• No results found

A Survey of Economic Growth

N/A
N/A
Protected

Academic year: 2022

Share "A Survey of Economic Growth "

Copied!
40
0
0

Loading.... (view fulltext now)

Full text

(1)

16/07/2002

A Survey of Economic Growth

Mark Rogers

Harris Manchester College, Oxford

Paper for Economic Record

Introduction

Economic growth is the term economists use to describe the growth in output from an economy. Some economies achieve large increases in output over extended periods of time which, in turn, dramatically changes the economic, political and social landscape. The US, for example, is estimated to produce about 30 times as much in 1999 as it did in 1899.

Increasing output alone cannot be a sensible goal for any society. A more useful measure is the amount of output per person in the economy. When output per person (or GDP per capita) is high, people have more goods and services, and this may increase societal well-being. In this survey the term ‘economic growth’ is used to refer to a situation when output per capita is increasing. Of course, society is also interested in health, inequality, education, pollution, sustainability, freedom and many other issues. The link between economic growth and a wider concept of ‘well-being’ is controversial, but this survey will leave aside this important issue and instead focus solely on economic growth.

The aim of this survey is to explain how economists try to understand the process of economic growth. To make the task manageable, the focus is on major issues and current debates. Models and conceptual frameworks are discussed in section 3. Section 4 summarises empirical studies, with a particular focus on econometric studies of groups of countries. This is not to say that case studies of single countries are not valuable, but space precludes

covering everything. The following section sets out some facts about economic growth and, hopefully, motivates the further effort needed to tackle the theory and econometrics.

(2)

Motivation

Economic growth in perspective

A short summary of economic growth in the world over the last 1000 years might be as follows. From 1000 to 1700 the limited data available suggest countries experienced low, sporadic growth. While some countries, for some periods, did grow, in general growth was slow or non-existence. England, for example, may have achieved a 0.3% per annum growth in GDP per capita from 1086 to 1688 (Snooks, 1992). The industrial revolution, starting in Britain, changed the ability of countries to raise output. In Britain, GDP per capita growth increased to 1.3% per annum in the last two decades of the 18th century, and rose still further in the 19th century (Crafts, 1996). Growth rates in other countries also increased in the 19th century, with a further acceleration in the 20th century (see Table 1). Table 1 also shows that the OECD countries fared much better than others in the 20th century.1 Within these

aggregates, however, individual countries had very different experiences. Australia, for example, is estimated to have had the world’s highest GDP per capita in 1870 (50% higher than the USA), but Australia averaged 0.9% growth in GDP per capita over 1870 to 1960.

The US growth rate of 1.7% for this period meant Americans had powered ahead by 1960 (Maddison, 1995).

Table 1 Growth of GDP per capita (average annual percentage changes)

1500-1820 1820-1900 1900-2000

OECD 1.2 2.0

Non-OECD 0.4 0.6

World 0.04 0.8 1.9

Source: Boltho and Toniolo (1999, Table 1) OECD refers to North America, Western Europe, Japan, Australia and New Zealand.

Two big questions

There are real difficulties in obtaining and interpreting data for GDP and population going back 100 or more years. For more recent periods, the various problems and measurement issues involved have been tackled by Summers and Heston (1991). The data base produced,

1 It is worth pointing out the implications of small differences in growth rates. A 0.6% growth rate will not even double GDP per capita over 100 years; a 2% growth rate will raise GDP per capita by 7.4 times.

(3)

the Penn World Tables (PWT), covers the post-1960 period and has become the most frequently used data base for empirical growth studies.2 The most recent version of this data base (PWT5.6) covers 152 countries for the period 1950 to 1992 (although gaps in series are present). Figure 1 plots the average annual GDP per worker growth3 for 1960 to 1990 against GDP per worker in 1960. The figure shows the variance in country growth rates, with poorer countries (in 1960) showing more variance. Some countries had negative average growth rates over this period, while others have done spectacularly well. The first big question, therefore, is what drives the differences between the ‘losers’ and ‘winners’, or to use even more emotive terms, the ‘miracle’ and ‘disaster’ stories?

Figure 1 Growth in GDP per worker (1960-90) vs. GDP per worker (1960)

-0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05 0.06

0 5000 10000 15000 20000 25000

While it is useful to focus on average growth over long periods (e.g. 30 years), it should not hide the fact that, on average, a country’s growth rate is not highly correlated over shorter time periods. For example, the correlation coefficient for countries’ average growth rates between the 1960s and 1970s is 0.15, between the 1970s and 1980s it is 0.16. The low

2 This is not to say there is universal agreement that the data are robust enough for all purposes. Temple (1999) discusses some of the data quality issues.

3 The growth rates is defined as b in a regression of lny = a + bt. There are 121 countries in PWT 5.6 with at least 30 years of real GDP per worker data.

(4)

persistence in growth suggests that random shocks may be important, or that the determinants of growth themselves have low persistence (Easterly et al, 1993). Behind these aggregate correlations, however, is a huge difference between OECD and non-OECD countries. The poorer, non-OECD countries exhibit large intertemporal variation in growth rates. For example, Brazil achieved 4.2% per annum growth in GDP per capita from 1965-80, but then growth fell to –0.2% in 1980-92; Cote d’Ivoire averaged 3.1% from 1960-80 then –4.1%

from 1980-92 (Pritchett, 1998). OECD countries do experience changes in growth over time – witness the extensive literature on the growth slowdown in the 1970s and 1980s (Maddison, 1987) – but growth rates are much more persistent than developing countries. The second big question is therefore: why do the growth rates of some countries, or groups of countries, vary so much over time?

The convergence debate

The two big questions can be framed in a different way, one that focuses our interest on what is, perhaps, one of the most important issues in economics. This concerns whether countries have been converging to a common level of GDP per capita, and is known as the

‘convergence’ debate. One approach to this issue is to analyse how the variance of the distribution of GDP per worker across countries has changed over time – this is known as testing for σ-convergence. For the countries in Figure 1, the variance in GDP per worker increases between 1960 and 19904. Lant Pritchett (1997) considers the period from 1870 and estimates that the gap between rich and poor countries has risen five fold. The title of

Pritchett’s paper sums up a worrying trend : “Divergence, Big Time”. However, using the variance, or the standard deviation, as summary statistics might hide more complex patterns.

Various analyses for the 1960 to 1990 period have found that the world distribution of GDP per worker has become bimodal (i.e. “twin peaks”) with, perhaps, each country a member of either a poor or rich ‘convergence club’ (Quah, 1996, Galor, 1996a, Kremer et al, 2001).

From this perspective, the poor growth performance of many African and Latin American countries appear to form a ‘club’, whereas OECD countries plus a few success stories form another.

4 In fact, there are some years when the variance falls, but overall the trend is rising from a var(lnGDP p.w.) of 0.89 in 1960 to 1.18 in 1990.

