Stages of Diversification

 Most of the existing economic theories indicate that income level and sectoral concentration have an inverse relationship. They suggest that although economies want to diversify to dampen sector-specific shocks, they wait until factors are accumulated because opening new sectors is costly. Hence, sectoral concentration decreases as income increases.

On the other hand, a significant body of literature has highlighted why sectoral specialization happens in the development process. Ricardian trade theory suggests that open economies specialize in a specific range of goods; specialization is expected as trade impediments are reduced. Analysts like Paul Krugman (1991) and J. Peter Neary (2001) have also taken interest in the geographic clustering of economic activity. They observed that demand linkages and costly trade makes clustering optimal for competitors. This then results to the increasing degrees of concentration at the sector level within defined geographic zones.

It seems that diversification happens among low-income countries, and concentration among richer economies. Empirical tests show that different measures of sectoral concentration follow a U-shaped pattern. Countries first diversify, but at a certain point in the development process, they start specializing again.

Sectoral data from the International Labor Office (ILO, 1997) and United Nations Industrial Development Organization (UNIDO, 1997) were gathered to study sectoral concentration across time and in relation to development level. Both sets of data include a wide range of industrial and developing countries. The difference is that ILO encompasses all economic activities, while UNIDO comprises only manufacturing. The variable that was the most given attention was the GINI coefficient, which stipulates that the more equal the sector shares, the more diversified an economy.

Nonparametric Estimation

Nonparametric methods similar to the robust locally weighted scatterplot smoothing were used to identify the shape of the relationship between indices of sectoral diversification and income levels. These techniques explore the shape of the relationship linking sectoral concentration and income for a typical country, as well as the sign and statistical significance of the coefficients on income in a within-country regression.

Across all data sets, it was found that the relationship between sectoral diversification and income per capita is nonmonotonic. It follows an asymmetric U-shaped pattern, wherein the shift towards reconcentration occurs at the late stages of development. Reconcentration happens earlier for the manufacturing sector than across other sectors.

Since the curve is asymmetric – the upward bend of the curve does not return to its original level of concentration, it was essential to establish the significance of the slope coefficient on income for all data. For various data sets, it has been confirmed that there exists a range of income per capita after the minimum point (the dip of the U-curve) wherein the slope estimates become positive and generally statistically significant. The upward-bending portion of the curve is also much more distinct for the OECD countries than for developing countries.

The shape of the estimated curves is similar across different sectors, data sources, level of disaggregation, and measures of concentration. But as for the extent of reconcentration in the second phase, the smoothed curves from the UNIDO data are flatter, indicating that within-manufacturing specialization is slower in late stages of development than specialization from other sectors in the economy.

Extension of Analysis

Nonparametric methods do not identify whether the relationship of income and sectoral concentration is driven by between or within-country variation. Quadratic specification answers this question. It allows out-of-sample predictions on the stages of diversification and provides accurate parameter estimates for the shape of the curve.

The results from nonparametric estimation are similar with within-country results. The curve demonstrates a U-shaped pattern, although the estimated fixed-effects coefficients on income and on the square of income are statistically significant. The estimates of the minimum point of sectoral concentration are found to be comparable to nonparametric results; the only difference is that the second stage was estimated to occur later for the ILO. Results can also be extended to cross-country levels. The U-shaped pattern is more distinct than under fixed effects regression, and it remains statistically significant despite use of fewer data points. Overall, it has been observed that within and between-country outcomes are very similar to each other.

Country and Period-Specific Analyses

Most of the large countries that have a relatively closed economy in the OECD sample are specializing. This is also true for small and open economies like Belgium. Countries that are open to trade and similarly sized – like Italy and France are at different stages of diversification. Data from UNIDO and ILO also show that countries which have similar incomes but with different sizes and international trade exposure are on similar stages in diversification. It is noteworthy to mention, however, that there are differences in the income corresponding to maximum diversification. For Singapore this seems to have occurred around $2,500 of per capita income, in Cyprus at $5,800, while in Ireland, it was estimated to have occurred around $7,000.25. All of them are small open economies and seem to have started specializing at levels of income lower than the benchmark estimate from the pooled sample ($9,000). These examples suggest that open economies tend to specialize at low levels of income per capita.

For the U-curves of the sample, it is difficult to describe the dispersion of the minimum point. Moreover, countries that went through a minimum level of specialization relatively early tend to be more open to trade on average, while countries that are late specializers were richer on the average when they did so. For closed economies, they specialize once they reach a certain level of income. This proposes again that income per capita and openness are substitutes in determining the stages of diversification.

Robustness

For the 4-Digit UNIDO Data, the U-curve pattern was still in place for all measures of sectoral labor concentration. On the other hand, logistic transformation verified that the results of the U-curve were still consistent and statistically significant even if they are not bounded anymore by zero and one. The possible bias concerning country size was eliminated by constructing two subsamples, one excluding countries in the highest quartile of the distribution of country sizes, the other excluding countries in the smallest quartile. No evidence contradicting the behaviour of the U-curve was observed. The agricultural sector was also removed to eliminate the possibility that industrialization was the reason for the nonmonotonicity of the curve. Period and region specific analyses were also conducted by splitting the sample according to time and geographic regions; again, all results still hold under these cases.

Diversification and Specialization

It was observed that diversification is influenced by the structure of preferences of the country and their portfolio arguments. Consumption patterns change as income increases, boosting the diversity of goods increases as well. This leads to increasing sectoral diversification throughout development. Acemoglu and Zilibotti (1997) suggested that diversification stems from agents’ decision to invest in imperfectly correlated, risky markets to diversify idiosyncratic risk. This implies that “development goes hand in hand with the expansion of markets and with better diversification opportunities”.

For specialization, Ricardian theory relates specialization to the intensity of trade. For example, specialization is fostered by falling transport costs (or tariffs) as the range of goods produced domestically drops. Economic geography literature (Krugman, 1991) also explains specialization. It emphasizes the importance of demand externalities in explaining the accumulation of economic activities in specific regions or cities. It suggests that it is optimal for monopolistic competitors to cluster because their profits increase in local expenditures, themselves a positive function of the number of local firms.

Endogenizing the Stages of Diversification

Specialization and diversification occur at different stages of development. The observed stages of diversification depend on which force dominates the country’s growth path at that point. Under certain assumptions on the dynamic evolution of relative productivity and the fall in transport costs, countries are predicted to first diversify, and then reach a point at which the force of concentration increases again.

Gilles Saint-Paul (1992) adds that when some countries have limited access to financial markets they resort to sectoral diversification as that is the only available approach to diversify away sector-specific income shocks and smooth consumption.

Conclusion

Poor countries tend to diversify. But as they become richer and they reach a certain level of per capita income, incentives to specialize become the dominant economic force and countries resort to specialization.

Source:

Imbs, Jean and Romain Wacziarg, “Stages of Diversification.” The American Economic Review (March 2003), Volume 93, Number 1, pp. 63-86.