Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of...

12
Draft, January 7 th , 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido * . Summary This document presents an ad hoc literature review on Structural Change and Patterns of Development. It is written with the purpose of providing a summary on the main theoretical foundations, definitions and empirical approach used to identify and measure uniform development patterns across countries of the World. Such patterns are proxied by the time behavior of variables considered to be as representative of different development process in relation to the level of per capita income. It is based on the empirical work conducted by Hollis Chenery and Moises Syrquin (Chenery & Syrquin, 1975), which, in turn, builds on earlier pioneering work conducted on the subjects by Kuznets 1 and others. Given the relevance of that work, this Literature Review will attempt to provide a summary of definitions, hypotheses, methods and overall results proposed by (Chenery & Syrquin, 1975). It will then describe subsequent updates in methods and results, with special attention being paid to (Chenery & Syrquin, 1989) and to the description of the underlying model used by the authors, as described by Chenery in (Chenery, 1960) and (Chenery, Robinson, & Syrquin, 1986). Further references will be also described. In particular, the Literature Review pays special attention to recent work on structural change, including that of (McMillan & Rodrik, 2011)…[more] Summary of (Chenery & Syrquin, 1975) The transformation of an underdeveloped to a developed economy can be defined “by the set of structural changes required to sustain a continuing increase in income and social welfare” (Chenery, 1982). Although these requirements tend to vary according to country characteristics -such as natural endowments and each country’s social objectives- there exist factors that, hypothetically, produce a degree of uniformity in this transition. Such factors include the changes in consumer patterns with the level of income, the need to accumulate physical and human capital to increase output and the access to common sources of technology and international trade. Based on these ideas and building upon pioneering work by Kuznets on quantitative inter-country analysis of economic structures 2 , Chenery and Syrquin took upon the subject of investigating development patterns 3 across countries and across time within countries, under a uniform and * Consultant, for World Bank, Development Prospects Group (DECPG) 1 (Kuznets, 1967) 2 Kuznets contribution on the subject was distributed across several pieces published in the Journal Economic Development and Cultural Change, between 1956 and 1967. See (Chenery & Syrquin, 1975) for references. 3 Development patterns being defined as “systematic variation in any significant aspect of the economic and social structure associated with a rising level of income or other index of development” (Chenery & Syrquin, 1975).

Transcript of Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of...

Page 1: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

Draft, January 7th, 2014

Patterns of Development. A Review of the Literature for an Empirical

Update.

Leonardo Garrido*.

Summary This document presents an ad hoc literature review on Structural Change and Patterns of Development.

It is written with the purpose of providing a summary on the main theoretical foundations, definitions

and empirical approach used to identify and measure uniform development patterns across countries of

the World. Such patterns are proxied by the time behavior of variables considered to be as

representative of different development process in relation to the level of per capita income. It is based

on the empirical work conducted by Hollis Chenery and Moises Syrquin (Chenery & Syrquin, 1975),

which, in turn, builds on earlier pioneering work conducted on the subjects by Kuznets1 and others.

Given the relevance of that work, this Literature Review will attempt to provide a summary of

definitions, hypotheses, methods and overall results proposed by (Chenery & Syrquin, 1975). It will then

describe subsequent updates in methods and results, with special attention being paid to (Chenery &

Syrquin, 1989) and to the description of the underlying model used by the authors, as described by

Chenery in (Chenery, 1960) and (Chenery, Robinson, & Syrquin, 1986). Further references will be also

described. In particular, the Literature Review pays special attention to recent work on structural

change, including that of (McMillan & Rodrik, 2011)…[more]

Summary of (Chenery & Syrquin, 1975) The transformation of an underdeveloped to a developed economy can be defined “by the set of

structural changes required to sustain a continuing increase in income and social welfare” (Chenery,

1982). Although these requirements tend to vary according to country characteristics -such as natural

endowments and each country’s social objectives- there exist factors that, hypothetically, produce a

degree of uniformity in this transition. Such factors include the changes in consumer patterns with the

level of income, the need to accumulate physical and human capital to increase output and the access to

common sources of technology and international trade.