(5)

Another method of approaching the convergence question is to use regression analysis to test whether poorer countries are growing faster. This is done by regressing average annual growth rates of GDP per capita or worker (gi), say from 1960-90, on the log of initial GDP per capita of worker (yio) (i.e. gi = a + βyio). If β<0 then there would be evidence of poorer countries growing faster.5 In fact, for the full sample, there is no evidence of “β-

convergence” (i.e. the regression coefficient β is not negative). Additional analysis has then included other explanatory variables, for example investment or trade openness, in the regression to control for other aspects of a country’s circumstances. When this is done the results show that β<0 and this is called conditional convergence: poorer countries grow faster conditional on other factors being present. This might be exactly what intuition would

suggest. The question is, of course, what are these other factors and can policy affect them?

Models and frameworks

Introduction

The aim of this section is to outline how modern economists think about the process of economic growth. The starting point is a consideration of the neoclassical growth model and

‘new’, or endogenous, growth theory. 6 In both cases the focus is on the main ideas rather than a detailed summary or critique. The third sub-section considers how the models change when the issue of international openness is introduced. Following these three sub-sections, which contain what might be called ‘formal’ models, two sub-sections widen the discussion by considering fundamental factors: institutions, society and geography; and technology, firms and competition. A final sub-section considers multiple equilibria models.

5 You might think that β<0 would also mean more equality in world incomes (i.e. σ-convergence), but statistically this is not the case: σ-convergence does imply β-convergence, but not vice versa. This is sometimes referred to as Galton’s fallacy and requires a few equations to illustrate (see for example, Valdes, 1999, p.49)

6 There is, of course, a long history of ideas on economic growth before these models. See Eltis (2000) for a discussion of the classical economists – Smith, Malthus, Marx and Ricardo. Hahn and Mathews (1964) provide a well known early survey, recent comprehensive books that review thinking on economic growth include Rostow (1990), Scherer (1999) and Ruttan (2001).

(6)

Neoclassical models

The Solow (1956) and Swan (1956) models are based on an aggregate production function of the form

Y = Af K L( , ) , [1]

with output (Y) depending on capital (K), labour (L) and technology (A). Other inputs, such as raw materials or energy, are abstracted from in the basic models.7 Assuming f(.)exhibits constant returns to scale, [1] can be re-written as,

Y

L = =y Af k( , )1 = Af k( ) , [2]

where k equals the capital to labour ratio. If we assume growth in L and A are zero, growth in output will only occur if there is capital accumulation. Gross capital accumulation is assumed to be equivalent to total savings (sY), where s is the fraction of income saved. Net investment is given by

dK

dt sY K or dK

dt K sY

= −δ / = Kδ , [3]

where δ is depreciation. This equation implies that capital accumulation (and output growth) must stop if the average (marginal) product of capital declines without limit.8 The level of output per worker when (net) capital accumulation stops is called the ‘steady state’. As might be expected, as an economy approaches its steady state its rate of growth will decline.

If there is labour growth the outcome of the model is that output grows in proportion, hence output per worker is constant. If there is technology growth, this transmits directly to a positive, and equal, level of output per worker growth. Underlying this is the fact that an increase in A raises the marginal product of capital, which can be thought of as maintaining

7 Hence such neoclassical models do not apply to primary resource dominated economies such as OPEC countries. As Lucas (1988, p7) notes these models were originally created to understand the US economy.

8 The average product of capital may not fall to this level if the elasticity of factor substitution is greater than one. Pitchford (1960) provides a discussion of this, Jones and Manuelli (1990) provide a more modern (endogenous growth) viewpoint.

(7)

the incentive to invest.9 A criticism of the neoclassical model is that it leaves technology growth as an exogenous factor (i.e. external to the main variables of the model). Without technology growth the model asserts that economic growth will, ultimately, cease.10

Endogenous growth models

In 1986 Paul Romer published a seminal paper entitled ‘Increasing Returns and Long Run Growth’. This paper provided a model that yielded positive, long run growth rates without assuming exogenous technical change. Instead, Romer modelled technology growth (he termed it ‘knowledge’ growth) as the outcome of competitive firms that invested in

knowledge generation. The central idea that allowed this was that while individual firms face diminishing returns to investing in knowledge, at the societal level returns to knowledge can be increasing. The basic ideas are contained in a seminal paper by Arrow (1962) (see Solow, 1997 for a recent discussion of this paper) which assumes that the stock of knowledge (A) is a function of the entire capital stock of the economy (i.e. A=Kφ). Consider the aggregate production function,

Y =Kα(AL)β =Kα([Kφ] )L β =Kα φβ β+ L . [6]

If α+βφ equals 1, which implies φ=1, the economy- wide marginal product of capital will be constant, implying the incentive to accumulate capital may always be present. Note also that the production function in [6] exhibits increasing returns to scale (if φ>0). Individual firms, however, still view their own production function as having constant returns and diminishing returns to capital. The fact that knowledge can spillover (i.e. positive externalities) is at the centre of the growth process. Romer (1986) develops these ideas into a competitive

9 In the text there is no explicit decision to invest since saving is exogenous. However, saving can be made endogenous by modeling forward looking consumers with a time preference (ρ), this results in a condition for positive growth where the marginal product of capital must be greater than ρ+δ (see Aghion and Howitt, 1998, or other text books).

10 It should be noted that the original authors realized this shortcoming. Swan (1956, p.338), for example, states

“To this anti-accumulation [of capital], pro-technology line of argument there are at least two possible answers.

First, the rate of technical progress may not be independent of the rate of accumulation, or … accumulation may give rise to external economies, so that the true social yield of capital is greater than ... private experience.

Second, the rate of growth of labour may not be independent of the rate of accumulation”.

(8)

equilibrium model which yields long run positive growth. The model also suggests that a) the competitive growth rate is below the socially optimal level (due to the presence of knowledge externalities), b) large countries may grow faster (a scale effect), and c) shocks to a country’s growth may have permanent effects.

Romer’s paper gave rise to an upsur ge in interest in endogenous growth. Many authors have modified and tested Romer’s findings, producing a huge – and technical – literature.11 Important modifications and re-interpretations have included the role of human capital (Lucas, 1988) and imperfect competition and creative destruction (e.g. Romer, 1990, Grossman and Helpman, 1991, Aghion and Howitt, 1992). To gain insight, consider the equation that describes the accumulation of human capital (h) in Lucas (1988)

h h G v h

h G v

= ξ ( ) or = ( ) if = 1 . ξ [7]

In this equation v is the share of economy wide effort devoted to human capital creation (education and training) and we assume dG/dv>0. Human capital is then used in the production of goods. Long run economic growth is possible if v>0 and ξ=1, as can be inferred from the far right of equation [7].

