Macroeconomics and the Chinese economy

What GDP measures

GDP is value added. Gross domestic product (GDP) is the value of final goods and services produced in a region over a period. “Final” matters: the chips inside a phone are intermediate inputs, and counting both the chips and the phone would count the chips twice.

Statisticians add up value added, not final goods. Whether something is final or intermediate is often ambiguous. Crude oil is an input for a refinery but a final export for the country that pumps it; electricity and artificial-intelligence (AI) tokens are final when used for entertainment and intermediate when used for research. So each producer’s value added, its sales revenue minus purchased intermediate inputs, is counted, and GDP is the sum over all producers.

Value added is not profit. Wages are not subtracted. Value added is the income the producer generates for its workers, its owners and the government together.

Investment goods are final by convention. A new machine or building counts in the year it is made. Buying an existing machine or apartment is a transfer of ownership, not production; only the broker’s commission is new output.

The test for every case: is somebody making a production effort that creates value this period, and is anything being counted twice?

Place versus passport. GDP is geographic: it counts production inside the borders. Gross national product (GNP), also called gross national income (GNI), counts the income of a country’s nationals wherever it is earned. Professor Bai argues GNP is the better objective, because it is what benefits a country’s people, but it needs ownership and dividend data that statistics bureaus cannot collect, so GDP is the working measure.

Can the numbers be trusted?

Incentives cut both ways. Firms have reasons to under-report to reduce tax; governments have reasons to over-report growth. China keeps a legal firewall between the tax bureau and the statistics bureau so that reporting to statisticians has no tax consequence, but firms do not fully believe it.

Coverage is uneven. Large firms report directly; small firms are sampled and scaled up. Housework, domestic help and the informal sector are structurally missed, so economies with larger informal sectors tend to be more under-measured.

Government output is valued at cost. A ministry’s value added is its total spending minus purchased equipment and inputs. There is no market price for defence, so what it costs is taken as what it is worth.

Nobody audits a country’s GDP. No government would let the World Bank or the International Monetary Fund (IMF) do it. One careful study cited in class concluded that China’s statistics are imperfect but comparable in quality to those of peers at a similar stage of development.

Proxies help but can mislead. Electricity use, night-time light seen from satellites and tax revenue are common cross-checks. Each assumes a stable relationship with output that can break: data centres raise electricity use without a matching rise in measured GDP.

Three ways to measure the same total

Production approach: add up value added by producer.

Income approach: split each producer’s value added into labour cost (wages, payroll taxes and social-security contributions), capital cost (interest, dividends, retained profit and taxes on capital) and taxes on production. Taxes on production are those that are neither labour nor capital taxes: value-added tax (VAT), China’s most important tax, and sales taxes such as the roughly one-third tax on gasoline.

Expenditure approach: Y = C + I + G + NX, where Y is output, C household consumption, I investment, G government purchases and NX net exports (exports minus imports). Sales between firms net out unless they are investment goods.

Why the three always agree: output that is not sold is counted as inventory investment, as if the producer had bought it. By construction, what is produced equals what is spent, and every unit of spending is someone’s income.

Investment is new productive capacity, bought by firms, government or households. A newly built apartment is household investment; buying shares or a second-hand home is not, because nothing new is produced.

G is purchases, not transfers. Civil-servant salaries and public vaccination programmes are in G. Pensions, medical reimbursements and subsidies to firms are transfers: nothing is bought, and the money shows up in C or I when it is spent. Counting it in G as well would count it twice.

Choose one approach and stay in it. Mixing income-side and expenditure-side reasoning in one sentence is the most common source of confusion.

Assets, housing and comparing countries

Changes in asset values are not production. An apartment falling in value from 10 million to 8 million does not reduce GDP, and a resale adds only the agent’s commission.

Homes produce housing services. Rent is a purchase of housing services. Owner-occupiers receive the same service, so an imputed rent is added. China’s imputation has been based on construction cost rather than market rents, a large under-estimate; moving toward market values would let house prices feed into measured GDP.