Based on these ideas and building upon pioneering work by Kuznets on quantitative inter-country

analysis of economic structures2, Chenery and Syrquin took upon the subject of investigating

development patterns3 across countries and across time within countries, under a uniform and

* Consultant, for World Bank, Development Prospects Group (DECPG)

1 (Kuznets, 1967)

2 Kuznets contribution on the subject was distributed across several pieces published in the Journal Economic

Development and Cultural Change, between 1956 and 1967. See (Chenery & Syrquin, 1975) for references. 3 Development patterns being defined as “systematic variation in any significant aspect of the economic and social

structure associated with a rising level of income or other index of development” (Chenery & Syrquin, 1975).

Page 2: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

comprehensive framework4 (Chenery & Syrquin, 1975). The work was an attempt for comprehending

and measuring underlying growth processes that resulted in structural change over a 2 decade period

(1950-1970), and the role played by so called “universal factors affecting all countries” as well as by

“particular characteristics such as natural endowments or market size”5. Results of the work were

expected to shed light on aspects of transformation that could be deemed as important for the design of

economic policy.

Ten basic processes or dimensions representatives of countries’ development paths were analyzed by

the authors in relation to the evolution of their income per capita (See Box 1). The rationale for the

relationships was founded, in some cases, on pure economic theory and, in others, on simple empirical

observations (stylized facts). Such processes were organized in three groups related to:

i) The accumulation of physical and human capital;

ii) The allocation of resources (demand, supply and trade); and,

iii) Distribution and demographic trends.

The choice of proxies for each process was made based on considerations of theoretical significance,

universality, data availability and policy relevance, yielding a total of 29 variables, as summarized in

Table 1. Taken together, these processes described different dimensions of the structural

transformation of countries as per capita income levels increase. From this point of view, the

development dynamics could be understood as a “multidimensional transition from one structure to

another” with lower and upper bounds (asymptotes) for the analyzed variables6.

The general model

(Chenery & Syrquin, 1975) did not provide an explicit general equilibrium model for an endogenous

explanation of changes in identified variables. Instead, econometric specifications in a reduced form

were presented for each of the variables included Table 1 as a function of (per capita) income, and other

controls (a proxy for domestic market size, namely, total country population; and net foreign flows)

which summarized “universal” and other “general” factors determining their behavior over time7.

Examples of such processes are Engels’ Law on the income elasticity for the demand of food, and Arthur Lewis observation that the savings ratio increases with income levels. 4 As opposed to inter-country quantitative analysis of isolated elements in savings, taxation, trade, investment, etc.

5 To be precise, (Chenery & Syrquin, 1975) explain that a country’s development pattern can be described in terms

of 3 components: 1) The effect of universal factors that are related to the income level; 2) Other general factors such as market size and natural resources endowments to which government has little or no control; 3) A country’s history including political and social objectives of governments and their choices of policy. The authors’ study pertains to the identification of uniforms factors in 1) and 2) which “account for well over half the observed variation among countries in most structural characteristics”. It is also important to explain that the authors’ work focuses in growth processes but not into the resulting rates of economic growth. 6 That, hypothetically, move following a logistic approximation

7 See, for instance, (Chenery, 1965) for a derivation of such reduced form equations from an inter-industry model

where the observed patterns of resource allocation are produced by changes in demand from rising incomes, and differences in trade patterns resulting from variations in market size and changes in factor proportions.

Page 3: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

Income and population entered non-linearly8 and a time trend was added to look at hypothetical time

shifts in cross country relations9. Thus way, the authors’ main specifications were:

( ) ( ) ( ) ( )

∑ eq. 1.1

( ) ( ) ( ) ( )

∑ eq. 1.2

Where X corresponds to each of the variables in Table 1 (left column)10; Y is real per capita income

(Gross National Product) which serves as an “overall index of [a country’s level of] development”, N is

total population, Tj is time periods (i=1,2,3,4)11 and F is net imports of goods and non-factor services (the

negative of the trade balance) as a percentage of GDP12. The authors had a maximum of 101 countries in

the sample13 for the years 1950-197014. The population variable was introduced as an independent

variable (uncorrelated with per capita income [check]) to allow for the effects of economies of scale and

transport costs on patterns of trade and production15. The net resource inflow variable was introduced

based on the ideas that trade balances result from trade orientation policies which have an impact in

the allocation of resources; and that, facing pressures from resource reallocation, the ability to balance

supply and demand for individual commodities is substantially affected by the magnitude and

composition of trade. It thus considers direct and indirect effect on development processes via

interactions between exports, imports, savings and investment, among others. The possibility that F

could not be an exogenous variable (in terms of its impact over the dependent variable, X) led the