The ‘knife edge’ property of the ξ=1 assumption is common to many endogenous growth models. Romer (1990) has a similar equation for the accumulation of “designs” that are used to produce new goods (this is dA/dt=δhaA, where ha is the human capital devoted to ‘R&D’

activity, with the implicit exponent on A equal to 1). As Romer notes, it is the specific functional forms of these accumulation equations that make long run growth possible. This issue is focused on by Jones (1995) who notes that these types of model contain a ‘scale effect’ (e.g. higher levels of human capital in R&D activities increases the growth rate, hence larger economies should have faster growth rates). Jones notes that the data on scientists and engineers in R&D shows a five fold rise in US (1950-88), while growth rates have clearly not increased five fold. Scale effects can be removed by alternative modelling strategies, for example, Jones assumes ξ<1 which yields a steady state level of growth equal to n/(1-ξ), where n is population growth (see also Jones, 1999). The importance of this discussion is that

11 Reviews are contained in Romer (1994), Pack (1994), Jones and Manuelli (1997) and Aghion and Howitt (1998) amongst others.

(9)

the policy implications between the two models are sharp. In Romer’s version any subsidy to R&D can raise long run growth; in Jones’ a subsidy will have no effect in the long run (although it can raise growth rates in the short term).

Part of the endogenous growth achievement is to model imperfect competition (i.e. firms have some market power and set price above marginal cost). Technically, this is done by modelling an economy of symmetric firms that each have the monopoly right to a distinct good. The goods come into existence through the process of innovation or R&D, with the incentive to undertake R&D derived from the expected future monopoly profits. In

equilibrium, competition between firms equates the expected monopoly profits to the fixed costs of innovating. In some models new goods are invented continuously (product variety models), while in others new goods displace older versions (creative destruction or product ladder models) (see Grossman and Helpman, 1991, for a comprehensive treatment of these models ).

The existence of imperfect competition implies welfare may be sub-optimal. In particular, would a social planner would prefer more or less resources devoted to innovation? In general, the social benefit of an innovation cannot be fully captured by a monopolist – implying too little private investment in innovation (draw a demand curve for a new good; the monopolist cannot extract all the consumer surplus unless it can perfectly price discriminate). Profit seeking firms, however, will not take into consideration the reduction of other firms’ profits (creative destruction), suggesting too much innovation. The models also contain knowledge externalities, suggesting firms under invest since they do not appropriate the full benefits.

The overall outcome of these three market failures depends on the functional forms in the model, with the general presumption that the presence of externalities will tip the balance to (private) under investment and the need for intervention. But the models are too general and stylised for any robust conclusions on this issue.

International openness and catch-up

The models described above are closed economy models. A major question, fundamental to the convergence debate, concerns how the process of economic growth is affected as a country becomes more open to world trade, knowledge flows and financial flows.

Consider a country that becomes more open to trade alone (i.e. knowledge and financial flows are absent), various static international trade theories analyse how consumption, production and factor prices may change. For example, if the factor price equalisation (FPE)

(10)

theory holds then convergence in factor prices, and wages in particular, suggests convergence in GDP per capita. However, FPE depends on the nature of tastes, technology and

endowments in specific countries. Equally, even if factor prices are equalised, movements in the stock of factors may cause divergence in income per head. In general, international trade theory suggests increases in trade have ambiguous effects on factor prices and rates of factor accumulation (see Slaughter, 1997). A neoclassical perspective had led some to argue that international trade can be a method of postponing diminishing returns to capital since an open economy is exposed to diminishing returns at the global, not local, level (Ventura, 1997).

However, the neoclassical model stressed technical change as the ultimate source of long run growth. How might increased trade openness alone impact on technology accumulation? One route may be in shifting resources into, or out of, the R&D sector depending on the country’s comparative advantage. This may mean that growth slows in some countries and rises in others, although the overall welfare outcome is more complex (Grossman and Helpman, 1991).

Young (1991) uses a model where knowledge spillovers are solely national. His paper assumes that all products experience learning by doing, which improves productivity, at a rate dependent on economy wide activity. The model considers two economies, an advanced economy, which has higher productivity (more learning has occurred), and a less advanced economy. Young finds that free trade will not diminish the growth rate of the advanced economy but cannot increase the growth rate of the less advanced (i.e. free trade might cause growth rates to diverge). Underlying this is the fact that trade causes a reallocation of

productive activity across products which affect the aggregate rate of learning. Note that real consumption growth may not reflect output growth, hence the welfare effects are unclear.

Obviously, a key assumption is the national boundary of spillovers, but the general idea of trade shifting the pattern of production, perhaps into low growth products is universal. These issues have similarities with the ‘infant industry’ argument, which suggests industrial policy should shelter some industries while they learn to be competitive.

Grossman and Helpman (1991) argue that considering only trade flows is unrealistic and that knowledge flows should be considered. It should be noted that trade and knowledge flows are related, hence distinguishing between the effects of each is problematic (e.g. knowledge might be supplied with the purchase of capital goods, or obtained by reverse engineering, or via personal contacts and observation through trade visits). Grossman and Helpman (1991) analyse how the incentives to conduct R&D change when a country becomes open to

(11)

knowledge flows. They stress that the international flow of knowledge will raise the

productivity of R&D and, overall, raise world growth rates (and welfare levels). There is also the possibility that international flows of knowledge will reduce duplication of R&D. Of course, individual countries may experience a decline in R&D and innovation dependent on their comparative advantage in R&D productivity.

Although Grossman and Helpman’s use endogenous growth models to illustrate their ideas, the concept of international knowledge and technology flows has a long history.

Gerschenkron (1962), Abramovitz (1986) and others have argued that technological catch-up by follower countries can be an important determinant of economic growth. Since all

countries have a technology gap with other countries in at least some industries, this is an important issue even for developed countries. Catching- up is, however, not always

straightforward. A country needs to have appropriate human capital, along with conducive institutions, incentives and policies. Abramovitz (1986) used the term ‘social capability’ to describe these factors, although a more descriptive term would be ‘absorptive capability’.

These ideas suggest that technology growth (dA/dt) could be written dA dt

A/ d GAP

(.) (.)

= +φ [8]

where d(.) represents a function that determines domestic generation of technology while φ(.) represents absorptive capability and GAP is the technology gap with rest of world. While the process of catch-up will affect growth rates in the short and medium term, the dynamics of an equation like [8] mean that in the long run all countries will grow at a common rate. This common rate – in effect the composite d(.) function for the world – must be determined by something, hence we are back to endogenous growth models again.

Finally, there is the issue of greater openness to financial flows. Two broad types may be distinguished: long term foreign direct investment (FDI) and short term capital flows relating to shares or loans. Foreign direct investment is generally viewed as conducive to growth, although there are arguments that FDI can cause ‘enclave economies’ that detract from wider development in the host economy (Rodriguez-Clare, 1996). Attracting FDI depends on other characteristics of the economy with, for example, national differences in political stability, corruption, taxation and profit repatriation rules, and human capital levels all being relevant.

