households and firms are liquidity-constrained. A household that cannot borrow spends its income; a reduction in that income reduces spending one-for-one. In normal times, more agents can smooth through a temporary change.

And when trading partners consolidate simultaneously, the exports that would normally cushion a domestic contraction are also contracting. The standard multiplier estimate assumes an unchanged external environment.

All three conditions applied at once, which is why the error was so large and so consistent.

Why the parameter was wrong

The estimation problem is genuine and worth understanding, because it is not a case of anyone using a careless number.

How multipliers are estimated: by observing historical episodes of fiscal change and measuring what happened to output. This requires isolating the fiscal change from everything else, which is difficult and has generated a large and careful literature.

The problem is that the historical sample is dominated by normal conditions:

• Most fiscal consolidations occur in normal times, when a government decides to reduce debt.
• In those episodes the central bank was able to offset, and typically did.
• And trading partners were usually growing, so exports cushioned the contraction.

So the estimated multiplier is the net effect under conditions where two large offsets were operating.

Applying it to a situation where neither offset exists is a category error, and it is the kind that is easy to make because the parameter looks like a structural constant when it is actually conditional.

The multiplier is not a property of an economy. It is a property of an economy in a set of circumstances — principally whether monetary policy can respond and whether anyone else is doing the same thing. Using a value estimated under one set of circumstances in a completely different set is where the error was.

The empirical work that identified this compared forecast errors against the size of planned consolidation across countries, finding a strong relationship: the larger the planned consolidation, the larger the forecast error. That relationship is precisely what an understated multiplier produces, and it is difficult to explain otherwise.

The methodological lesson generalises well beyond fiscal policy:

When a model produces errors in one direction repeatedly, the diagnosis is a parameter, not noise. And the test is whether error size correlates with the input that parameter multiplies — which is a simple, checkable diagnostic available after two or three observations.

When cutting raises the ratio

The arithmetic consequence is counterintuitive and follows directly from the identity in the Europe Investment Report 2010.

The debt-to-GDP ratio has a numerator and a denominator, and consolidation affects both.

The numerator falls. Less borrowing means slower debt accumulation. This is the intended effect.

The denominator also falls, because consolidation reduces output — by the multiplier.

Which effect dominates depends on the multiplier's size:

• With a low multiplier, output barely falls, the numerator effect dominates, and the ratio improves. This is the assumed case.
• With a high multiplier, output falls substantially. If it falls proportionally more than debt, the ratio rises — the consolidation increases the thing it was intended to reduce.

The additional feedback that worsens it:

• Lower output means lower tax revenue, which partially offsets the spending cut. So the numerator improves by less than the headline consolidation.
• Lower output raises unemployment spending, offsetting further.
• And a weaker economy can raise the interest rate on the sovereign, which adds to the numerator directly.

Fiscal consolidation reduces debt and reduces output. Whether the ratio improves depends on which falls faster, and at a high multiplier the answer can be the wrong one. The policy then requires more of itself to reach the same target.

This does not mean consolidation is never right. A sovereign that cannot borrow has no choice, and unsustainable debt must eventually be addressed. The argument is about pace and timing, not direction:

• Consolidation when the multiplier is low — normal conditions, monetary policy available, trading partners growing — is far less costly.
• The same consolidation when the multiplier is high costs several times as much output for the same fiscal improvement.

The Europe Investment Report 2013 covers how the framework's pace was eventually adjusted.

Everyone at once

The simultaneity effect deserves separate treatment, because it is a coordination failure rather than a policy error by any individual government.

For a single small open economy consolidating alone:

• Domestic demand falls.
• Exports are unaffected, since trading partners are unchanged.
• The currency may weaken, supporting exports further.
• So the multiplier is relatively low, and the adjustment is manageable.

When every trading partner consolidates simultaneously:

• Domestic demand falls in each.
• Export demand also falls, because each economy's customers are contracting.
• The currency cannot adjust between them, since they share one.
• So the multiplier is substantially higher for every participant.

This is a genuine coordination problem:

• Each government's decision is individually reasonable given its own fiscal position and market pressure.
• The aggregate outcome is worse for everyone than a coordinated, more gradual path.
• And no individual government can improve it unilaterally, since consolidating less means facing the market alone.

Every government was doing the right thing for its own position, and the sum of the right things was worse than a coordinated alternative that no one had the authority to organise. That is a coordination failure, not a policy mistake.

The available solution — countries with fiscal space expanding while constrained countries consolidate — requires either a central fiscal authority or voluntary action by the unconstrained. Neither existed, which is the missing fiscal leg from the Global Investment Outlook 2010. The Europe Investment Report 2020 covers the first substantial attempt at it.

Which cuts, which taxes

A dimension frequently lost in the size debate: consolidation of the same magnitude has very different effects depending on its composition.