8 Following the observation that for developing countries one often identifies a period in which the rate of change

of a variable accelerates following an earlier period of little structural change, and continued to an approximation to some asymptotic value. That is, an S-Shaped growth. 9 Another specification alternative to equation 1.1 is also considered allowing each country to have its own

intercept, eliminating all the variation among countries and retaining only the variation within “i” countries:

( ) ( ) ∑

10 So 27 regressions were conducted for each equation 1.1 and 1.2

11 The time trend is included as dummies for the periods 1950-54, 1955-59, 1960-64m and 1965-69 since,

according to (Chenery & Syrquin, 1975) “a year may be too short a unit of time to identify these [structural relationships] shifts” 12

For the variable X, whenever it was represented by shares (but not in logarithms), the sum of the values of the observation across regressions adds up to 100 percent, and the partial effect of each exogenous variable must add to zero. 13

Excluding then-Communist countries and countries that by 1960 had a total population under 1 million people. 14

As it was explained earlier (footnote 6), a non-linear, logistic approximation was, in principle, a better specification than the one presented above given the observation that structural change implies a movement from a lower to an upper boundary at a speed that is proportional to the remaining gap. However, (Chenery & Syrquin, 1975) observed that given the relatively low number of observations for the 1950-1970 period and the concentration of data towards the lower boundary, the logistic form did not provide significant improvements relative to the simpler form used. Results of equations 1.1 and 1.2, however, can be though as valid for the “middle income range of $100 to $1,000” for the period 1950-1970, which represents “75 to 80 percent of the structural change” (Chenery & Syrquin, 1975). 15

[In updating observe the effect of globalization on the relationship of population size and patterns of trade and production]

Page 4: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

authors to cautiously estimate regressions with and without the former, so equations, 1.1 and 1.2 allow

to check whether the potentially non-exogenous variable affects other coefficients in the specification16.

These simple specifications permit, according to the authors, “a uniform analysis of all aspects of

structural change” and provide a basis for further analyzing (in more elaborated specifications)

interdependence issues in a consistent framework. A fundamental hypothesis in this context is that

development processes “occur with sufficient uniformity across countries to produce a consistent

pattern of change in resource allocation, factor use and other structural features as the level of per

capita income rises”. The choice by (Chenery & Syrquin, 1975) of a narrow set controls was deliberate,

so as to yield the biggest possible sample size, and to provide a basic set of results that could be

compared against more elaborated models (in terms of variables included in the reduced form or

complexity of the functional form). In order to derive uniform cross section estimates of structural

transformation the two basic equations 1.1 and 1.2 were applied to the 29 variables chosen to measure

the basic processes.

Box: Definitions, characterization and main hypotheses on fundamental processes associated

with structural change

Accumulation process is defined as the use of resources to augment the productive capacity of the

economy. It refers mostly to physical and human capital, with increases in the latter being linked to

education, public health, technical assistance and other public services that increase human

productivity17. In what refers to physical capital, both investment (the flow variable from which capital

stocks are formed) and savings are considered18. This provides the rationale for variables included in

Table 1, Section A. The main hypotheses for the behavior of those variables in relation to income per

capita are: i) An increase in the savings ratio following the Keynesian postulate of higher marginal

propensity to save; ii) Increase in the investment ratio, enabled from lower share of demand for food; iii)

Increase in government revenues from augmented tax base relative to income (buoyancy hypothesis);

and, iv) Higher budget-enable education expenditures that allow for a build-up of human capital.