With respect to short term capital flows, or international capital market liberalisation, there is considerable controversy over the impact on economic growth. In theory, openness to short

(12)

term capital flows should relax any constraints on investment and may act as a spur to create a competitive environment. In practice, there are real concerns over capital market

liberalisation increasing instability and thereby reducing investment, as well as leading to capital outflows rather than inflows in many poorer countries. The recent financial crises in East Asia, Russia and Latin America has focused attention on these issues and led some authors to argue for intervention in capital markets (Stiglitz, 2000). From a long run growth perspective a key issue is whether short run crises, and the associated falls in output, are permanent in any way. Empirical evidence suggests that output volatility reduces long run average growth rates (Ramey and Ramey, 1995).

Institutions, society and resources

A counterbalance to the formal, stylised growth models is provided by economists’ research into institutions.12 Rodrik (2000) highlights five key institutions: property rights, regulatory institutions, institutions for macroeconomic stability, institutions for social insurance, and institutions of conflict management. Two further points are stressed. First, there is no universal ‘blueprint’ for the optimal institutional design: institutions should vary with

national or local characteristics. Second, obtaining good institutions, in Rodrik’s view is best achieved by ‘participatory political institutions’ (a ‘meta- institution’). Institutions underpin market activity and incentives; without them opportunistic behaviour (“cheating”) is likely to deter production and investment (Bardhan, 2000). Recently, a related set of ideas comes under the heading ‘social capital’, which can be defined as “features of social organisations, such as trust, norms and networks, that can improve the efficiency of society by facilitating co-ordinating actions” (Putnam, 1993). An underlying theme is that high levels of social capital are associated with higher levels of GDP per capita and, possibly, promote economic growth. However, Olson (1982) argued that some ‘groups’ – such as unions, professional associations and cartels – can be detrimental to economic growth. The argument here is that powerful groups inhibit change. Olson argues that the slow evolution of powerful groups, if left unchecked, can ultimately slow economic growth. Delving further into ultimate causes, the economic historian David Landes considers culture – defined as inner values and attitudes

12 North (1993, p.3) defines “Institutions are the rules of the game in a society or, more formally, are the humanly devised constraints that shape human interaction”.

(13)

that guide a population – as the fundamental institution which drives economic wealth creation (Landes, 1999).

A related issue that has recently attracted considerable interest concerns inequality. Authors have pointed to four broad categories of potential links between inequality and growth. First, if credit markets are imperfect, the distribution of assets will affect the level and composition of investment. For example, poor people may not be able to borrow money to invest in education, suggesting a negative inequality-growth link. However, if some investment projects have large fixed costs, and there are credit market imperfections, an unequal distribution of assets may allow some (wealthy) agents to undertake these investment s.

Second, political economy models stress that high inequality may lead voters to choose policies that favour redistribution. Redistribution may create distortions in the economy which may, in turn, reduce growth. Third, inequality can be associated with crime, riots, disputes and alike, which reduce productive activities and raise uncertainty. This may affect growth through reduced accumulation of capital or technology. Lastly, some authors suggest that inequality and savings are linked, Keynes for example suggested that saving increases with the level of income. These points indicate that the effect of inequality on growth could be positive or negative. Aghion et al (1999) provide a summary of these issues and discuss the reverse link: growth can increase inequality depending on the nature of growth. This, in turn, links with an older tradition summarised by the Kuznets hypothesis, which proposed an inverted U-shape relationship between inequality and development.13

It has been argued that resource abundant economies tend to grow less rapidly than resource scarce economies. Various theories have been used to explain this observation (Sachs and Warner, 1995). For example, a natural resource boom can raise the exchange rate which tends to shift resources out of the tradable (predominantly manufacturing) sector. This so-called

‘Dutch disease’ is growth reducing if the tradable sector has innately higher growth potential, or if it has a positive effect on economy-wide growth through various linkages. A related view is that world demand for primary products will always grow less rapidly than

13 In short, agrarian societies tend to be more equal, as industrialization occurs inequality increases but, as the economy adjusts to these changes, inequality falls. More recent authors have considered the impact of so-called

‘general purpose technologies’ (GPTs) – for examp le, the steam engine, electric motors, or the laser – which tend to worsen inequality during the adjustment period (Helpman, 1998).

(14)

manufactured goods (the Prebisch-Singer hypothesis). Another argument is that abundant natural resources result in more rent seeking behaviour, which can damage business and government through corruption and diversion of human capital. Similarly, some argue that government policy can become mis- focused and unmotivated in tackling (difficult) growth- promoting policies. For example, there is evidence that resource abundant countries invest less in human capital (Gylfason, 2001). The underlying message is that various asymmetries or imbalances in an economy can affect policies, incentives and growth rates. The extreme cases of some oil or mineral rich economies provide clear examples, but the underlying issues are likely to be present in many economies.

Growth in GDP per capita can be broken down into growth in GDP and growth in population.

So far the general assumption has been that population growth is exogenous. More

realistically, levels of fertility and mortality are likely to be linked to economic factors and the growth process itself. Since 1800 the industrialising countries have broken the Malthusian

‘trap’, initially by raising GDP growth above the (increasing) rates of population growth, but then by reducing population growth as output growth accelerated. Underlying these broad trends are a number of potential mechanisms. For example, societies may undergo a switch from investing in the quantity of children to investing in quality or human capital (Becker, 1990, Becker et al, 1999). Underlying such a substitution may be changes in technology or specialisation that raise the return to human capital. Increasing investment in human capital may, in turn, feedback into higher rate of technology growth (Galor and Weil, 2000). The density of population, in particular the presence of cities, has also been linked to economic growth. High density (urban) populations may allow higher GDP per capita by enabling specialisation, reducing transport costs, improving labour mobility or speeding up knowledge diffusion (Ciccone and Hall, 1996). High density may also raise the rate of technology

creation as ideas, knowledge and human capital flow more easily between firms. Finally, some authors have linked population size itself to technological change (because ideas are non-rival and more people implies more idea generators). This is the ‘scale effect’ argument present in endogenous growth models, a feature that has been criticised in the recent past (see above), but appears to have some validity over the very long run (Kremer, 1993).

Technology, firms and competition

Both neoclassical and endogenous growth models place large emphasis on the accumulation of technology or (productive) knowledge. In the neoclassical model technical progress is

(15)

exogenous, which might be taken to imply that it is largely the result of scientific curiosity and, perhaps, publicly funded research (both outside the market system, but potentially testable empirically). Technological catch- up models stress that follower countries – if other conditions are met – can rapidly assimilate superior technology from overseas. Endogenous growth theorists emphasise that profit seeking firms generate techno logy through R&D and innovation, the implication being that incentives are paramount. Incentives, in turn, depend on property rights and competitive conditions. Additionally, these models draw attention to possible market failures from knowledge spillovers, monopoly power, and creative

destruction.