Market rates versus purchasing-power parity (PPP). Converting at market exchange rates compares what incomes buy internationally. PPP converts at the rate that makes the same basket cost the same in both places: if a basket costs 40,000 yuan in China and 8,000 dollars in the United States, the PPP rate is 5 yuan per dollar. Because services are cheaper in poorer economies, PPP raises their measured income. At PPP China’s economy is already about 130% the size of the US economy; at market rates it is about two-thirds.

Levels are nominal, growth is real. The GDP level is reported in current prices; the headline growth rate (China’s roughly 5%) is adjusted for inflation.

Stocks, flows and what GDP leaves out

Stocks and flows. Think of a bathtub: the faucet is a flow, the water level a stock. Investment is a flow that adds to the capital stock; a government deficit is a flow that adds to the stock of debt.

Depreciation and depletion are not deducted. Oil sold at 90 dollars a barrel counts fully in GDP even though the reserve shrinks. “Green GDP”, which would subtract environmental damage and resource depletion, is attractive in principle but has not been made operational.

GNP matters more for ageing, high-saving economies. Japan, Singapore, Hong Kong and increasingly China invest abroad; the returns are national income but not domestic product. Japan’s “lost decades” look noticeably better measured by GNP.

The limits of valuing at cost. The classic objection, paying people to dig holes and fill them in, is a real limitation of the measure; most government activity does create value.

How income is shared between labour and capital

The benchmark model. One representative producer uses capital K and labour L with a stable production function Y = F(K, L). Each factor has one market price: the real wage for labour and the real rental rate for capital.

Diminishing returns and constant returns. Adding workers to a fixed stock of machines raises output by less and less: the marginal product of labour (MPL) falls. Scaling all inputs together scales output in proportion.

Each factor is paid its marginal product. A worker who adds 8,000 of output will not be kept at a wage of 10,000; one who adds 12,000 will be hired. With constant returns and competition, labour income plus capital income exactly exhaust output.

Cobb-Douglas. The form Y = A·K^α·L^1−α gives each factor a constant share of income: α to capital and 1 − α to labour, where A is the level of technology. It is used because the US labour share has been remarkably stable at about two-thirds for decades. In China capital’s share is closer to one-half.

The natural rate of output (Ȳ) is what the economy produces with given technology and factor supplies at full employment, allowing for normal job-to-job movement. It is the benchmark for everything in the short-run part of the course.

Demand, saving and the trade balance

Consumption depends on disposable income: C = C(Y − T), where T is taxes net of transfers. The slope, the marginal propensity to consume (MPC), is the share of an extra yuan of disposable income that is spent; in China it is about 0.7.

Investment falls as the real interest rate r rises. The interest rate is a cost even when a firm uses its own money, because that money could have earned the market rate elsewhere: an opportunity cost.

Saving. Private saving is Y − T − C; public saving is T − G. Together they form national saving S. In a closed economy saving equals investment, and the interest rate is the price that brings them into line.

The external balance. In an open economy, S − I = NX: saving that is not invested at home is lent abroad, and the counterpart is a trade surplus. Professor Bai calls this the most important formula for understanding a country’s external balance.

Why tariffs rarely close a trade deficit. The United States invests more than it saves, so it runs a deficit. A tariff changes the deficit only if it changes saving or investment. It works mostly as a tax on US consumers; that could raise saving, but if the revenue is spent, S − I and hence the deficit barely move. China saves more than it invests and runs a surplus for the mirror-image reason.

Why China’s household consumption share is low

The pattern. Household consumption fell from about 47% of GDP in 2000 to about 35% in 2009, recovered somewhat to 2015, and has stayed near 40% since. For comparison, US household consumption is about two-thirds of GDP, and in Japan and the euro area it is a little over half. China’s GDP is about 140 trillion yuan, around 20 trillion dollars at market rates, against roughly 30 trillion for the United States.