Resource allocation process is defined as the systematic change on the composition of demand,

production and trade as the income level of a country rises19. These patterns occur due to the combined

effect (interaction) of the demand effect of rising income and the supply effects of changes in factor

16

In fact, the authors found a weak relationship between F and the level of income per capita [check]. 17

Such resources are commonly referred to as fundamental sources or determinants of growth. Another determinant, Institutions is notoriously absent from this group. 18

Saving behavior stems from the Keynesian hypothesis that marginal propensity savings out of income exceeds average rates of savings. (Chenery & Syrquin, 1975) results in equations 1.1 and 1.2 are based on this hypothesis (direct income effect) although they acknowledge further (indirect) effects from changes in the composition of production, trade, demand and employment and other elements associated with higher income. These indirect effects are considered in […] 19

Changes in the composition of demand must be balanced by corresponding changes in the composition of domestic supply and international trade. Such relations can be specified by se set of input-output or commodity balance equations of the following form relating gross output (X), consumption use (C), investment use (I), Intermediate inputs used by other sectors of the economy (W), Exports (E) and Imports (M), by sectors (i): ( ) ( ) ( ) where the first parenthesis corresponds to domestic demand, the second refers to intermediate demand, and the third to international net demand.

Page 5: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

proportions and technology. These ideas are supported by economic theories, including those of

balanced and unbalanced growth20, the theory of international trade based on factor proportions21 and

the Engel’s Law. The main hypotheses are: i) The rise of non-food consumption share in demand, and,

correspondingly, a higher share of industry and services output in value added at the expense of primary

output; ii) An increase of the share of the investment ratio to the detriment of the total private

consumption ratio; and, iii) A shift in primary exports to manufacturing and services exports; with

export variation being significantly affected by a country size. The process of resource allocation is one

that may contribute to economic growth from changes in average product of labor22 (transitionally, at

least), a subject that will be explained later in {…}.

Demographic and distributional process is referred to, from one part, as the demographic transition

(changes in mortality and fertility rates, plus migration patterns) –which affects the growth rate of

population, the structure of working age population and the amount of labor supply23- and the

urbanization trends that occur as the level of income grows; and, from the other part, as the changes in

the distribution of a welfare metric (income, consumption) that results from changes in human capital

(education, and health) embodied in labor force, in the structure of production, and in government

resources available for distribution programs, all associated with changes in the income level. (Chenery

& Syrquin, 1975) acknowledge the complexity of interactions that determine distribution and

demographic outcomes, which are influenced, among others by: a) The relative size and growth of

sectors and their modes of production; b) The size and characteristics of the population and labor force;

c) The level and behavior of savings and assets ownership; and, d) Political economy issues and the

government stance regarding redistribution policies. Notwithstanding, the authors use their empirical

approach to show general relationships between these processes, identifying aspects of these complex

interactions that could be measured and comparable across countries and across time. The main

hypotheses are that: i) Urbanization patterns follow changes in the structure of labor employment, with

population moving towards urban centers, seeking for jobs in modern sectors, so urbanization rates

increase with (expected) income; ii) mortality rates fall with improvements in health and technology,

followed by lowered birth rates from higher education ad augmented labor income opportunities, to

determine an immediate, short-lived increase in the growth rate of population, followed by a reduction

of the variable over time; coupled with a higher transitional growth of working age population (relative

to total population) as income growth, until population and labor growth equalize at lower rates; and,

20

Relating the evolution of domestic demand to that of production and trade. It uses 2 critical assumptions, namely, that the price elasticity of demand is low in relation to income elasticity of demand; and relatively low level of exports and imports in relation to domestic supply and demand. The former assumption tends to be invalid in smaller economies. 21

Based on the idea that comparative advantages are derived from factors proportions. As physical and human capital increase relative to unskilled labor, production and exports of manufactured products increase. Alternatively, the increased proportion of physical and human capital to unskilled labor leads to higher domestic production, which then, by means of specialization and learning by doing, production costs decrease leading to increases of exports in manufacturing. 22

Changes in the composition of production together with the available (and changes in) technology determine changes in the demand for labor. 23

In turn, the interaction of supply and demand for labor contributes to determine the allocation and (levels and change of) productivity of labor and income.

Page 6: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

iii) For low income countries, an initial worsening of income distribution occurs as income levels

increase, as a result of a growth process that is driven, at early stages, by modern sectors, that is

followed (with lags) by labor reallocation. Only at high levels of income, distribution becomes less

unequal, resulting from different contributing factors such as technological progress spreading more

evenly through the economy, a possible reduction of dispersion in labor productivity across sectors, and

the effect of income transfer programs.