The endogenous growth models, however, present a highly stylised view of identical firms pursuing well-defined profitable opportunities. Studies on the history and nature of

technological change provide a different perspective (Landes, 1969, Mokyr, 1990, Cardwell, 1994, Rosenberg, 1982, 1994). Major issues highlighted by these studies include: the path dependency of technical change (i.e. history matters); the freedom to travel, experiment, and change jobs; receptivity to new ideas, which in turn derives from cultural and religious norms; the diffusion of knowledge, often fostered by associations and mobility; a skilled labour force to ease uptake of new technologies; and capable scientists and engineers interested in tackling applied problems. The ability to accumulate technology, therefore, is embedded in the historical, cultural, social and institutional system of the country. Ultimately, however, implementation of technology must occur at the firm- level, a view that the

endogenous growth theorists share with some economic historians (in particular, some stress the role of large firms in the 20th century, Chandler and Hikono, 1997).

Some of these issues can be explored by considering the rapid growth of the Asian NICs (S.

Korea, Taiwan, Hong Kong and Singapore). In explaining these ‘miracles’ some argue that rapid technological progress, centred in private firms although supported by government policies, drove the high rates of growth (Amsden, 1989, Stiglitz, 1994, Rodrik, 1995). Others have stressed that the accumulation of capital and labour, combined with relatively free markets and trade, drove the high rates of growth (Krugman, 1994, Young, 1995).

Specifically, the study by Alwyn Young estimates TFP growth rates for the NICs, finding that their growth of TFP – which is sometimes equated with technological change – is similar

(16)

to many other countries.14 The implication is that accumulation of capital, and increases in participation rates and labour force quality, drove the high rates of growth; implying that technology accumulation, and the role of firms in this process, were not atypical. Nelson and Pack (1999) provide a critique of this viewpoint arguing that TFP cannot be used as an

indicator of the importance of technology and learning more generally. They use a model that stresses the role of firms, specifically, the levels of entrepreneurship and effectiveness (in implementing new technology). They argue central attention should be given to how firms responded aggressively to innovative opportunities, which was partly due to policy incentives (e.g. export subsidies and low cost loans) being tied to achievement and favourable human capital provision. It should be clear that both the ‘accumulation’ view and the ‘technology or learning’ view have a central role for investment, but the different views on the role of technology and firms, hence yielding very different implications for policy. A literature that reflects the latter view is called the ‘national systems of innovation’ approach (Lund vall, 1992, Nelson, 1993). This approach stresses the role of firms – both in their internal

organisation and the nature of interfirm networks – and how they learn to be more productive through innovation.

An important factor that affects firms’ decisions is the nature of competition. Standard endogenous growth models imply that more competition must reduce profits and lower the incentive to innovate – something which sits uncomfortably with economists’ general outlook. If the models are altered to assume, for example, managerial ‘slack’ or low effort levels, then it is possible that increased competition will raise innovation rates (Aghion et al, 1997). An alternative approach, which implies a non-monotonic relationship between growth and competition, is to link the level of competition with investment in human capital in an imperfect labour market (Duranton, 2000). Microeconomic theory also has ambiguous results concerning the role of competition on productivity and innovation (Boone, 2001), although many would argue that some competition is vital (Nickell, 1996). This issue is also related to trade openness, as more imports may raise competition faced by domestic firms. The process of competition is linked to the ideas of the entry and exit of firms, and structural change more generally – issues that provide motivation for the ‘creative destruction’ or Schumpeterian

14 For a detailed discussion of TFP in Asian countries see Dowling and Summers (1998). They note that the absolute levels of TFP growth in some Asian countries are higher than developed countries; however, as a proportion of the overall (high) rates of growth the contribution from TFP is relatively small.

(17)

endogenous growth models. Some argue, as did Schumpeter, that entrepreneurial activity is the basic force driving the growth process. Baumol (1990) notes that entrepreneurial activity can be diverted to productive or unproductive activities, with the latter including crime, tax evasion and speculation.

Poverty traps and multiple equilibria

All of the models and frameworks discussed above imply that different economies may have different GDP per capita levels or growth rates depending on their characteristics. Other growth models suggests that a single economy may have multiple equilibria. Solow (1956) notes that non-standard functional forms, and non- monotonic links between labour force growth and income per head, could result in multiple equilibria. This possibility has been highlighted by Galor (1996) and Deardorff (2001). The latter augments the basic Solow- Swan model with international trade in multiple goods finding that, if factor price

equalisation is not present, a poverty trap may occur. Another well-known example is the

“big push” model, originally attrib uted to Rosenstein-Rodan (1943) and formalised by Murphy et al (1989). These types of models suggest that coordination or policy may be required to guide many firms or sectors into industrialising at the same time (i.e. individual firms perceive no profit from industrialising but, if all did so, profits would rise due to external effects between firms). The central idea can be present in other circumstances. For example, high GDP per capita can lead to a more competitive and sophisticated financial sector which further boosts GDP per capita (Berthelemy and Varoudakis, 1996).

Alternatively, income levels may be inversely linked to fertility rates and this may lead to multiple equilibria (Galor and Weil, 1996).

The previous paragraph considers multiple equilibria in levels of GDP per capita. It is also possible to have multiple equilibria in economic growth rates. In some ways these models reflect the basic idea of economists like Rostow (1971) in his book The Stages of Economic Growth: countries may languish in low income, low growth outcomes, then some may experience take-off. At the centre of such models is a feedback mechanism between economic growth and an underlying cause of growth. Suppose, for example, that the incentive to be an entrepreneur depends on the rate of economic growth: high rates of economic growth imply rapid change, in which high entrepreneurs are especially valuable, ensuring more entrepreneurs who, in turn, drive growth. A low growth economy, however, provides few entrepreneurs, which reinforces low growth (Hassler and Rodriguez, 2000).

(18)

Other models focus on human capital more broadly and suggest various feedback and threshold effects that could support underdevelopment or poverty traps.15

Empirical Evidence

Introduction

This section focuses on cross-country empirical research where the dependent variable is the rate of economic growth. The aim is to summarise some of the major findings along with some recent controversies. Most empirical models are not based on a specific theoretical model. Instead, the empirical specifications are somewhat ad hoc, with motivation for the variables included being taken from a wide range of sources. This is less than ideal, but the complexity of economic growth and the lack of an encompassing model make it a necessity.

There are a host of econometric and data issues involved in empirical growth analysis, here there is only space to highlight four issues (see Temple, 1999a, Durlauf and Quah, 1999, Brock and Durlauff, 2000, for more discussion).

First, in most cases there are difficulties in obtaining data on the real variables of interest.

