Commodity Price Volatility and Inclusive Growth in Low-Income Countries

Chapter 9. Inclusive Growth in Natural Resource–Intensive Economies

Rabah Arezki, Catherine Pattillo, Marc Quintyn, and Min Zhu
Published Date:
October 2012
  • ShareShare
Show Summary Details
Andrew M. Warner


Lack of structural change, unsustainable growth, and noninclusive growth are frequently mentioned features of natural resource–intensive economies. There is renewed interest in these issues after the new wave of commodity booms over the past 10 years. It is often stated that the challenge is to leverage the billions of dollars in natural resource–related rents to achieve growth, diversification, and inclusive sustainable development. Yet based on theory and evidence from earlier years, the risk of such hopes being thwarted by a recurrence of the natural resource curse is well recognized. The perceived risk encompasses not only the possibility of disappointing growth, but also the possibility that growth will be noninclusive, with disappointing wage and employment growth and a widening dispersion of incomes.

This chapter examines the cases of Saudi Arabia and Azerbaijan, two countries in which the available data permit some empirical assessment of the related issues of diversification, sustainability, and noninclusive growth. Saudi Arabia experienced a boom in the 1970s, and Azerbaijan has had a recent boom. Both permit some analysis of diversification of output across economic sectors. Achieving diversification away from the natural resource sector is the near-universal objective of natural resource–intensive economies and is sometimes viewed as the key to achieving sustainable growth. Saudi Arabia is one of the few highly natural resource–intensive economies that publishes value-added data by economic sector over a long time period, that is, since 1970. It thus enables analysis of diversification during booms and busts as well as the long quieter period since the mid-1980s. Azerbaijan has experienced one of the most dramatic natural resource booms in the past decades. It permits a comparison with the earlier Saudi Arabian boom and also permits an examination of the inclusive growth issue because it is one of the few resource-intensive economies for which one can obtain employment and wage data.

To be specific about the subject, the natural resources being discussed are hydrocarbons and minerals rather than agriculture. The distinguishing feature is that hydrocarbons and most minerals yield high economic rents once they are extracted and produced. They are typically mined after an uncertain search process, which means that countries and companies can experience major windfalls once discoveries are made. World prices are also highly volatile, which adds to the large swings in revenues and periods of high rents.

Inclusive growth as defined here refers to relatively fast growth in incomes of lower-income groups or in which significant social groups are not left out of the growth process. The concept is continuous and thus does not lend itself to strict absolute standards. In a labor market context, growth would be more inclusive the faster the growth in labor incomes or the faster the growth of low-income labor. A benchmark could be to compare the growth in wages with that of per capita GDP, with inclusive growth defined as growth in which the former exceeds the latter. Growth that diminished an unemployment problem would also be considered inclusive growth.

Empirical examination of the degree to which resource-led development has been or is likely to be inclusive faces a number of issues. Data on incomes, wages, and employment are scarce in the highly natural resource–intensive economies, and thus the analysis must adapt to the available evidence. Of the data that are available, GDP by sector is the most common, followed by employment and wage levels. What can we know deductively and what requires data? It may be plausible to believe—through deductive reasoning that gains in the natural resource sector itself will not be widely dispersed because we know that the property rights (usually controlled either by private firms or the state) are concentrated. But this is not enough to settle the inclusive growth question; the full picture requires analysis of the rest of the economy. Whether or not resource-led development is inclusive hinges on sustainable increases in labor productivity outside the resource sector and whether such increases translate into higher earnings or improved employment conditions.

This directs us to the data on GDP by sector in search of evidence of lasting productivity changes. The critical evidence can come from something as simple as the identity of sectors themselves; for example, if a rise in GDP in public administration accompanies a resource boom, it is likely to reflect rent sharing rather than a sustainable rise in productivity. Critical evidence may also be found in the comovement of prices and quantities, because demand and productivity changes will cause these to covary in different ways. This chapter explores how much mileage can be obtained from (inevitably) limited data on these issues.