Accounting for heterogeneity in country characteristics

Equations 1.1 and 2.2 allow for estimating average variation in economic structures as the level of

income rises under a premise of uniform development patterns across countries. In reality,

heterogeneous patterns of development are observed after controlling for income levels, with the

former being associated to countries’ specific characteristics. Such characteristics are classified by

(Chenery & Syrquin, 1975) in two groups:

i) Those associated to a country’s “basic structure” (including features such as size and

resource endowments); and,

ii) Those linked to “development policy choices”24.

A crucial problem, however, is that the effect of variables included in those categories on economic

structures may not be independent, and, consequently, not additive to one another25. From the

empirical standpoint, such circumstance would require either adding interaction terms in the general

model, or conducting separate regressions from breaking down the data by sub-samples of countries

based on selected categories, so as to observe the differences in values of coefficients according to

country type26. In the latter case, one must be persuaded that the categories of choice are the ones for

which exist meaningful differences in the impact of controls in the dependent variable (and also ones

that lead to samples that are adequately described by the chosen specification). Following this approach

one would generate samples of countries based on, say, trade characteristics, size, whether a country is

resource poor or resource rich, etc. The approach, however, is limited in scope as a result of the smaller

sample sizes that result from incorporating additional country characteristics27.

Acknowledging these circumstances, and in order to be able, inasmuch as possible, to account for

heterogeneous patterns of development –as opposed to the uniform representation implied by

equations 1.1 and 1.2- (Chenery & Syrquin, 1975) expand their empirical study taking into consideration

24

The author’s acknowledge that, from the empirical point of view, it is difficult to separate the effects of one from the other. For instance, the existence of large reserves of oil tends, to a large extent, to dictate the nature of policy. 25

For instance, if one estimated equation 1.2 including control variables C1 and C2, their effect on X may not be independent from each other. Interaction terms may be required to capture the dependence of one variable, say C1, in the other, C2, in affecting X. In fact, (Chenery & Syrquin, 1975) show the relationship in 1.2 of the trade variable, F, and population N. Indeed, the size of domestic market, the nature of resources and the inflow of capital affect the sectoral allocation of of capital and labor through the impact on patterns of trade. 26

Interaction variable discussions 27

Breaking down country data based on m characteristics, each with n elements results in nm

groups of countries.

Page 7: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

just few country characteristics28 and focusing on a development process that results from the

relationships between trade, domestic allocation, production and factor use. The country characteristics

are:

i) Country size (large vs small) used as a scale index29;

ii) A trade orientation index that reflects both the level and composition of exports, reflecting a

country’s resource endowment, and its trade policy.

Once these factors are accounted for, the other exogenous variables (namely income level and capital

inflow) are, according to the authors, adequately represented in the model. The scale index produced

large and small country subsamples with a cutoff population defined by (Chenery & Syrquin, 1975) at 15

million in 196030. With regards to trade, the authors consider an indicator that includes the level and

composition of exports31, acknowledging the exogenous effect of natural endowments on the size of

primary exports, the endogenous effect of primary exports on other exports (industrial and services) and

the role of government policy in promoting non-primary exports. Thus way the trade orientation index

(TO) is defined as the difference between the actual and the predicted trade bias. The former being

computed as the observed difference between the ratio of primary exports to GDP (Ep) and the ratio of

non-primary exports (Em), all divided by the ratio of exports to GDP (E)32. The predicted trade bias ( T ) is

computed using equation 1.1 with X representing alternatively the ratio of primary, non-primary, and

total exports to GDP. Hence, TO is obtained as:

E

EE

E

EETTTO

mpmp

ˆ

ˆˆˆ definition 1.1

28

(Chenery & Syrquin, 1975) selected those characteristics based on tests of quantitative significance and further restoring to the criteria of resulting sample size, on data availability for the proxy of choice, independence of variables and significance of results. 29

An alternative scale proxy, total GNP was discarded in favor of keeping its 2 components (Total population, N, and per capita income, Y) in the regression, considering that some commodities have income elasticities different than one (<1 for foodstuff and >1 for some industrial goods and services). It is convenient to capture those elasticities separating Y from N. Also, a country group classification based on income tend to change much faster than one based on population size given the observed marked differences in GDP growth across countries relative to population growth. 30