Sometimes this means using proxies that do not have clear interpretations. It also means that measurement error, which can bias coefficients to zero, is a concern. Second, the results of any specific regression may be influenced by a relatively few, perhaps unusual, countries or time periods (e.g. OPEC (oil exporting) countries, or city states like Singapore and Hong Kong). The fact that results are driven by ‘influential observations’ is not, in itself, a problem, since interesting insights may result. However, the presence of influential observations

implies great care in interpreting results or selecting regression samples. Another way of describing this issue is to say that parameter heterogeneity, across countries or sub-samples of countries, is likely to be a problem. Third, there are hundreds of possible explanatory

variables that could be included in a growth regression. Since there are limited numbers of countries and time periods this necessitates some a priori selection procedure. Economists

15 See de la Croix (2001) and Eicher and Garcia-Penalosa (2001) for some recent examples and short literature reviews.

(19)

would normally appeal to theory at such a point, but the theory is equally diverse. 16 Lastly, there are considerable problems in interpreting regression coefficients. Causality is always an issue: econometric results give insights into associations between variables, including that a variable in period t is associated with growth in period t+1. Moreover, since groups of explanatory variables are often highly correlated, requiring only a sub-set to be entered as explanatory variables because of multicollinearity, there is a difficulty in interpreting the coefficient(s) as relating only to one variable.

Investment and finance

How does physical capital accumulation affect economic growth? De Long and Summers (1991, 1993) provide cross country regression evidence that high levels of physical

investment, and equipment investment in particular, for the period 1960-85 are linked to high levels of GDP per worker growth over this period (88 countries). A first issue is causality:

does investment cause growth, or does rapid growth drive investment. Their analysis uses instrumental variables to suggest investment drives growth. In contrast, Blomstrom et al (1996) argue that growth causes investment (by using a Granger-Sims causality framework and data on 5 year periods). More recently, Podrecca and Carmeci (2001), using more sophisticated econometrics, again find no evidence that increasing investment leads to

16 Two main empirical approaches have been suggested to this problem. Levine and Zervos (1992) use a form of Leamer’s extreme bounds analysis. In this a core set of four explanatory variables is always used in a cross- section regression of growth. Each variable is then tested in a series of regressions and the (statistical) lower and upper bounds on its coefficients are found (the series is defined by all combinations of three other explanatory variables from a total of around 50). Sala -i-Martin (1997) tackles this issue in a slightly different way in a paper entitled ‘I just ran two million regressions’. The regression number is from using 62 possible explanatory variables, with 3 core explanators, and all others in combinations of 4. The distributions of estimated coefficients are then analysed, rather than upper and lower bounds. The distributions are estimated in various ways, including weighting coefficients proportional to the regression model’s likelihood. Dopplehofer et al (2000) provide an alternative, but similar, method based on a Bayesian approach. These approaches are useful, but they still limit the search, for example, there are over 3 billion combinations if we allow 7 explanatory variables in any comb ination. Many more combinations are present if interaction terms or log transforms are added. There are also more than 62 possible explanators.

(20)

increased growth.17 Interpreting these findings requires some care. As might be expected, there is evidence that investme nt and growth are closely linked, and policies that hinder investment may well reduce growth. However, policies aimed solely at boosting physical investment in order to raise growth may be ineffective. Empirical evidence from Africa highlights these and other issues. Higher private investment rates in African countries show little association with higher growth (Botswana is an exception), suggesting aggregate measures hide important differences in quality and composition.

Another way of considering the issue is to say that physical investment is a proximate not ultimate cause of economic growth. Underlying high, effective rates of investment will be other more fundamental factors. One of these may be the nature of the financial system, something investigated by King and Levine (1993a,b). They argue that the ability of financial systems to evaluate investment projects, mobilise savings and channel these into the most productive investments is critical. Testing this requires data on the extent of financial

development, which they proxy by various measures of banking development. They find that higher levels of financial development are associated with higher subsequent growth rates in cross-sectional regressions. They do note, however, that their aggregate measures omit non- bank institutions which may be important (they cite S. Korea in the 1980s as such a case).

The general thrust of these results has been supported by more recent studies that use improved data sets, refined measures of financial development and different econometric methods (Beck et al, 2000; Neusser et al, 1998, and Rousseay, 1998, for OECD countries).

There is some evidence that financial development has its largest effect directly on productivity (TFP) growth, rather than on capital accumulation and savings rates. One interpretation of this result is that financial development acts in a Schumpeterian manner, allocating funds to more efficient firms and boosting creative destruction. Lastly, stock market development has also been found to be positively associated with economic growth, with the particular aspect being liquidity, rather than absolute size (Levine and Zervos, 1998, Rousseau and Wachtel, 2000).

17 In fact they find that increasing the investment to GDP (I/GDP) ratio leads to a fall in growth in subsequent periods. To explain this they argue that in Solow-Swan steady state a rise in I/GDP will have a contemporaneous increase in growth of GDP p.w., but in subsequent periods the growth rate will decline.

(21)

Human capital

Human capital can be thought of as affecting economic growth in two ways. First, if human capital (H) is a factor of production, e.g. Y = f(A, K, H, L), changes in H will be correlated with changes in Y (growth). For example, workers with higher levels of education or skills should, ceteris paribus, be more productive. Second, the level of human capital may affect the rate of accumulation of other factors. For example, Romer (1990) assumes that the growth of knowledge or technology (A) depends on the level of H. This appeals to the idea that more educated and skilled people are more inventive and innovative. Higher levels of human capital may also encourage capital accumulation, or may raise the rate of technological catch- up for follower countries (Nelson and Phelps, 1966).

Empirically testing the role of human capital requires its measurement and this turns out to be difficult. An obvious place to start is schooling data. A number of cross-country data sets use enrolment and other data to estimate the average years of schooling in the working population (e.g. Barro and Lee, 1996, Kryriacou, 1991). These data can then be used in regression

analysis using either the change or level of schooling as explanatory variables. Benhabib and Spiegel (1994) find that changes in schooling capital are unrelated to growth, suggesting the first mechanism mentioned above is not strong, at least for their 78 country sample.18 They then test whether the level of schooling capital is related to technology growth. This could be a direct effect or through technological catch-up (suggesting that the empirical specification should have an interaction term between schooling capital and the technology gap). Their results suggest human capital raises growth through the technological catch- up mechanism for all countries, but also raises growth directly in the wealthiest countries. A drawback of this view is that it implies increasing the level of education can raise growth rates without bound, something that, in the limit, appears difficult to accept. Their study also shows that level of human capital is positively correlated with investment rates. Hence they find that the level of human capital is pivotal in allowing rapid accumulation of technology and physical capital.

The result that changes in average schooling per worker itself is unrelated to growth ha s, however, been challenged on a number of grounds. First, measurement error in the schooling

18 Other studies of individual advanced countries using the ‘growth accounting’ approach have found increases in schooling capital to be important (e.g. Jorgenson et al, 1987, Maddison, 1991).

