Growth Strategies: The empirics of policy reform and growth

March 2014

• Over the past 50 years the world’s richest economies have seen gradual declines in growth rates of GDP per person. The median rate has fallen from over 4% in the 1960s to just 0.02% in the aftermath of the financial crisis, according to the IMF. After a slump in the 1980s and 1990s, middle-income economies rebounded in the 2000s and have maintained a healthy growth rate more recently. Low-income countries have seen the most improvement over time: median growth was 3% in 2008-11. Even better, since the 1990s these countries’ “take-offs”— an expansion of income per person averaging at least 3.5% for a period of at least five years—have been more numerous and longer-lasting.

• The outlook for economic growth in the West is bleak, according to the IMF’s latest World Economic Outlook, released on April 16th. Worldwide output is expected to grow at just over 3% in 2013, but rich countries will lag behind, expanding at 1.2%. Growth in emerging markets, by contrast, will exceed 5%, with Asia and sub-Saharan Africa motoring along at 7% and 5.6% respectively. The IMF's report praises euro-area policymakers for avoiding a break-up, but notes that parts of the currency union are still uncompetitive and constrained by austerity. The failure of cash injections to trickle down to households and businesses is dampening prospects, too. The American economy, though buoyed by recovering housing and credit markets, must endure the effects of the budget sequester. 

Starting point: a change in emphasis

• Presumptive strategies – ISI– Washington Consensus– U.N. Millennium Project, MDGs– Governance agenda

• Long laundry list of reforms– In a wide range of areas: trade, fiscal, legal,

regulatory, health, education, etc.• Focus on complementarity of reforms rather than

prioritization or sequencing• A bias towards universal recipes, “best-

practices,” and rules of thumb

Starting point: a change in emphasis

• Diagnostic strategies – We do not know ex ante what works and what doesn’t– Need to look for binding constraints

• Which tend to be context-specific– Experimentation central part of discovery– Monitoring and evaluation equally central

• Focus on selective, more narrowly targeted reforms

• Based on the idea that there exists lots of slack– Well targeted reforms can produce a big bang

• Suspicious of “best-practice,” universal remedies– Looking for policy innovations that unlock local

second-best/political complications

The Litmus test

• Do you believe there is an unconditional and unambiguous mapping from specific policies to economic outcomes?– If yes, you are in the presumptive camp– If no, you must rely on the diagnostic strategy

The Outline

• Why this mapping does not exist (in general)– Arguments from theory– Arguments from evidence

• How conventional wisdom skirts the difficulty

Why rules of thumb don’t work: an illustration from trade policy

• The liberalization must be complete – or else the reduction in import restrictions must take into account the potentially quite

complicated structure of substitutability and complementarity across restricted commodities. • There must be no microeconomic market imperfections other than the trade restrictions in question

– or else the second-best interactions that are entailed must not be adverse • The government must be able to undertake a compensatory devaluation of the currency

– or else nominal wages must be downwardly flexible• The home economy must be “small” in world markets

– or else the liberalization must not put the economy on the wrong side of the “optimum tariff”• The economy must be in reasonably full employment

– or else the monetary and fiscal authorities must have effective tools of demand management at their disposal.

• The income redistributive effects of the liberalization should not be judged undesirable by society at large

– or else there must be compensatory tax-transfer schemes with low enough excess burden.• There must be no adverse effects on the fiscal balance

– or else there must be alternative and expedient ways of making up for the lost fiscal revenues.

• The liberalization must be politically sustainable and hence credible– or else the fear of reversal must not lead to too large a consumption boom in imported

durables

And none of this guarantees growth (as opposed to level) effects, which would require an even longer list of prerequisites!

When is trade liberalization desirable (in theory)?