Share is not level. Over 2000 to 2009 real GDP grew to roughly two and a half times its size, so real consumption still rose by about 80% even as its share fell: 0.47 × 100 = 47 became roughly 0.35 × 246 ≈ 86. Nobody consumed less; consumption grew more slowly than output.

A framework for consumption. Households consume out of current disposable income (labour income plus investment income minus taxes), out of wealth (past saving, inheritance, luck), and with an eye on the future: expected income, expected expenses such as old-age care, and uncertainty, which a social safety net for health and pensions reduces. All in real terms.

Demographics. Longer lives without a later retirement age force more saving; China began raising its retirement age gradually in 2025 after long resistance. Lower fertility reduces child-related spending and, through expected housing demand, house prices and household wealth. China’s population has already peaked, earlier than the United Nations’ 2019 projection of around 2030.

The puzzle of the 2000s. After entry to the World Trade Organization (WTO), prospects, house prices and the safety net were all improving, yet the share fell. Five explanations were developed in class:

1. Housing reform. From the late 1990s housing moved from employer allocation to purchase, so household income went into buying homes. Land is a government monopoly, and land sales financed infrastructure when tax and bond markets were weak, so households effectively paid for it through land prices.

2. Migration under hukou. About 300 million rural migrants work in cities without urban household registration (hukou), and so without equal schooling, safety net or the right to buy urban housing. They keep their roots in the village and save most of their much higher income.

3. A falling labour share. A farmer keeps all of the output; a factory worker keeps roughly half, with the rest going to more concentrated capital owners who save much more of it. Professor Bai’s research finds migration was the single most important cause of the falling labour share.

4. Habit. Lifestyles adjust more slowly than income, and older cohorts are more frugal than younger ones. Culture explains differences in levels between countries but not a ten-year trend.

5. Weak wealth accumulation. Figures shown in class put 20-year returns on China’s mainland A-share market at about 6.5% a year including dividends, well below the returns on comparable Hong Kong and US listings, and most small investors lost money trading in and out. Housing no longer builds wealth either.

The mirror image abroad. China’s trade surplus peaked near 9% of GDP in 2007, fell to about 2% and has recently risen to around 4%. With S − I = NX, weak consumption at home shows up as surplus abroad.

What could raise household consumption

Mandated pay rises do not help. Firms substitute machines for workers. The elasticity cited in class is about one: a 1% rise in labour cost cuts employment by about 1%, leaving labour income unchanged. Creating jobs works better.

Inflation is not the constraint. China has had mild deflation, which squeezes profits and tax revenue. Moving from about −1% to +2% inflation would help.

Interest rates have little room. China’s 10-year government bond yield was about 1.8% against nearly 5% in the United States; a wider gap would encourage capital to leave.

Services need market prices. Utility charges, road tolls, bus fares and public-university tuition (around 5,000 yuan a year at many universities, frozen for most of two decades) are far below cost. Raising them is politically painful, but future jobs will come from services such as medical and elderly care, and those sectors need prices that cover costs to grow. There are some 40 million disabled elderly people and care beds for only about a third of them, so the demand is there; the question is who pays.

Transfers and the safety net. Cash transfers and better social insurance raise consumption, but once granted they are expected permanently and must be financed. China runs two systems: employment-based insurance for about 375 million people, and a residents’ scheme for about 1 billion with much thinner benefits.

Equal treatment for migrants would raise their consumption. Urban residents resist because services and university places would be shared more widely.

Foreign reserves are not spare money. About 3 trillion dollars, the world’s largest (Japan is second at a little over 1 trillion), is roughly 15% of GDP; against a broad fiscal deficit near 8% of GDP it would last about two years. Professor Bai also doubted China could count on outside rescue.

Jobs must come from services. Unemployment was about 5.3% overall but about 18% for ages 16 to 24, roughly double the US rate for that age group (China’s figure excludes students). Manufacturing is highly automated, so new jobs will be in services.

Profits and competition. Corporate profitability is the real problem for the stock market. Competition is too fierce: some 3,000 solar-panel plants, roughly one per county on average, in an industry losing money. Local governments keep “zombie” firms from exiting, and weak protection of property rights lets innovations be copied immediately.