(22)

data, which is aggravated when changes in schooling are used, could bias the results to finding no association (Krueger and Lindahl, 2000).19 Improving the measurement of schooling might show that changes in schooling are important, something with empirical support for OECD countries (de la Fuente and Domenech, 2001). Second, human and

physical capital accumulation are related and, if you omit the change in physical capital from the regression, the coefficient on the change in human capital becomes significant. Third, the form of the production function can lead to different results (Temple, 2001). Fourth, the insignificant results may be driven by relatively few influential observations, put another way the ‘impact’ of education varies across countries (Temple, 1999b). Overall, however, there are still those that argue that the dramatic rise in schooling in developing countries (e.g.

secondary schooling enrolment rates have risen from 14% in 1960 to around 40% today) has had little impact on economic growth (Pritchett, 2001).

Researchers have also used the results from international test scores on the mathematical, scientific and reading ability of students as a means of ‘quality adjusting’ the crude schooling data. Results suggest that mathematics and science scores have a significant partial

correlation with growth (Hanushek and Kimko, 2000), although reading scores appear to show no significant association (Barro, 2001). Other results suggest that primary schooling and higher education appear to have less association with growth than secondary schooling, and that female education has a less stable association with growth than male (Barro and Sala-i-Martin, 1995, Barro, 1999). These results, however, should be interpreted with extreme caution. For example, primary schooling is a prerequisite for secondary schooling.

Equally, the coefficient on female schooling can become significant if the fertility rate is omitted from the regression (i.e. more female schooling tends to reduce fertility rates which, on average, raises growth rates).20

19 For the econometrically minded, a random measurement error in an explanatory variable attenuates coefficients in a standard OLS regression. Using the first difference, or within transformation, as the explanatory variable worsens the situation if the levels of the explanatory variable are correlated over time (Johnston and DiNardo, 1997, p. 399). Wolff (2000), however, notes that if measurement error is constant over time within each country (i.e. a permanent, non random difference across countries) first differences or within transformation removes the measurement error.

20 There is also a large development economics literature on the importance of female education and its links to child health, productivity and governance (see Stern, 2001, for a recent statement).

(23)

A strong message from the literature is that there is great difficulty in measuring human capital. The issues of estimating stocks of schooling in the population and adjusting for quality make it clear that data issues can be critical. But the issue is even wider than this since

‘human capital’ should really be the capital used for productive activity. One interpretation of the poor result s on female schooling is that many countries do not utilise the (female) human capital that they have. Similarly, some might argue that having, say, more lawyers from higher education does not raise productivity if they are engaged in rent seeking activities (Murphy et al, 1991, model this issue and provide some support with data on enrolments in engineering versus law degree courses). Conceptually, human capital should include

knowledge learnt from formal training and on-the-job learning (for which schooling per se is only an enabler). Also, an individual’s human capital is related to the knowledge they can access (i.e. we may not know the answer now, but we know where to find it – the role of the internet springs to mind). All of these difficulties imply that schooling on its own does not affect growth directly, rather schooling interacts with firm- and economy- level processes – a much more difficult aspect to test empirically.

Macroeconomic factors and the role of government

“It is now widely accepted tha t a stable macroeconomic framework is necessary though not sufficient for sustainable growth”, is how Stanley Fisher opens a 1993 article. In this article Fisher uses regression analysis to assess the links between economic performance and inflation, the black market premium on foreign exchange and government deficits. All of these are used as indicators of the overall ability of the government to manage the economy, with the assumption that poor management distorts incentives and resource allocation. High inflation, high black market premia, or high government deficits are all found to have negative associations with economic growth. Fisher uses a range of methods to investigate causality and non- linear relationships, concluding that there is causality from the factors to economic growth.

Many endogenous growth models have stressed the role of private firms in driving the growth process. This idea links to the oft held view that too much interference from government may be detrimental to efficient productio n and (high) rates of accumulation. This type of thinking has led economists to empirically analyse the relationship between size of public sector (e.g.

government expenditure to GDP) and economic growth. This might appear sensible: data are available and growth regressions investigate broad relationships. On the other hand, the size

(24)

of the public sector is a crude measure of the intended latent variables (marginal tax rates, over-regulation, crowding out), moreover, some forms of government expenditure (e.g.

education) may already be included in the regression and have, a priori, positive impacts.

Ram (1986) found a positive link between government size and economic growth, however, subsequent studies have tended to find either no relationship, a negative association, or an inverted-U association with economic growth (Barro, 1991, Easterly and Rebelo, 1993;

Radovano and Glli, 2001, and Folster and Henrekson, 2001, for OECD countries). These results, however, are part of a continuing debate, both concerning methodological robustness and interpretation (e.g. do other factors, such as demographics, drive the results) (Carr, 1989, Rao, 1989, Agell et al, 1997, 1999).

The debate over the size of the public sector links to another set of empirical research on the role of public infrastructure (especially transport, water and sewage, and communications).

Aschauer (1989) – in a paper that started a wave of studies – linked the decline in US productivity growth in the 1970s with falls in the stock of public capital in the 1960s. This created a lively debate in the US over the role of public investment and challenged the traditional view that high energy prices, inflation and declines in R&D were linked to the productivity decline. Subsequently, many papers criticised Aschauer’s findings of

(extremely) high rates of return to public capital. The two main lines of attack were causation (i.e. high rates of economic growth boost infrastructure spending) and omitted variables in the regression analysis (see Gramlich, 1994). One method of controlling for the latter is a fixed effect, or first difference model, and studies on US state level data using this approach have shown no effect of public capital (Holtz-Eakin, 1994). Again, one can criticise the

‘public capital stock’ as too broad a measure, for example studies have found that

communication infrastructure appears to be robustly linked to economic growth for 1970-90 (Easterly et al, 1997, Roller and Waverman, 2000, for OECD countries).

Rule of law, political systems and social capital

Knack and Keefer (1995) test the idea that secure property rights and efficient, consistent government policies are linked to economic growth. They use data from the International Country Risk Guide on ‘expropriation risk’ and ‘the rule of law’ as explanatory variables, finding a strong, positive partial correlation with contemporaneous economic growth (97 countries, 1974-89). Some authors have focused more narrowly on the importance of intellectual property rights, again finding a positive association with economic growth or

(25)

factor accumulation (Gould and Gruben, 1997, Park and Gunarte, 1997). A wide range of empirical research has also includes variables for political violence (revolution, coups and assassinations), which have a negative association with economic growth, and the Gastil indices of civil liberties and political freedom, which have a positive association (e.g. Barro and Sala-i-Martin, 1995). These results are reassuring, if not surprising, but they do raise questions about causality – perhaps high growth countries create better property rights, allow more freedom and have less political violence?