Alternative cutoff sizes were tried (10 million and 20 million population). The division was largely arbitrary, drawing from (Chenery & Taylor, 1968) who, in turn, follow Kuznets’ suggestion about what a small country was based on population data for 1960 [Reference needed, see Chenery and Taylor 1968 page 11 footnote]. It also takes advantage of the fact that there were only few countries (3) in the interval 14-22 million people by 1960. A more rigorous test for separating the sample could be based in the Chow Test for structural change across models. 31

As opposed to a simpler indicator regarding the export orientation of an economy based only in levels (built as the ratio of actual exports to GDP (E ) relative to the ratio of exports to GDP predicted by country size and income level resulting from applying equation 1.1 to a country’s values of Y and N. 32

Actually, (Chenery & Syrquin, 1975) compute the difference between the ratio of primary exports and the ratio of industrial exports to the ratio of exports to GDP, from observing the low share of services exports in total exports [nowadays this may not be the case].

Page 8: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

Thus way, TO could be interpreted as the lead or lag in changing the export composition relative to a

“normal” (average) country pattern. Countries with high TO are classified as primary oriented (TO>0.1);

Normal countries in terms of TO are those with (0.1<TO<-0.1); Finally, Industry (or, more appropriately,

Non-Primary) oriented countries are those with (TO<-0.1).

Combining the two-way country size category with the 3-way trade orientation index yields 6 groups of

countries; however, after conducting statistical analysis for the 6 groups, (Chenery & Syrquin, 1975)

observed uniform results among the large country categories and, for small country countries little

improvement of results for the 3-way split compared to a two-way split (primary with TO>0 and non-

primary with TO<0). Hence, overall results were reported for the following categories: i) Large; ii) Small,

primary oriented; and iii) Small, non-primary oriented. Results for the accumulation and allocation

resources variables were compared to those generated from applying equations 1.1 and 1.2 to the full

sample.

Overall results and Country Typologies

In general, the regression analysis applied to groups of countries yielded different values of coefficients

for the relationship between proxies for development processes and income levels, by groups, and also

compared to those obtained from the basic regression applied to the full sample. Most prominent is the

effect of country size on the volume of trade and specialization, with larger countries having a lower

ratio of exports and imports and lower specialization in either primary or non-primary exports,

compared to small countries. On the flipside, larger countries show a less variable pattern of domestic

production and –it is inferred- tended to adopt more inward looking development policies that had an

impact in accumulation and resource allocation. As a result, in general, larger countries transformed

their economic structure at earlier points in their process of development (measured by per capita

income) even if the economic structures did not differ significantly at the lowest income levels relative

to small countries. This implies, for large countries, accelerated initial changes in production and

investment patterns at early income stages (relative to small countries) followed by a deceleration at

high income levels. Small countries began to catch up only at a relatively high level of income33. Larger

countries also showed higher transitional early increases in savings and investment with small countries

catching at higher levels of income. In addition and (possibly) related to earlier industrialization, larger

countries showed early increases in urbanization rates, larger industrial labor force share, and marginal

higher education than small ones34. However, government revenues ratios lagged in large countries

relative to small countries35.

33

For instance, large countries have a much larger share of industry value added at lower levels of income compared to small countries. Small countries tend to close the gap only at higher levels of income as cross country results converge. In turn, these results reflect in export structure and levels, with large countries concentrating in industrial exports and small countries concentrating in primary exports. 34

(Chenery & Syrquin, 1975) hypothesize that the differences in savings and investments are due to higher larger capital requirement and infrastructure investment in larger countries. The authors also contrast the favorable effect of industrialization on urbanization, education and industrial labor employment with the greater political and administrative obstacles faced by larger economies. 35

Explained, seemingly by the larger tax base on natural resource based activities on smaller countries.

Page 9: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

Akin to the case of country classification by size, dividing the sample according to trade orientation

yielded different values of coefficients for development processes associated with the structure of trade

in relation to income per capita. Indeed (Chenery & Syrquin, 1975) observed that for small countries,

development processes were all affected by external markets and capital inflows. Their results indicate

that small countries with a primary export orientation (Reflecting, in all likelihood, the rich natural

resource base of these economies) register a lag in the average transformation of their export structure.