The identity mismatch. Jobs in demand, such as care work, nursing, kindergarten teaching and welding, go unfilled because many young people feel they are not “their” jobs, even where training is cheap. Professor Bai calls this a cultural failure rather than a market failure, and the hardest barrier of all.

The summary framework: raise current income through jobs; raise wealth through profitable firms and a working stock market; improve expectations through a stable policy environment and a stronger safety net; and break down the identity mismatch.

Local governments and the incentive to invest

Local governments hold the levers of land, licensing and regulation, and need discretion because the formal rules are imperfect.

Their incentives favour investment over consumption. Officials are judged on GDP growth and typically serve about three years in a post, so borrowing for visible, fast-built investment pays off for them while the debt passes to successors. Investment can be attracted by local policy; consumption is a household decision.

Taxes follow production, not consumption. Because VAT is collected where goods are made, unlike a US-style sales tax collected where they are bought, localities compete for factories and protect local producers rather than consumers. That helped early in development; it is now a drag.

Professor Bai’s proposal: make household consumption growth a performance indicator for local officials. That would need reliable provincial consumption data, which the national household survey is too small to provide, and tax revenue collected or shared by where consumption happens, which digital payments now make feasible.

The stage has changed. Production capacity is no longer China’s constraint; demand is. Fixing it is structural “plumbing”, not something monetary policy alone can do.

Saving, capital and the steady state

Why a 1950s model still matters. Robert Solow’s model makes strong simplifying assumptions and gets clean, durable results. The question for China is not whether growth has fallen from about 10% (it naturally has) but whether the economy is reaching its lower potential.

Per-worker terms. With constant returns, output per worker y depends only on capital per worker k: y = f(k). A fixed share s of income is saved and invested, and capital wears out at the depreciation rate δ.

The bathtub again. Investment s·f(k) flows in; depreciation δ·k leaks out. Because returns to capital diminish while depreciation rises in proportion to capital, the leak eventually matches the inflow. That level is the steady state k*: the economy converges to it from above or below and then stops growing in per-worker terms.

Prediction 1: saving changes the level, not the long-run growth rate. A higher saving rate moves the economy to a higher steady state, with a burst of growth on the way, and then per-worker growth stops again.

Steady state is not standstill. The economy keeps producing, investing and replacing capital; only capital per worker stops changing.

Population growth

More workers need more capital. If the workforce grows at rate n, keeping capital per worker constant requires replacing worn capital (δk) and equipping the newcomers (nk). Break-even investment is (δ + n)·k.

Prediction 2: faster population growth lowers income per worker but not its long-run growth. The break-even line is steeper, so k* and y* are lower.

Bigger is not richer. In the steady state income per worker is constant, but total GDP grows at n because there are ever more workers. A larger population makes a larger economy, not a richer one.

Extra investment cannot escape the steady state. Beyond k* the extra output is less than what is needed to maintain the extra capital, so the economy slides back. Only a change in the production function itself can lift the path.

Efficiency: the only source of lasting growth

Labour-augmenting technology. Let E be the efficiency of each worker: one worker with E = 2 counts as two effective workers, so the effective labour force is L·E. If E grows at rate g, the mechanics are the same as for population: measure everything per effective worker, and break-even investment becomes (δ + n + g)·k.

Steady-state growth rates. Capital and output per effective worker are constant; output per worker grows at g; total output grows at n + g.

Solow’s most powerful result: sustained growth in income per person comes only from efficiency improvement. Saving cannot deliver it, and neither can population growth.

Read “technology” broadly. Professor Bai prefers “efficiency”: science and technology, but also education, management, institutions and a labour market that puts people to work. Textbooks use technological progress, productivity growth and total factor productivity (TFP) almost interchangeably.

What the model leaves out: unemployment, differences between workers, the distribution of income and any shortage of demand. It describes growth potential only.

Does the model fit the data?

Models are tested, not proved. Trust grows as a model’s implications keep matching the data.