The chapter therefore is a combination of empirical examination of GDP by sector for information on potentially lasting productivity changes and examination of wage and employment data where available.

The Saudi Arabian Case

Changes in natural resource income dominate aggregate GDP in the highly resource-rich economies. A rather obvious point, although one that is sometimes not heeded, is that a rise in aggregate GDP driven by resource discoveries or price windfalls is not necessarily a reliable indicator of economic progress. Saudi Arabia provides a simple illustration of why it is critical to go behind the veil cast over the economy by the natural resource sector. Figure 9.1 shows real GDP per capita for the whole economy and in the oil sector between 1970 and 2009. A naïve observer might conclude that the economy was highly successful during the early 1970s, when in fact the growth was mostly a function of the boom in the oil sector. For this reason especially, but also because economic mechanisms are likely different in the resource sector and the rest of the economy, it is helpful to separate the natural resource sector from the rest of the economy. In addition, it is potentially helpful to distinguish boom periods from more normal periods. Even though growth may be noninclusive during the boom periods, it may still be inclusive over the long term if the seeds for higher productivity are sowed during the boom period.

Figure 9.1Saudi Arabia: GDP in the Whole Economy and the Oil Sector

(Thousands of Saudi Arabian riyals, 1999 prices)

Source: Saudi national accounts.

Hence, this chapter will examine the inclusive growth issue separately for the resource-intensive sector and the other sectors. We first consider the resource sector itself. There are several reasons to think that natural resource discovery and production will have noninclusive features. Property rights to natural resource wealth are usually held by the state or ceded to private enterprises in exchange for exploration. The immediate effect of discoveries is to enrich those with property rights or residual claims over the resource. Even if concessions to explore for natural resources are offered competitively to investors in a transparent manner so that ex ante expected profits are normal, ex post profits to the lucky investors can still be enormous. This outcome may be inevitable given the uncertain nature of natural resource exploration and production and the need to motivate exploration. The scope for altering revenue-sharing contracts between states and private companies is constrained by the economic incentives of taxation, as too high a tax rate can discourage the investment required to discover natural resources in the first place. Another channel through which natural resource–driven growth can be noninclusive is through powerful interest groups gaining preferential access to rents from natural resources. Such first-order impacts are likely to be significant in determining the inclusive nature of resource-driven development during booms.

Next, consider the rest of the economy. Resource wealth can either be consumed or invested, at home or abroad, and the full impact on inclusive growth depends on a number of factors. These include (1) what sectors the greater spending falls in, (2) the degree to which sectors produce traded or nontraded products or services, (3) the efficiency of the investments, (4) the presence and nature of unemployment, and (5) the labor intensity of expanding and contracting sectors. Whether such second-order impacts are beneficial is governed in part by the efficiency of state-directed investment and the institutional structure under which those decisions get taken. What happens in the labor market is also an important channel governing the inclusiveness of resource-driven growth. As a resource-driven spending boom raises demand for low-income workers, it will show up in some combination of wage increases or employment changes. Skill shortages or other labor supply constraints can govern the extent to which the impact is felt on the wage side or the employment side. The extent to which greater inclusive-ness of growth can be achieved through directed public expenditures can also be constrained by supply conditions. If governments decide to boost health expenditures in an environment of limited supply of health professionals, the higher demand would raise salaries with little impact on employment of health professionals and the total supply of health services.


Diversification, sustainable growth, and inclusive growth are often linked. Diversification out of the natural resource sector helps sustainability of GDP growth because production in other sectors is not as susceptible to booms and busts as is the natural resource sector. Diversification also helps inclusiveness because production in other sectors is believed to be more inclusive than in the natural resource sector. This section offers a note of caution on the link between diversification and growth. It will be shown that Saudi Arabia did indeed diversify in the long term. On that score, it has done well. However, this diversification has not been associated with overall positive real GDP growth per capita over the very long term. Desirable though it is, the benefits of diversification can be overplayed. The Saudi Arabia case illustrates that diversification, even if apparently successful, is not sufficient for successful development.

