China Stimulus Report 2009 — The Floor Under the World

By early 2009 global trade was contracting at a rate without modern precedent and the Western policy response was primarily monetary — rates at the lower bound, balance sheets expanding, described in the Global Investment Outlook 2009. What was missing was demand.

China supplied it. The programme announced in late 2008 and executed through 2009 was, in its effect on global output, the single largest discretionary demand injection of the crisis. And it was structurally different from anything attempted in the West, in a way that matters for understanding both its success and its cost.

Western stimulus was fiscal: governments borrowed and spent, and the cost appeared as a deficit and a rising public debt ratio. China's was executed principally through the banking system. Banks were directed to lend; state-owned enterprises and local governments were directed to invest. The result was a very large increase in credit that did not show up as a central government deficit, because the borrowing was done by entities that were not the central government.

This is the report's central point. Judging the programme by the headline fiscal figure understates its scale severely, and — more importantly — misidentifies where the liability sits. The obligation was distributed across local government financing vehicles, state-owned enterprises and the banks that lent to them. A liability that does not appear in the statistic being monitored is not a smaller liability. It is a less observable one — the same lesson the US Housing & Mortgage Report 2008 draws from off-balance-sheet vehicles, arriving through an entirely different route.

Two consequences shaped the following decade.

Globally, it put a floor under demand. Commodity exporters recovered far faster than the developed economies whose banking systems had failed. Australia, Brazil, and the resource complex generally experienced the crisis as a sharp interruption rather than a lost decade — and the reason was a single buyer's investment rate.

Domestically, it built a debt structure that outlived its purpose. Infrastructure and property assets financed with credit must generate returns sufficient to service that credit. Where they do not, the debt remains after the demand impulse has passed.

Why the headline number understates it

Comparing China's programme to Western packages using the announced figures produces a misleading result, and the reason is mechanical.

What the announced figure captured: a headline investment programme over roughly two years, concentrated in transport infrastructure, post-earthquake reconstruction, housing and rural development.

What it did not capture, and this is the larger part:

• Directed bank lending. New credit extended at the authorities' direction, which is not a fiscal transaction and appears in no deficit.
• Local government investment, financed by borrowing through vehicles created for the purpose.
• State-owned enterprise capital expenditure, undertaken at policy direction rather than on commercial assessment.

The distinction between fiscal and credit stimulus is not accounting pedantry, because the two behave differently:

• A fiscal transfer is spent and gone. It raises public debt and stops.
• Credit-financed investment creates an asset and a matching liability. Whether it was worth doing depends on whether the asset earns enough to service the debt — a question that cannot be answered at the time and is answered slowly.

Fiscal stimulus is measured by what it costs. Credit stimulus is measured by what the assets return, and that measurement takes a decade. The first is visible immediately and the second is not visible at all when the decision is made.

A second-order effect compounded this. Directing banks to lend on policy grounds changes what a loan decision is. A loan extended because it was instructed carries an implicit expectation of support if it fails, which means the credit assessment that would normally constrain it does not bind. The lending was not mispriced by accident; the pricing mechanism had been set aside deliberately, for a reason that was defensible in early 2009.

Local government financing vehicles

The instrument that made the programme executable is also the one that created its most durable liability, and it is a clean example of a constraint being circumvented rather than removed.

The constraint. Local governments were, at the time, prohibited from borrowing directly. That rule existed to prevent exactly the accumulation that followed.

The requirement. Local governments were nonetheless expected to deliver a large share of the investment programme, and lacked the revenue to fund it.

The resolution. Separate legal entities — financing vehicles — were established, owned or sponsored by local governments. The vehicle could borrow where the government could not. It borrowed from banks, invested in infrastructure, and was capitalised substantially with land.

Four features made this structurally significant:

• The liability sat outside the government's formal accounts, so the constraint was satisfied in form while being defeated in substance.
• Land was the collateral and the equity. Vehicles were endowed with land whose value depended on continued property-price appreciation — so the collateral and the activity being financed shared a single driver.
• The revenue model depended on land sales. Local governments increasingly funded themselves by selling land use rights, making fiscal capacity a function of the property market.
• The guarantee was implicit. Lenders believed the sponsoring government stood behind the vehicle without any explicit guarantee, so the credit was priced on a promise that had no legal form.

Each of these appears elsewhere in this archive. Off-balance-sheet vehicles with implicit sponsor support are precisely the structure the US Housing & Mortgage Report 2008 describes returning to bank balance sheets when the vehicles could not fund themselves. Risk that is moved contractually and retained economically is the recurring pattern, and the fact that it appears in a completely different institutional system is what makes it structural rather than a feature of Western finance.

What "it worked" means

The programme is frequently described as successful, and it was — provided the claim is stated precisely.

What it achieved, and these are real:

• Chinese growth was sustained through a global contraction, avoiding an employment shock in a country where that carried consequences beyond the economic.
• Global demand had a floor. With the US and Europe contracting simultaneously, Chinese investment was the marginal source of demand for a wide range of traded goods, particularly industrial inputs.
• Commodity exporters recovered quickly, experiencing an interruption rather than a lost decade.
• The deflationary spiral was arrested. In 2009 the tail risk was a self-reinforcing collapse in global demand, and the programme removed it.