Balanced growth. In the steady state capital and output grow at the same rate, so the capital-output ratio K/Y should be roughly constant. For most economies it moves very little. China’s rose about 0.5% a year from 1978 to 2007: consistent with an economy converging up to its steady state from below.

How the capital stock is estimated. Nobody counts every machine. Statisticians cumulate past investment (China’s series starts in 1952), subtract depreciation using assumed lifetimes (around 40 years for structures, 10 for machinery) and adjust for price changes. Intangible capital such as research and development (R&D) is handled crudely.

Efficiency is estimated, not observed. It is the residual left after measurable inputs are accounted for. In advanced economies it has grown about 2% a year since the Industrial Revolution; before then, hardly at all.

Other predictions that hold: real wages rise with output per worker, so the labour share is stable (clearly in the United States; less so in fast-changing China), and the real interest rate fluctuates around 2% in advanced economies.

Convergence

The most hopeful idea in growth theory. Other things equal, poorer economies grow faster. Far below the steady state a little investment adds a lot of output; at the steady state it adds nothing. When Professor Bai first went to the United States in 1985, China’s income per head was about one-fiftieth of America’s. Singapore, independent in 1965, now has a higher income per head than the United States.

Two engines of catch-up. The first is accumulating capital up to the steady state with a given technology. The second is learning from the frontier, which shifts the production function itself. Knowledge flows across borders, and imitation, often scorned, is how development works. Both engines weaken as the gap closes, which is a large part of why China’s growth has slowed from about 10% to about 5%.

Learning requires the capacity to learn. Education, R&D and openness switch convergence on.

Convergence is conditional, not automatic. Only a handful of economies have escaped the middle-income trap. Among economies with similar saving, population growth and human capital, income gaps close by about 2% a year; across all economies there is no such pattern.

How much saving is best? The golden rule

Maximise consumption, not output. In the steady state, investment just covers depreciation, so consumption per worker is c* = f(k*) − δk*: the gap between the output curve and the depreciation line. Saving nothing leaves no capital and no output; saving everything leaves nothing to consume. The golden-rule capital stock is where the gap is widest, where the marginal product of capital (MPK) equals δ.

With population and efficiency growth the test becomes MPK − δ compared with n + g. If the net return to capital exceeds the growth rate, the economy should save more; if it is lower, it saves too much.

The US saves too little. Using numbers from class: capital is about 2.5 times GDP and capital income about 30% of GDP, so MPK ≈ 0.30 ÷ 2.5 = 12%; depreciation is about 10% of GDP, so δ ≈ 4%. The net return of 8% is far above growth of about 3%.

China saves too much, Professor Bai argues. Its net return to investment is below its roughly 5% growth rate, so it is beyond the golden rule and should consume more. The exact golden-rule saving rate is unknown, but, as Professor Bai put it, it is easier to tell which side of it you are on.

Why China grew 10% a year for three decades

Suppressed potential, released. From 1978 to 2007 growth averaged about 10% a year, doubling the economy every seven years. Professor Bai’s explanation is that reform released potential the planned economy had held back.

The ingredients: a population with widespread basic education, built up during the planning era even though universities lost a decade to the Cultural Revolution; urbanisation from about 15% to about 60%, freeing some 300 million people from subsistence farming; markets allowed to work; infrastructure built quickly; learning from abroad, including sending students overseas; and bold reforms under pressure, such as entry to the WTO in 2001 and the restructuring of state-owned enterprises.

The recipe does not transfer. Other developing countries did not have the same backlog of suppressed potential to release.

No return to 7%. The scope for learning is narrower and the international environment tougher. Professor Bai would be pleased with 5% and thinks reform could add at most half a point.

Investment and efficiency

Across the world, investment and efficiency growth move together. New equipment embodies new technology, investment has spillovers, efficient economies attract capital, and a good business environment encourages both.

China is the exception. Across Chinese provinces, higher investment rates have gone with slower efficiency growth. China invests more than 40% of GDP, about double the average of the Group of Twenty (G20) economies excluding China, and around twice the US rate. That may put it on the far side of an inverted U: invest too fast and the projects get worse.