The empirics of policy reform and growth

• What do cross-national growth empirics show?• Actually, not much• Some serious methodological flaws

– As practiced:• Fragility of results (non-robustness, non-linearities) • Outcome versus policy variables on the RHS

– Inherent:• governments use policy not randomly, but to achieve objectives--whether

good or bad• cross-national correlations can take any sign, depending on the source of

variation in the data

• Therefore, the systematic statistical evidence gives us little confidence that there is a determinate, uniform, and non-context specific relationships between specific policy reforms and economic growth

The tyranny of extreme outcomes

• Easterly (2005) runs growth regression with panel of 5-year averages over 1960-2000

• “Policy” variables (inflation, budget deficits, black-market premia, overvaluation, M2/GDP and trade/GDP) are highly significant

• He re-estimates the regression by removing observations with extreme policies (inflation and BMP > 35%, overvaluation > 68%, budget deficits/GDP > 12%, M2/GDP > 100%, trade/GDP > 120%)

• “Policy” variables no longer enter significantly, individually or collectively

• Bottom line: no reason to expect growth effects from “moderate” changes in policies

Policies versus outcomes

“Our results indicate that all 20 Latin American and Caribbean countries in our sample experienced a positive contribution from structural policies to growth…. For most reforming countries, the estimated growth contribution from improvements in … policies ranges from 2.5 to 3.0 percentage points.” (Loayza, Fajnzylber, and Calderon 2005, emphasis added)

What are these “structural policies”?– secondary school enrolment ratio (not, say, spending on education)

– private credit/GDP ratio (not, say, specific financial-market interventions)

– adjusted trade-GDP ratio (not import tariffs)

The trouble is that the regressions do not tell us which “structural polices,” if any, actually led to improvements in these outcomes.This is not a point about endogeneity, it is about the inappropriateness of

ascribing improvements in these areas directly to specific policy levers included under the conventional “structural reform agenda.”

Modeling the cross-national variation in policies and growth: an illustration (I)

Modeling the cross-national variation in policies and growth: an illustration (II)

We get a negative relationship between s and g even though policy intervention maximizes the growth rate and we would never want to restrain governments from using policy intervention s.

Modeling the cross-national variation in policies and growth: an illustration (III)

How has conventional wisdom accommodated the “anomalies”?

Non-operational truisms

“I would suggest that the rate at which countries grow is substantially determined by three things: their ability to integrate with the global economy through trade and investment; their capacity to maintain sustainable government finances and sound money; and their ability to put in place an institutional environment in which contracts can be enforced and property rights can be established. I would challenge anyone to identify a country that has done all three of these things and has not grown at a substantial rate.”

-- Larry Summers (2003) But:

• “ability” to do X and “capacity” to manage Y do not tell us what the requisite policies area• and if we try to give it operational content, it turns out that the immediate implications are not quite consistent with the evidence

“[Reforms in Latin America] were uneven and remained incomplete…. More progress was made with measures that had low up-front costs, such as privatization, relative to reforms that promised greater long-term benefits, such as improving macroeconomic and labor market institutions, and strengthening legal and judicial systems.” IMF (2005)

“The moral of the Argentine story is that being an emerging market country is not easy: not only do you have to have good institutions to support an efficient financial system, but you need good institutions to support fiscal probity. It is not enough to be good: you have to be very good.” (Mishkin 2006)

“[O]penness to trade is not by itself sufficient to promote growth—macroeconomic and political stability and other policies are needed as well….” (Panagariya 2003)

A recommendation contingent on all the requisite “complementary reforms”—many of which cannot be specified ex ante--is impractical (and defeats the purpose of reform).

How has conventional wisdom accommodated the “anomalies”?

Open-ended list of complementary reforms

Where to next?

• A taxonomy of reform strategies– taking second-best issues seriously

• How does the “binding constraints” approach relate to other strategies?

Analytics of Reform

• Suppose an economy starts out with k distortions, τ={τ1,τ2,…,τk} , with marginal social valuations of activities diverging from marginal private valuations:

μjs(X) – μj

p(X) – τj = 0 (X = vector of activities)

• Interpretations of τi:

– “tax” on investment μIs(X) – μI

p(X) > 0

– human capital externality μHs(X) – μH

p(X) > 0

– learning spillover μRs(X) – μR

p(X) > 0

– excessive logging of forests μFs(X) – μF

p(X) < 0

• Problem is we have all these (and many other) distortions and market failures simultaneously

• Important: a distortion in one market enters other margins

Analytics of Reform

• What is the effect on growth (or welfare) of removing one distortion leaving the other k-1 unchanged?