What it did not achieve:

• It did not rebalance the economy toward consumption, which was already the stated long-term objective. It did the opposite, raising investment's share of output.
• It did not resolve the excess-capacity question, and in several industries it enlarged it.
• It did not produce assets that uniformly earned their cost of capital. Some infrastructure was plainly valuable; some was built because building was the instrument.

The honest assessment is that it was the right decision with a real cost, and that framing matters more than a verdict. Faced with a collapse in external demand, the available instruments were the ones that could be deployed quickly — and directing banks to lend into investment is fast, whereas building a consumption-led economy is a decade-long project.

The cost was not the spending. It was the deferral. An investment-led response creates demand today and capacity tomorrow. If the capacity is used, the debt is serviced and the deferral resolves. If it is not, the debt remains and the excess capacity suppresses prices in the industries carrying it. That question was left open in 2009 and has been answered slowly and unevenly since.

The commodity channel

The mechanism by which Chinese domestic policy became the dominant variable for a set of unrelated economies is worth setting out, because it was widely misread as structural strength.

The chain is short and direct:

1. Infrastructure and property investment is exceptionally commodity-intensive — steel, copper, cement, energy — far more so than consumption.
2. China was already the marginal buyer in several of these markets, so a change in its investment rate moved the global price rather than being absorbed.
3. Higher prices transformed the terms of trade for exporters of those commodities.
4. Improved terms of trade raised export revenues, currencies, fiscal receipts and equity markets across the resource complex.
5. Capital flowed toward those economies, which appeared to be growing strongly on their own account.

The misreading was to treat step five as independent. Resource economies in 2010–2013 exhibited every appearance of structural strength: strong currencies, healthy fiscal positions, rising investment, upgraded ratings. The underlying variable was a single buyer's investment rate, driven by a policy decision.

A terms-of-trade improvement caused by one buyer's stimulus is a cyclical position wearing structural clothing. The distinction is invisible while the buyer keeps buying, and it is the only thing that matters afterwards.

The corollary is the archive's common-factor principle again. A portfolio diversified across Australia, Brazil, Chile, Russia and the resource sector looked geographically broad and was one position with respect to Chinese fixed-asset investment. The Russia & Commodity Exporters Report 2014 describes what happened when that variable turned, and the US Housing & Mortgage Report 2008 describes the identical error in mortgage pools.

What was accumulated

The programme's legacy is a balance sheet, and its composition determines how the following years played out.

On the asset side: a very large stock of transport infrastructure, urban development and industrial capacity, built quickly. The quality is genuinely mixed and the average is not the useful statistic — high-return trunk infrastructure and low-return duplicative capacity were financed on the same terms.

On the liability side: credit extended by banks to vehicles and enterprises whose ability to service it depends on those assets earning returns, and on land values.

Three characteristics made the liability durable:

• It was not marked. Loans to policy-directed borrowers with implicit government support are not repriced in a market. The absence of a price is not the absence of a loss — the same point the US Venture Capital Report 2008 makes about untested private marks, in a completely different asset class.
• Rolling was available. A borrower who cannot repay but can refinance does not default. Extension converts a credit event into a slow drag on capital allocation, which is less disruptive and more persistent.
• The land dependency was circular. Vehicles were capitalised with land, funded by land sales, building assets whose value depended on the surrounding land. A single variable supported the collateral, the revenue and the asset, and the diversification implied by having three of them was illusory.

The result is the pattern this archive calls a deferral that compounds. Nothing was written off. The test of whether the assets justified their cost was postponed, and postponement had a price — paid in capital that remained allocated to activities that could not service their debt.

What an allocator could act on

Measure stimulus by credit, not by the deficit. A programme executed through directed bank lending does not appear in fiscal statistics. Comparing packages on headline fiscal numbers will understate anything delivered through the banking system, sometimes by a very wide margin.

Look for the vehicle when a constraint appears to have been satisfied. A borrowing prohibition met by creating an entity that can borrow has not been met. The question worth asking of any binding constraint is what structure was created to accommodate it.

Treat an implicit guarantee as a real exposure with no documentation. Lenders priced local vehicle debt on sponsor support that was never legally given. That belief can be revised without any default occurring, and the revision is the risk.

Separate cyclical terms of trade from structural strength. Resource economies in 2010–2013 displayed every marker of durable improvement, driven by one buyer's investment rate. Ask which variable the improvement depends on and whether it is a policy decision.

Check whether collateral, revenue and asset share a driver. Land as collateral, land sales as revenue, and property as the asset is one exposure held three times. Apparent diversification within a structure is worth decomposing to its underlying factor.

Distinguish an unmarked exposure from a sound one. Loans that are rolled rather than repriced produce no visible loss and a persistent misallocation of capital. The absence of a mark is a statement about the market's existence, not about value.

What 2009 established

• Credit-directed stimulus is a distinct instrument from fiscal stimulus, faster to deploy, invisible in deficit statistics, and settled over a decade rather than immediately.
• A borrowing constraint met through a financing vehicle relocates a liability rather than preventing it, reproducing the off-balance-sheet structure of the Western crisis in a different institutional system.
• China became the marginal source of global demand, making its domestic policy the dominant variable for commodity exporters that appeared to be strengthening independently.
• Investment-led responses defer rather than resolve, because they create demand now and capacity later, and the return on that capacity determines whether the financing was sound.
• Unmarked, rollable debt converts a credit event into a persistent drag on capital allocation, which is less visible and considerably more durable.