The bridge with no river. The extreme case: a bridge built on dry land for a river still to be dug, paid for with an IOU, to hit a growth target.

Regional balance versus efficiency. Markets send capital where returns are high; governments push it to lagging regions, where a weak business environment is often why they lag. People then move away, weakening those regions’ finances further.

Who should allocate investment? Industrial policy

The one-good model hides real choices. Real economies invest in private buildings and machinery, human capital, R&D and public infrastructure, with different horizons and different gaps between private and social returns.

Industrial policy is back. Long out of favour, it is now practised almost everywhere. The case against: can governments pick winners, and will politics override economics? The case for: private investors ignore social benefits and costs, so markets under-invest in some areas and over-invest in others.

How China plans: the five-year plan

The cycle. A plan is adopted in March of its first year; ministries, regions and institutions then write their own plans beneath it. A mid-term review in year three revises the current plan and starts research for the next. The National Development and Reform Commission (NDRC) drafts; the Communist Party’s Central Committee publishes its recommendations in the October before; after consultation, the National People’s Congress votes the following March. Professor Bai sits on the national expert committee for the 15th plan (2026–2030).

Plan first, budget second. The NDRC sets the plan and the Ministry of Finance budgets to fit it, the reverse of the US Treasury’s central role. Most financing is domestic and private.

Targets are taken seriously. The 2012 pledge to double GDP between 2010 and 2020 shaped every later plan and was nearly met despite COVID-19. The 15th plan sets no numerical growth target because of uncertainty, but keeps the 2035 goal and about 20 numerical targets planners are confident of, such as life expectancy and care beds for the elderly. A higher household consumption share is a goal without a number, because households decide it.

The 2035 goal: a “moderately developed” economy, read as about 20,000 dollars per head against 14,000 to 15,000 today (the US figure is about 90,000). Set in 2020, it implied doubling GDP over 15 years, about 4.7% a year.

Rigid targets can backfire. An energy-use cap derived from a carbon target collided with booming exports; power was rationed and factories closed for months to meet it, and planners later dropped the indicator.

How business feeds in: major firms are consulted by sampling, emerging industries are assessed with scientists and engineers, and entrepreneurs can use the People’s Congress, the Chinese People’s Political Consultative Conference (CPPCC) and the All-China Federation of Industry and Commerce.

Estimating potential growth

The 8% view of 2008. An influential argument held that China could average 8% for 20 years, because economies at one-fifth of US income per head had done so: Japan from 1951 grew 9.2% a year for 20 years, Singapore from 1967 8.6%, Taiwan 8.3%, Korea 7.6%. The government chose a more cautious 7%. Professor Bai sees the 8% view as a source of later problems: over-ambitious targets, inefficient investment and an over-built property sector.

What the analogy missed: demographics. Japan’s dependency ratio (people under 15 and over 65 relative to those of working age) fell for decades after 1951; China’s bottomed out in 2010 and has risen since. China’s labour-force participation was also already high.

The method. Split GDP growth into labour-productivity growth and labour-force growth. Plot productivity growth against productivity relative to the United States for 18 East Asian and Western European economies: the upper edge forms a tight, downward-sloping band, the growth potential at each level of development. Labour-force growth is projected separately from demographics.

Track record. Projections made with pre-2008 data matched China’s actual productivity growth for 2008 to 2014 closely. For the 13th plan (2016–2020) the method gave 6.36% against the 6.5% needed for the doubling target, and helped stop a higher target being set; actual growth came in a little below 6.5%.

Now. China’s labour productivity is about a quarter of the US level, and the suggested potential for 2026–2030 is about 4.5%. If the potential is realised, by 2050 China’s economy would be about 1.8 times the size of the US economy at PPP, but income per head only about 53% of the US level. The point of the exercise, in Professor Bai’s words, is to make policymakers more conservative: this is the potential if everything goes right.