Total effect = direct effect (positive) + sum of all other indirect

effects (positive or negative)

i

j

ijj

i

ii

i

XX

d

du

Where do indirect effects come from?

• Lowering one distortion can have good or bad effects, depending on which way other distorted quantities move– Example from agriculture: export promotion – Example from trade: liberalizing intermediate inputs– Example from macro: relaxing borrowing restrictions under moral hazard– Two sort of “budget” constraints

• Fiscal revenue constraint: reducing one tax requires a rise in another• Political budget constraint: may need to compensate for loss of rents

• Key point: Undertaking partial reform while leaving other distortions in place can have large, small or even negative effects on growth and welfare

• We can envisage 5 alternative reform strategies in these circumstances

Strategy 1: Wholesale reform

• Eliminate all distortions at once• Sure to raise welfare• Problem: impossible to do

– Full list not knowable– Administrative, political, human-resource constraints

Strategy 2: Do as much as you can, as best as you can

• Implicitly assumes:– Any reform is good– The more areas reformed, the better– The deeper the reform in any area, the better

• Trouble is, none of these assumptions holds under second-best environments

Strategy 3: Sophisticated second-best reform

• Take into account all possible indirect effects• Problem: technically infeasible, even if

administratively feasible– Most of these second-best interactions are very

difficult to figure out and quantify ex ante– Some distortions (for instance lack of credibility) are

not even observable

The pitfalls of “sophisticated” reform:A Chinese counterfactual c.1978

solutionproblem

Low agricultural productivity Price liberalization

Private incentives Privatization

Fiscal revenues Tax reform

Urban wages Corporatization

Monopoly Trade liberalization

Unemployment Labor market flexibility

Enterprise restructuring Financial sector reform

And so on...

Contract enforcement Legal reform

Strategy 4: Target the largest distortion(s)

• Under certain assumptions* (unlikely to hold in practice), can be guaranteed to raise welfare

• But:– It does require us to have a complete list of

distortions– It does not guarantee that the biggest welfare

bang is achieved

* The (sufficient) condition is that the activity whose tax is being reduced be a net substitute (in general equilibrium) to all the other goods.

Strategy 5: Focus on the most binding constraints

• Pursue those reforms where direct beneficial effect is largest

• This increases likelihood that the total effect (including hard-to-measure indirect effects) will be positive and large

• Advantage: less technically demanding, more likely to result in big welfare bang

• Disadvantage: still need to check about key likely indirect effects

How China did it instead

• Pragmatic, often heterodox solutions to overcome political constraints and second-best complications – Two-track pricing insulates public finance from the provision of supply

incentives – Household responsibility system obviates the need for explicit

privatization– Township and village enterprises overcome weaknesses in legal (third-

party) enforcement of contracts– Special economic zones provide export incentives without removing

protection for state firms (and hence safeguard employment)– Federalism, “Chinese-style” generates incentives for policy competition

and institutional innovation

• Strategic and sequential approach targeting one binding constraint at a time– First agriculture, then industry, then foreign trade, now finance…

• Remaining institutional challenges– Especially with regard to political democracy and the rule of law

Chinese experimentation

Source: Heilmann (2008)

Toward a growth diagnostics

• What we need is method for identifying the most binding constraints

• This can be done in the following way– Start from the proximate determinants of growth, X(i)– Figure out which of those pose greatest impediment to higher

growth– Identify the specific distortion(s) that lie behind these

impediments (tau’s)– Come up with policies that target these distortions as closely

as possible, while bearing in mind potential interactions with distortions in other, related areas

High cost of financeLow return to economic activity

Low social returns Low appropriability

government failures

market failures

poor geography

low human capital

bad infra-structure

micro risks: property rights,

corruption, taxes

macro risks: financial,

monetary, fiscal instability

information externalities:

“self-discovery”

coordination externalities

bad international finance

bad local finance

low domestic saving

poor inter-

mediation

Growth Diagnostics

Reasons for low private investment

An empirical diagnostic framework • Explicit search for “diagnostic signals”:

– “If story A is correct, signals x, y, z must be present…”

• Direct evidence– “shadow” prices: returns to education, real interest rates, cost of

transport,… – benchmarking potentially helpful

• Indirect evidence– if a constraint binds, effects must show up in differential outcomes

for activities that differ in their intensiveness in that constraint• informality, internalization of finance, self-enforcement of contracts

– elimination of other plausible constraints

• A constant cannot explain a change– high growth episodes of the past cause us to ask what has

changed

A policy reform agenda that is consistent with a country’s recent growth history

• Episodes of high or low growth provide evidence on necessary and sufficient conditions for growth in a given setting

• Reforms under consideration must be consistent with this evidence

• What to do:– Develop internally consistent “stories” about the

causal mechanisms underlying recent growth history

Illustrations

• El Salvador: low investment demand due to low incentives for “self-discovery”– Need to find new high-return investment opportunities– Solution: industrial policy?– What will not work: Improving “institutional environment” will

not be very effective when constraint is low appropriability due to “cost discovery” and coordination externalities

• Brazil: low investment due to high cost of capital– Need to increase domestic savings and enhance access to

foreign savings– Solution: adjust fiscal policy?– What will not work: improving “business climate” not very

effective when problem does not lie with low investment demand

Step 2: Policy design

• First-best logic often not helpful– targeting policy on relevant distortion may not work due to second-

best interactions and political-economy or administrative constraints• Requires instead creative solutions that overcome these complications

– policies that can decouple complementary areas of reform often work best, even if heterodox (e.g. China)

– Taking advantage of multiplicity of institutional solutions:• the functions that good institutional arrangements perform (protect

property rights, ensure macro stability, internalize externalities, etc.) do not map into unique institutional forms

– local contingencies require local solutions• Policy solutions may lie in areas that did not appear to be the binding

constraint– E.g. may recommend saving-augmenting strategy even if an economy is not

saving constrained, if that enables a depreciated currency that increases tradables profitability

• Experimentation and learning are necessary components of reform• Implication for government-business relations

– government needs to be close enough to business to elicit information, far enough not to be captured

Step 3: Institutionalizing the diagnostic process

• Nature of binding constraints change over time

• Growth will slow down if diagnostic process not ongoing– Argentina, Indonesia, Cote d’Ivoire,..– China’s future challenges

• Sustaining growth requires ongoing institutional reform to– Maintain productive dynamism

• “industrial policy” institutions?

– Increase resilience of economy to external shocks• “institutions of conflict management”

– democracy, rule of law, social pacts, social safety nets

Arguments against GD

• Poor nations suffer from many constraints– Of course– But this is no argument against the need to prioritize

• GD is a way of thinking about how to prioritize• Diagnostic “signals” are model-specific

– Yes, the GD framework does place a premium on being explicit about the underlying model of the economy one has in mind

– But that is probably an advantage, not a disadvantage• Can never be sure you have identified binding constraint(s) correctly

– Yes– But even then, it provides a useful way of framing the debate over

different recommendations• “Since you recommend a, you must presume the binding constraint is x;

what evidence can you adduce for it?”• It is hard

– Yes!

General lessons

• Binding constraints to growth differ across countries and over time– clear evidence that growth is unlocked in a large variety of ways– different strokes for different folks: CHN was constrained by poor supply

incentives in agriculture; BRA is constrained by inadequate supply of credit, SLV by inadequate production incentives in tradables, ZAF by inadequate employment incentives in manufacturing, ZWE by poor governance …

• Relaxing binding constraints requires well-targeted reforms that are cognizant of prevailing second-best and political complications

– selectivity instead of a laundry list– pragmatism in lieu of “best practice” and rules of thumb

• Over time, strengthening institutional underpinnings is critical– institutionalizing “diagnostics”– building resilience to external shocks– institutional reform is key, but to sustain rather than ignite economic growth