The general findings from the empirical literature, stated as tendencies rather than precise values:

Public investment cuts have among the highest multipliers. The spending is domestic and immediate, and cutting it also reduces future capacity — so the cost is both short-run output and long-run potential. It is also the easiest cut politically, since no one loses a job or a payment today. The political ease and the economic cost point in opposite directions.

Transfers to lower-income households have high multipliers, because recipients are liquidity-constrained and spend nearly all of what they receive.

Public sector wage cuts have moderate multipliers, and effects that extend into private wage-setting.

Consumption taxes have moderate multipliers and raise measured inflation, which in a deflationary environment has an ambiguous sign.

Taxes on high incomes and wealth have lower multipliers, since the affected households smooth consumption. They raise less revenue per unit of rate, so the trade is real.

Two structural implications:

• A consolidation weighted toward public investment cuts is the most damaging composition for the same headline size, and it is the composition most frequently chosen because it is politically cheapest today.
• The distributional and macroeconomic effects run together. Measures falling on liquidity-constrained households have the largest output cost and the largest distributional impact — so the efficiency and equity arguments point the same way, which is unusual.

The practical point for an investor assessing a consolidation programme is that the composition is disclosed in budget documents, and it predicts the output cost far better than the headline number does.

Internal devaluation, and what it costs

The adjustment mechanism available to a member state that cannot devalue deserves its own treatment, because it is referenced constantly and its mechanics are rarely set out.

The problem: an economy has become uncompetitive — its costs are too high relative to trading partners, so it runs a persistent external deficit.

The normal remedy is currency depreciation. A weaker currency reduces domestic costs in foreign terms immediately, restoring competitiveness in weeks.

Inside a currency union that is unavailable, so the same relative price change must be achieved by lowering domestic wages and prices directly. The economic destination is similar and the path is not, as the UK Investment Report 2008 notes from the other side.

Why it is so much more costly:

It is slow. Wages are set in contracts, negotiated collectively, and resistant to nominal reduction. Achieving a substantial fall takes years, during which the uncompetitiveness persists.

It raises real debt burdens while it happens. This is the crucial difference. A devaluation reduces the value of domestic-currency debt in foreign terms but leaves the domestic burden unchanged. Internal devaluation reduces incomes while debts stay fixed in nominal terms — so borrowers become less able to service unchanged obligations. The adjustment mechanism actively damages the balance sheets it is trying to help.

It works mainly through unemployment. Nominal wages fall reluctantly, so the adjustment comes substantially through job losses rather than pay cuts — which is a far more concentrated and socially costly distribution of the same aggregate adjustment.

And it interacts with the multiplier. Falling wages and prices mean falling nominal output, which raises the debt-to-GDP ratio through the denominator exactly as the consolidation arithmetic above describes. The two policies compound each other's costs.

A devaluation changes one price overnight and nobody negotiates it. An internal devaluation changes millions of prices over years, each one contested, while the debt stays exactly where it was.

What made it partially work anyway: unit labour costs did fall substantially in several adjustment countries, and external balances moved from large deficits to surplus. The competitiveness adjustment was real.

How much came from each source is the contested part. Some came from genuine productivity improvement and wage restraint; a substantial share came from the composition effect of unemployment — when low-productivity jobs disappear, measured average productivity rises without anything improving. The statistic improves and the economy has not.

The Europe Investment Report 2015 covers the recovery that followed and the extent to which the adjustment proved durable.

What an allocator could act on

Treat the fiscal multiplier as conditional, not structural. Its value depends on whether monetary policy can offset and whether trading partners are consolidating too, and estimates from normal periods do not transfer.

Diagnose systematic forecast errors as a wrong parameter. Test whether error size correlates with the input that parameter multiplies — the check is available after two or three observations and it identified this one.

Check whether consolidation can raise the debt ratio. At a high multiplier the denominator falls faster than the numerator, and revenue and unemployment feedbacks make the numerator improve by less than the headline.

Read composition, not just size. Public investment cuts carry the highest output cost and the lowest political cost, which is why they are chosen and why they are the most damaging.

Look for the coordination failure rather than the policy error. Simultaneous consolidation across a currency union raises everyone's multiplier, and no individual government can fix it alone.

Note when efficiency and distribution point the same way. Measures falling on liquidity-constrained households have both the largest output cost and the largest distributional impact, which is a rare alignment worth using.

What 2012 established

• The multiplier is conditional on circumstances, principally monetary offset and simultaneity, and was underestimated by a large factor.
• Systematic one-directional forecast errors identify a parameter problem, detectable early through a simple correlation.
• Consolidation can raise the debt ratio in the short run when the multiplier is high.
• Simultaneous consolidation is a coordination failure, unfixable by any individual government.
• Composition determines the output cost as much as size, with public investment cuts the most damaging and most chosen.