Growth accounting: where the growth came from

The standard decomposition. With Y = A·K^α·H^1−α, where H is human capital (workers weighted by schooling), output growth splits into efficiency growth plus α times capital growth plus (1 − α) times human-capital growth. Everything except A can be measured, so efficiency growth is the residual: the Solow residual.

Professor Bai’s variant uses the capital-output ratio instead of capital. Because higher efficiency also induces more investment, this credits that induced investment to efficiency and human capital, doubling the weight on efficiency. What remains in K/Y is investment for other reasons, such as stimulus.

1978 to 2007, growth about 10%: roughly 60% from efficiency, 35% from human capital and 5% from a rising capital-output ratio. Textbook growth. Human capital mattered because the working-age share was rising and college intake grew from a few hundred thousand a year in the late 1970s to about 9 to 10 million now.

2008 to 2015, growth 8.7%: efficiency’s contribution fell from about 6 to 2.6 percentage points a year, and most growth came from capital deepening. That cannot last, because a rising capital-output ratio needs ever more investment. The room for it came from the trade surplus shrinking rather than from lower consumption.

Why efficiency growth slowed: diminishing catch-up, a smaller farm sector to draw workers from, a growing service sector with slow productivity growth (the Baumol effect), a falling trade-to-GDP ratio, and falling labour-force participation as people stay in school longer and retire at 60.

Capital’s share. α is about 0.5 in China against about one-third in the United States, because industry pays a smaller share to labour than agriculture or services. As the economy shifts toward services, the labour share should rise.

The 2009 stimulus and its long shadow

The investment rate tells the story. It rose after WTO entry, dipped, and then the 2009 stimulus pushed it to about 47% of GDP in 2010 and 2011 before a slow decline.

How the credit flowed. After the clean-up of bad loans around 2000, banks had become commercially minded and reluctant to lend to local governments. In late 2008 regulators signalled that a bank refusing a local-government loan would have to justify the refusal. Bank credit rose by about 13% of GDP in the first quarter of 2009. As the economist John Taylor has shown, US federal grants to states in the same crisis largely went to reducing state borrowing; Chinese local governments treated the signal as a mandate to expand.

The lesson, perhaps learned too well. The central government is now reluctant to stimulate even though its own debt is only about 23% of GDP (US federal debt held by the public is about 100% of GDP). Professor Bai’s view is that stronger stimulus is warranted after the property correction, but through new channels: the 15th plan emphasises networks such as the power grid, data centres, logistics, underground utilities, water and communications, rather than roads and housing.

Long run and short run

Two halves of macroeconomics. Growth is about the long run, when everything adjusts, and the Solow model has no prices in it at all. Stabilisation is about the short run, when wages and prices are rigid and shocks move output away from its natural rate Ȳ. The IS-LM model, old but still useful, is the short-run tool.

Why prices are sticky. Wages are set in contracts for a period. Reprinting price lists is cheap, but changing prices costs customer goodwill unless there is an obvious reason such as a war. So the model first holds the price level P fixed, then asks how it moves.

The policy questions: has China invested too much, and what did the post-2008 stimulus do?

The IS curve and the multiplier

IS: the goods market. The IS curve (for investment and saving) is every combination of interest rate r and output Y at which the goods market clears: Y = C(Y − T) + I(r) + G. A lower interest rate raises investment, and output must rise to meet the extra demand, so the curve slopes down.

Fiscal policy shifts IS. A rise in G raises demand at every interest rate. The government’s yuan becomes a firm’s revenue, then a household’s income, of which 0.7 is spent, and so on. The multiplier is 1 ÷ (1 − MPC): about 3.3 with China’s MPC of 0.7, if the interest rate stays put.

3.3 is a ceiling, not a forecast. It is too large to be realistic, because the interest rate does not stay put. That is where the money market comes in.

Money and the LM curve

Money is the liquid part of wealth: cash and checking accounts, not the one-year deposit. People hold it to make transactions, so demand for money rises with income. Holding it means giving up the interest on less liquid assets, so demand falls as the interest rate rises. Only real balances, money M divided by the price level P, matter.