Property rights, and the quality of institutions in general, can be thought of as the outcome of more fundamental characteristics of a society. Some have defined social capital in this regard, namely, the ‘norms and networks that allow collective action’ Others, however, extend social capital to include the quality of institutions and the level of trust in a society. Testing these ideas requires data on social capital, a key source of such data is the World Value Survey (WVS). The WVS, which has been conducted on up to 29 market economies post 1981, asks individuals whether they can trust others. The answers can be averaged to form a ‘trust’

measure for each country and, when entered as an explanatory variable in a growth regression, this turns out to have a positive association with economic growth (Knack and Keefer, 1997). Research has also shown that trust has strong, positive associations with government efficiency, educational attainment and the importance of large firms in the economy, amongst other variables (La Porta et al, 1997). This is a good example of the correlations between the explanatory variable if interest (trust) and other variables making the interpretation of results unclear. The WVS also asks questions about involvement in networks (e.g. religious organisations, trade unions, professional associations). Knack and Keefer (1997) find no evidence for the role of such networks in economic growth. They also attempt to distinguish between Olson-type networks (e.g. trade unions, professional associations) and Putnam- type networks (e.g. religious, art, community) – but again find no significant

associations.

Other empirical studies have used data from a variety of sources to investigate what might be called fundamental characteristics of societies. Temple and Johnson (1988) use an extensive cross-national data set compiled by Adelman and Morris (1967) which looks at a range of social development indicators including the dominance of kinship, extent of mass

communications and social mobility. Using these data from the 1960s in country regressions explaining economic growth from 1960 to 1985, the results suggest higher social

development raises growth. The association appears to go beyond a link through lower

(26)

fertility or higher human capital accumulation with Temple and Johnson stating “it may be that society matters because it influences the quality of investment, the level of overall technical efficiency, or the ability of countries to assimilate technology from abroad” (p987).

In particular, they find mass communications (based on newspaper circulation and radios per head) has a positive association with TFP growth.

Empirical analysis on the relationship between inequality and growth has used measures of inequality such as the Gini coefficient and quintile shares (Deininger and Squire, 1996).

Initially, cross-sectional analysis tended to find a negative link: high initial inequality was associated with lower growth (e.g. Benabou, 1996). Subsequent studies has introduced uncertainty over this finding. In particular, recently available panel data on a sub-set of countries has allowed country specific effects to be included (i.e. fixed or random effect estimators). Panel studies have found a positive relationship between inequality and growth (Forbes, 2000). Why do the cross-sectional and panel coefficient estimates differ so

dramatically? A traditional response is that the country specific effect is correlated with inequality, implying an omitted variable bias in coefficient estimate from a cross-sectional (OLS) estimator. Alternatively, the time periods considered in cross-sectional analysis (often 20 years), compared to panel studies (often 5 year periods), might imply that the short term effect differs from the long term effect. More specifically, a fixed effect or first difference specification effectively estimates coefficients based partial correlations between changes in inequality and changes in growth. The implication of this has been explored by Banerjee and Duflo (2000). They consider a simple ‘hold up’ model that implies a change in inequality in any direction will reduce growth in the short run. Although this conflicts with the results of panel studies that find positive effects, they argue that the inequality-growth link is non- linear and has both level and change effects. This complexity results in previous empirical

specifications being mis-specified which, they assert, can lead to panel estimates finding positive coefficients. Barro (1999) explores a further complication, namely that the impact of inequality may differ between poor and rich countries. Using cross-sectional analysis he finds that reducing inequality in poorer countries (less than US$2070, 1985 prices) tends to

increase growth, while reducing inequality in richer countries tends to reduce growth. These issues mean that there are still major uncertainties over the inter-relationship between inequality and growth. They also provide an example of the debates about cross-sectional versus panel estimators, and the issue of coefficient heterogeneity, which are common to many empirical studies.

(27)

International openness

There is a real need to understand how greater trade openness affects economic growth. The theory relating to this question is ambiguous, so the scene is set for empirical research to clarify the issue. Although empirical research into this issue has a long history, there is an argument for starting our discussion in 1998. By 1998 four influential empirical papers had been published – based on analysis of multiple countries and time periods - that supported the idea that increased trade openness raised economic growth (Dollar, 1992, Ben-David, 1993, Sachs and Warner, 1995, and Edwards, 1998).

The paper by David Dollar estimates measures of the wedge between the price index of tradable goods in a country and the US (this is a proxy for trade distortions, on the assumption that the ‘law of one price’ holds on average). The distortion measure, and the volatility of this measure over time, are then used in a regression analysis on economic growth. The results suggest trade distortions reduce growth. Sachs and Warner take a different approach. They construct a trade openness dummy variable based on data for tariff and non-tariff barriers, black market premia, whether the state had a monopoly on major exports, and if the country had a socialist system. Again, they find trade openness fosters growth. Ben-David (1993) uses the idea of σ-convergence and studies how the GDP per capita of European countries evolved. The basic finding is that European countries that pursued trade liberalisation experienced σ-convergence. Lastly, Edwards (1998) considers growth in TFP in a cross-section of countries and how it is associated with a nine measures of openness, finding “the regressions reported here are robust to the use of openness indicator, estimation technique, time period and functional form, and suggest that more open countries have indeed experienced faster productivity growth” (p.396). These studies, and many others not mentioned, support an increasingly held view that openness and growth are positively linked.

There are, however, dissenting voices. As noted in the introduction to the empirical section, this type of analysis needs to be aware of (at least) four issues: imperfect proxies, influential observations, controlling for other factors, and causality. Rodriguez and Rodrik (2000) provide a critique of the above studies along these lines. For example, they consider that Dollar’s distortion measures are poor proxies for openness, with the volatility of price distortions (the most robust correlate with growth) being an indicator of general economic instability. In a similar way they note that the Sachs and Warner dummy variable can be decomposed into its relevant components, with the most important components for the link to

References

Related documents

The Macroeconomic Policy and Financing for Development Division of ESCAP is undertaking an evaluation of this publication, A Review of Access to Finance by Micro, Small and Medium

Percentage of countries with DRR integrated in climate change adaptation frameworks, mechanisms and processes Disaster risk reduction is an integral objective of

The Congo has ratified CITES and other international conventions relevant to shark conservation and management, notably the Convention on the Conservation of Migratory

SaLt MaRSheS The latest data indicates salt marshes may be unable to keep pace with sea-level rise and drown, transforming the coastal landscape and depriv- ing us of a

These gains in crop production are unprecedented which is why 5 million small farmers in India in 2008 elected to plant 7.6 million hectares of Bt cotton which

INDEPENDENT MONITORING BOARD | RECOMMENDED ACTION.. Rationale: Repeatedly, in field surveys, from front-line polio workers, and in meeting after meeting, it has become clear that

To break the impasse, the World Bank’s Energy Sector Management Assistance Program (ESMAP), in collaboration with Loughborough University and in consultation with multiple

Angola Benin Burkina Faso Burundi Central African Republic Chad Comoros Democratic Republic of the Congo Djibouti Eritrea Ethiopia Gambia Guinea Guinea-Bissau Haiti Lesotho