Regional Productivity Convergence in Peru

Keyword: 
Poverty
Topic: 
Microeconomics - Competition - Productivity
Poverty - Inequality - Aid Effectiveness

Peru has emerged as a new star in terms of its growth during the last decade nearly doubling its income while halving its poverty rate. However with its large differences in income across its regions even more important has been its process of geographic convergence. This column analyzes this process of convergence during 2002-2012 and shows that while productivity has convergence this has been driven only by manufacturing and services sector. No convergence occurred in agriculture and services. These results are consistent with recent cross-country evidence on convergence patterns, and with the hypothesis that productivity convergence is more likely in sectors with greater scope for market integration. Because most workers in Peru are employed in agriculture and services, aggregate labor productivity is converging only slowly across regions. Finally, the column suggests that the heterogeneous nature of productivity convergence could also be one of the factors holding back a narrowing down of regional differences in poverty rates.

During 2002 and 2012, Peru experienced unprecedented economic growth, becoming one of South America’s fastest growing economies. GDP per capita grew 65% (at an annualized rate of 5.2%) and the rate of poverty fell from 52% to 26% in the same period. However, regional inequalities in incomes remain quite large. For example, in 2012 GDP per capita in the poorest province, Apurimac, was only 16% of GDP per capita in the Moquegua, the richest province.

Could regional gaps in income be widening during this growth episode? It has been known to happen in other countries. For instance, incomes levels of Chinese provinces seem to have diverged following the open door economic reforms of the late 1970s (Pedroni and Yao, 2006). On the other hand, there is evidence that regional per capita incomes in the US and Japan (Sala-i-Martin, 1996; Barro and Sala-i-Martin, 1991) have had a tendency to converge over time. 

The evolution of regional differences in labor productivity is a key aspect of this question.  A large part of the spatial inequality in incomes is accounted for by spatial differences in labor productivity (output per worker). For instance, labor productivity in Moquegua, for example, is more than 5 times as high as labor productivity in Apurimac.

Moreover, there could be differences across sectors in the evolution of regional differences in labor productivity. A recently proposed hypothesis is that the potential for productivity catch up across countries varies across sectors. Rodrik (2013) suggests that manufacturing sector has greater scope for catch-up because manufactured goods are traded and can be integrated into global production chains.[1] Global integration provides a channel for technological diffusion, and also imposes competitive pressure on lagging firms. A similar dynamic could play out across regions within a country.

 

Manufacturing has been the main engine of regional convergence 

So how have labor productivity gaps across Peruvian regions evolved during 2002-2012? The data show that there are important sector-level differences in this process (Iacovone et. al, 2015). Let us start with agriculture, which in 2012 employed 24% and remains a major employer in nearly all provinces (in spite of having its share reduced from 33% in 2002). Figure 1 plots the growth rate in agricultural labor productivity (measured as value added per worker) versus its initial level in the 24 provinces (“departments”) of Peru. A downward sloping relationship implies that less productive regions experienced faster labor productivity growth; in other words, they caught up. While the fitted slope in Figure 1 is indeed sloping downwards, our econometric analysis reveals that it is statistically indistinguishable from a flat slope. Thus, regional labor productivity gaps in agriculture have not been narrowing down. 

Figure 2 depicts the relationship between the growth rate in mining sector labor productivity versus its initial level in the provinces. Here, in contrast to agriculture, the slope is indeed significantly negative, implying that labor productivity in mining has been converging across the provinces.  However, given the small of share of mining in total employment in most provinces (as an example, the department with the highest share of mining in total employment in 2012 was Madre De Dios, with only 5.6% of its labor working in such sector; other regions show much lower shares), this has limited implications for regional productivity catch up. 

 

As Figure 3 shows, like mining, manufacturing sector labor productivity grew significantly faster in provinces that started out with lower initial levels. Even after ignoring some outliers which could be biasing the “convergence rate” upwards, our estimate is that departments in the bottom 25% of productivity in manufacturing are growing faster than the top 25% most productive departments by about 3 percentage points per annum. 

 

In contrast, there is no evidence of unconditional productivity convergence within the tertiary sector (Figure 4). If anything, it appears that the opposite is happening: regions with higher initial level of productivity were the ones experiencing fastest growth rates. 

These findings are consistent with the results of Rodrik (2013) and the hypothesis that productivity convergence is more likely in sectors with greater scope for market integration.

 

Decomposing the convergence process

The services sector, within which regional gaps in labor productivity are persisting, employs an ever increasing share of Peru’s workforce (52% in 2002, and 58% in 2012). The secondary sector, within which regional productivity is converging, remains relatively small. Hence, most workers in Peru are excluded from the convergence process simply because of their sector of employment.