Treat r as a price. The interest rate is the price of holding money. “Interpret r as a price and a lot becomes easier.”

LM: the money market. The LM curve (for liquidity and money) is every combination of r and Y at which real money supply equals money demand L(r, Y). Higher income raises money demand, so with supply fixed the interest rate must rise: the curve slopes up. Here L means liquidity, not labour.

Monetary policy shifts LM. Printing money lowers the price of money, the interest rate; tightening raises it. “Accommodative” policy means more money and lower rates.

Payment technology matters. Electronic payments linked to interest-bearing accounts reduce the need for cash. A loss of trust in e-payments would raise money demand, push up interest rates and cut investment: a real effect from a change in payment habits.

Policy in IS-LM: crowding out and aggregate demand

Equilibrium is where IS and LM cross: two markets, linked by r and Y, give one interest rate and one level of output.

Fiscal expansion and crowding out. A rise in G shifts IS right. Higher output raises money demand; with money supply fixed, the interest rate rises; the higher rate cuts private investment. Output still rises, but by less than the multiplier. Government spending crowds out private investment, although infrastructure that lowers business costs can crowd it in.

Monetary expansion. More money shifts LM down and to the right: the interest rate falls, and investment and output rise.

Aggregate demand (AD). A higher price level shrinks real balances M/P, raising the interest rate and lowering output. So demand for output falls as prices rise: a downward-sloping AD curve. Monetary or fiscal expansion shifts it right.

From the short run to the long run

How prices move. Above the natural rate Ȳ people work overtime, costs rise and firms raise prices. Below it there is slack and unemployment, and prices fall. China today has demand too weak for its capacity, so prices are under downward pressure.

After a fall in government purchases, such as at the end of a war: output falls below Ȳ; over time prices fall, real money balances rise, the interest rate falls and investment recovers until output is back at Ȳ, with less government and more private investment. That is the “peace dividend”.

Money is neutral in the long run. Starting at Ȳ, extra money overheats the economy; prices rise until real balances are back where they were, and output, the interest rate, consumption and investment all return. Only the price level is higher. Below Ȳ, monetary policy still has real effects.

Markets can tighten without the central bank. A central bank that refuses to raise rates while inflation persists loses the argument anyway: shrinking real balances push market yields up.

Professor Bai’s closing line: “These are very simple models, but if you understand the insights, you can make a lot of applications.”

Rules on paper, discretion in practice

The puzzle. In 2012 the World Bank’s Doing Business indicators ranked China only 91st of 185 economies. How could an economy with such a poor measured business environment grow so fast?

What the indicators miss. They measure formal rules on paper, assessed by simulating a start-up in one or two large cities, not how rules are implemented or worked around.

Selective help from local governments. Local officials are very business-friendly but cannot help everyone, so they help selectively. Firms in industries favoured by the five-year plan are natural choices; in one city studied, each senior official looked after a portfolio of firms and escalated their problems with land, environmental approvals and regulation. Professor Bai’s research suggests selection is driven mainly by business prospects. In hindsight he describes it as the exercise of discretionary power rather than preferential treatment.

The evidence. Because investment agreements with local governments are treated as commercial secrets, the research traces firms’ ultimate shareholders through registration data. Private firms with a state entity among their ultimate shareholders are more productive and obtain cheaper credit.

The direction of travel. As formal rules improve, there is less need for discretion, and exercising it is becoming riskier for officials.

Rule-based versus pragmatic governance. Rules sound right, but when the rules are bad, pragmatism is needed. Professor Bai calls this probably the most important tension in development.

A case in point: the property tax

China has no recurrent tax on residential property, although businesses pay one. Professor Bai proposed a residential property tax more than twenty years ago; it never passed, and a weak property market now makes it impractical.

The main resistance comes from local governments. Land-sale revenue is paid up front, and buyers would pay less for land once a future tax was priced in. For a local government, a large lump sum now beats a small annual tax.