They lag in total and manufacturing export shares even at high levels of income, compared to industry

oriented small countries36. The industry orientation of the latter group appears to be linked to their lack

of natural resources to provide an export base37. On these counts, small industry oriented countries

tended to mirror the patterns of large countries. However, these results were not carried over to

accumulation and other development processes to any significant degree.

Based on this augmented empirical exercise that accounts for some of the heterogeneous country

characteristics, a typology of countries was produced by (Chenery & Syrquin, 1975). The authors

identified 4 types of countries grounded on some criteria regarding the level and composition of

exports, sectoral composition of production and the level of capital inflows, which are linked to the

nature of demand and comparative advantage of countries38. The criteria elements were:

i) Production patterns, summarized by the production orientation index (PO). This index was

built on a fashion analogous to the trade orientation index described above, as the

difference in the actual ratio of primary production (Vp) and non-primary production (Vm)

relative to the difference implied by a country’s income and size, according to the equation

1.1, which was recomputed for the relevant country size sub-sample;

( ) ( ) definition 1.2

ii) The trade orientation index, based on the level and composition of exports, recomputed

according to country size;

iii) The export levels (shares) relative to predicted export levels, E/ (see footnote 28).

iv) The ratio of net capital inflow

Results from the PO index helped to classify countries as primary oriented (PO>0.07), normal (0.07>PO>-

-0.07) or non-primary oriented (PO<-0.07). Countries were classified according to TO in the same fashion

as PO: Primary Export Oriented (TO>0.1), Normal (0.1>TO>-0.1), and non-primary export oriented (TO<-

0.1). Three groups also were generated according to Export levels: Export above normal (E/ >1.25);

36

Except oil producing economies. 37

These countries also show an abnormally high inflow of external capital., which replaces primary exports as source for enabling the development of the manufacturing sector at early stages of development. 38

For the purposes of grouping countries based on this typology the sample was restricted to countries that: i) Were “large enough to be representative” (Population of 1.5 million or more by 1965); ii) “Transitional” countries, excluding those with income per capita too low (under 100$) or too high (over 1,000$) for at least a part of the decade 1960-1970; and, iii) Economies not disrupted by socio-political events that could prevent growth.

Page 10: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

normal exports (1.25>E/ >0.75) and exports below normal (E/ <0.75). Three groups were derived from

the net capital inflow indicator: High (F>0.04), Normal (0.04>F>0) and negative (F<0)39.

The resulting country types distilled from (Chenery & Syrquin, 1975) empirical exercise were:

i) Primary specialization: PO>0.07; TO>0.1; E/ >1.25

ii) Balanced production and trade: 0.07>PO>--0.07; 0.1>TO>-0.1. The group can be subdivided

between those with normal F and normal E/ and those with high F.

iii) Import substitution: TO>0.1; E/ <0.75; PO<0.07 (normal plus non-primary oriented)

iv) Industrial specialization: TO<-0.1; PO<-0.07. This can also be subdivided between high F and

normal F countries.

Such classification allowed for the allocation of most countries in the transitional stage, based on the

(Chenery & Syrquin, 1975) sample. The typology is informative inasmuch as it picks up similar

development strategies within members of each group and also inasmuch as there are not significant

differences between the intended and actual allocation of resources40. If that is the case, the typology

can help in comparative analysis of development strategies and to provide benchmark for country

performances. The former line of analysis can help to assess the likely effect of policies. The latter refers

to the idea that each development pattern may require difference sequence of structural changes and

that comparisons are better made in relation to countries sharing similar characteristics and facing

similar challenges.

Developed vs Developing countries

A final experiment conducted by (Chenery & Syrquin, 1975) pertained to splitting the country sample by

income groups and following similar steps as in the case of large vs small countries41. The rationale for

this last split was the need to take into consideration the lags in development processes which are not

properly captured by the current income variable. Since the regression analysis includes only the level of

income and not its growth rates, it may be the case that in fast growing economies –that move rapidly

into higher levels of income- development process that take much longer to occur (say, fertility rates

that takes decades to fall [check]) will be overestimated based on results of equations 1.1 and 1.2.