Figure 5 delves further into this issue by showing the evolution of sectoral employment share, comparing provinces grouped into two sets based on whether they were in the top or bottom half of the overall labor productivity distribution in 2002.  In summary, what we observe here – in an accounting sense – is a set of counteracting forces, some helping and others hindering lagging provinces from catching up.

The change in regional employment compositions is giving a helping hand to regional convergence. As we see in Figure 5, lagging provinces have a much larger primary sector. This relative dependence on the primary sector keeps their average labor productivity low because (barring mining), the primary sector has lower labor productivity than other sectors. However, compared to leading provinces, lagging provinces have experienced faster labor reallocation from the primary to the tertiary sector.  Everything else being the same, this compositional change acts a force for close regional gaps in average labor productivity.

But there is a countervailing force. Because there is no regional productivity catch up within the services sector, even if all the workers were to move into services, regional productivity gaps would persist. The secondary sector, which seems to be the main engine of regional convergence, employs a persistently low share of workers, particularly in lagging provinces. 

Currently, on balance, pro-convergence forces seem to be winning. Pooling all sectors together, aggregate labor productivity is indeed converging across regions (Figure 6). But not surprisingly, the rate of aggregate labor productivity convergence is lower than that within manufacturing and mining.  Assuming that workers will continue to move out of agriculture in lagging provinces and that there are limit to how many workers mining can absorb, there are really just two possibilities for faster regional convergence in productivity: faster regional productivity convergence within services, or if that is not possible, greater labor movement into manufacturing activities. 

Regional gaps in poverty rates persist

Peru experienced poverty reduction across all regions during 2002-2012. In most regions, the rate of poverty fell annually at a rate of around 5%. However, even though the incidence of poverty is falling everywhere, regional gaps in poverty rates seem to be persisting (Figure 7). 

 This is puzzling. In general, Peruvian regions that lag behind in labor productivity also have the highest incidence of poverty. Hence, regional catch up in labor productivity should have helped to alleviate spatial differences in poverty rates as well.

 Sectoral differences in regional productivity convergence could help to explain this inconsistency. The key is that those below the poverty line might be disproportionately dependent on certain sectors for their livelihood. Indeed, Loayza and Raddatz (2010) argue that the extent to which output growth is supportive of poverty reduction depends on the extent it is driven by labor-intensive sectors, such as agriculture and manufacturing.  As we just saw, the main labor-intensive sector in Peru’s lagging regions is agriculture, not manufacturing. Moreover, labor productivity in agriculture has not grown faster in poorer regions. This could be one reason why they have not experienced faster declines in the rate of poverty.

 How to speed up regional convergence?

More equitable access to health and educational services, and better integration of labor, capital and goods markets across regions are usually seen as the key to providing lagging areas more opportunities to catch up with leading areas (World Development Report, 2009). Sector-specific policies, such as place-based incentive for specific industries, are not in general recommended because of their potential for distorting the efficient location of economic activity. Are there grounds for recommending such policies based on the observation that manufacturing and mining sectors alone are converging in labor productivity across Peruvian regions?  Not until we understand why this is the case. Surely, given that a similar pattern is observed in cross-country data, a better understanding of the underlying mechanism is called for.

A key question is to what extent the underlying factors which determine the potential for catch-up are not inherently sector-specific, and instead, linked to the policy environment. Rodrik (2013) suggests that trade and contestability of markets promote convergence through competition and knowledge flows.  Thus, greater market integration in agriculture and services could help to expand the process of convergence beyond manufacturing and mining. But since many services have limited tradability, market integration alone is probably not going to be enough, therefore policies to strengthen competition in the market for services and supply-side policies to promote technology diffusion could also play a key role.


[1] Productivity in farming, mining and certain industries may depend, to a large extent, on location-specific factors such as climate and soil quality. In such sectors, it is not obvious that labor productivity should converge across regions. 

References:

Barro, Robert J., and Sala-i-Martin, Xavier, 1991. “Convergence across States and Regions.” Brookings Papers Econ. Activity, no. 1 (1991), pp. 107–58.

Iacovone, Leonardo, LF Sanchez-Bayardo and Siddharth Sharma, 2015. “Regional Productivity Convergence in Peru”. Policy Research working paper; no. WPS 7499. Washington, D.C.: World Bank Group.

Loayza, Norman V. and Claudio Raddatz, 2010. The composition of growth matters for poverty alleviation. Journal of Development Economics, 93 (2010) 137–151

Pedroni, P. and J. Yao, 2006, Regional Income Divergence in China, Journal of Asian Economics, Vol. 17, 2, pp. 294-315.

Rodrik, Dani, 2013. Unconditional Convergence in Manufacturing, Quarterly Journal of Economics, February 2013.

Sala-i-Martin, Xavier, 1996. "Regional cohesion: Evidence and theories of regional growth and convergence". European Economic Review 40 (6), 1325-1352.

Share this