Separating the sample by income levels may help to overcome this problem. In addition, (Chenery &

Syrquin, 1975) argue that fast growing developing countries (which began in 1950 at lower level of

incomes but were by 1970 in the group of higher income economies) had political and social

characteristic different from those in the (more homogeneous) initial group of high income countries, so

it became apparent that a separation of countries based on some income level (say, at initial year or

midway in the transition) could yield different sets of results. Finally, and more importantly, separating

39

The threshold being chosen in such a way that 40% of observations are included in the Normal groups). 40

(Chenery & Syrquin, 1975) argue that some of the countries classified in the Import substitution group actually tried (and failed) in promoting a more manufacturing led growth (which would place them in the Balanced production and trade group). However they observed sufficiently widespread features of the Import substitution type in many countries to persuade them to keep it as a separate category. 41

The authors considered 2 types of split: One for countries above or below the 500$ per capita mark; and a 3-way split for incomes under 200$, between 200$ and 800$ and above 800$. The latter was used when the first split showed significant difference between the 2 groups.

Page 11: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

countries by income level was in line with a hypothesis of non-linear (s-shaped) transition from low

income and high income countries, further building on the assumptions used to define functional forms.

Results of this procedure ratify the s-shaped dynamics for several development processes42, with a

significant acceleration in the rate of change in the middle income group in the case of a 3-way split of

income43, with lower and higher asymptotes in the low and high income groups, respectively. Results

also indicate different scale effect, by income groups, with time trends being slightly larger for high

income countries.

Time Series versus Cross section patterns

In bringing together times series and cross section data (Chenery & Syrquin, 1975) attempt to shed light

for reconciling apparent empirical discrepancies regarding development processes estimated based on

within country (times series) analyses versus country cross section estimates of production, demand and

savings functions. The idea was to build on the cross country framework summarized by equations 1.1

and 1.2, now incorporating available times series data for a large number or countries, thus estimating

the following specification for each of the accumulation and allocation processes variables included in

Table 1:

( ) ( ) ∑ eq. 1.3

Several observations are in order: First, equation 1.3 is a (country) fixed effect specification where i

represents a constant (dummy) variable or own intercept for country i, so the variation between

countries is eliminated and only the within component of the total variation is considered. Figure 1 plots

a hypothetical case of a variable X representative of a development process as a function of the income

level. The coefficients in equation 1.3 yield the weighted average values of income effects (1, 2), the

time trends (j) and net capital inflows () within countries; and the i summarizes the average times

series results across countries, which picks up the effect of omitted variables related to income that

affect X. Second, the population variable is omitted given its dynamic relationship with the time trend.

Third, two types of regressions were conducted: An unbalanced panel (with all countries from the cross

section analysis being included), and a smaller (in terms of countries included) balanced panel data44.

Figure 1

42

Mainly for savings in the case of accumulation processes, and for exports in the case of allocation processes. 43

And at the low income group, in the case of the 2-way split. 44

The balanced panel has the the advantage that it permits better comparison of short term (within country) patterns with the cross section results, and the disadvantage that it underrepresents lower income countries (those with more missing data for the full period). Furthermore, the balanced panel in (Chenery & Syrquin, 1975) showed an undersirable high correlation (collinearity) between income and the time trend while the full sample allowed for a larger variation for income and a lower association between income and time.

Page 12: Patterns of Development. A Review of the Literature for … January 7th, 2014 Patterns of Development. A Review of the Literature for an Empirical Update. Leonardo Garrido*. Summary

In (Chenery & Syrquin, 1975) dataset for 1950-1970, the resulting (short run, within country) elasticities

for several variables representative of accumulation and allocation processes, as derived from equation

1.3, differ from the (long run) estimates obtained using equations 1.1 and 1.2. Differences in times

trends were also observed (normally larger than in cross country analysis).

Equalization of factor returns is hampered by immobility and unequal access to capital, land and other

complementary factors of production.

Results:

Distributional process: Income level does not explain much of the inter-country variation. There is a lot

of variability. (Chenery & Syrquin, 1975) use proxies for factors (beyond those included in the standard

regressions) that have a hypothetical effect on income distribution: Education (measured by school

enrolment), dualism (measured by the share of primary exports) and the extent of agriculture sector

(measured by the share of primary sector). The inclusion of these variables improve regression results.

Perform an analysis including changes in income distribution over time

pool

Income level

X

Country 1

Country 2

Country 3