Saudi Arabia is a major country that both experienced a boom and has some data on economic activity by sector that goes back to the boom period. Figure 9.2 shows per capita real GDP in total and then subdivided into in that for the oil sector and all other sectors. In 1970, it can be seen that the two subsectors were the same size, indicating that the oil sector was half the economy in terms of generation of value added. By the end of the period, the non-oil sector was about twice the size of the oil sector, indicating that the non-oil sector had grown at the expense of the oil sector. At first pass, this evidence suggests that there were indeed investments made during the boom period that paid off in terms of higher output in the rest of the economy but not overall, because total GDP per person was not higher in 2009 than it was in 1972.

Figure 9.2Saudi Arabia: GDP per Person in the Whole Economy, the Natural Resource Sector, and the Rest of the Economy

(Thousands of Saudi Arabian riyals, 1999 prices)

Source: Saudi national accounts.

On further examination, the growth in the rest of the economy was highly concentrated in a few sectors. Over the full 39 years (1970–2009), real growth per capita in the non-oil economy averaged 1.6 percent per year. Much of the long-term rise, however, occurred in four key sectors. Figure 9.3 shows the same graph with the non-oil economy further subdivided into two parts. One part is the sum of four sectors: construction, trade and hotels, finance and real estate, and government services. The second part is the rest of the economy, which comprises agriculture; manufacturing; electricity, gas, and water; transport and communications; and all other services.

Figure 9.3Saudi Arabia: GDP in the Whole Economy; the Natural Resource Sector; the Construction, Trade, Real Estate, and Government Sectors; and the Rest of the Economy

(Thousands of Saudi Arabian riyals, 1999 prices)

Source: Saudi national accounts.

As Figure 9.3 shows, much of the reaction of the non-oil sector to the oil boom occurred in the three real estate–related sectors plus government services. The construction, trade and hotels, finance and real estate, and government services series shows a rise and decline that appears to be a lagged image of the boom in the oil sector. In contrast, the rest of the economy barely moved at all in response to the oil boom—the rest of economy series is extremely flat and shows little sensitivity to the oil boom. Over the full 39-year period, real growth per capita in the rest of the economy (agriculture, manufacturing, services, etc.) grew by an average of 1.9 percent per year.

The Saudi Arabia data afford the opportunity to analyze, with the perspective of 30 to 40 years of hindsight, the net impact of the oil boom between 1970 and 1986. Note that the forcing variable, the oil boom, was roughly symmetric in the size of the boom and the bust. The net effect by sector is summarized in Table 9.1. In this table, the oil sector is shown under the “Mining and quarrying” heading. By 1976, oil value added had increased from 1970 levels by 13,999 Saudi Arabian riyals (SRIs) per person (all figures are in constant 1999 prices). By 1986, it had fully reversed this trend, dropping by 14,625 SRIs per person compared to 1980 levels. As for the other sectors, most of the impact was felt in four key sectors: finance, insurance, and real estate; construction; government services and trade; and restaurants and hotels, in that order. Although these sectors bore the brunt of the boom and bust, they also experienced some long-term increase in GDP once the dust settled, as they came down in the bust by less than they had gone up in the boom. GDP in two of the sectors rose substantially during the boom and partially fell back during the bust (construction and finance and real estate); in the other two sectors, GDP rose and then did not decline much (trade, hotels, and government services). The full impact of the boom and bust, shown in the last column, was that four sectors were higher than before, and the major winner was government services, followed by finance and real estate, construction, and manufacturing. In fact, most sectors, except mining, construction, and real estate, kept growing at modest rates even during the bust period.

Table 9.1Saudi Arabia: The Net Effect of the Boom and the Bust by Sector
Change in GDP per person (constant 1999 SRIs)
Boom periodBust periodNet
1970-19761980-1986(1) + (2)
Agriculture, forestry, and fishing281579860
Mining and quarrying13,999−14,625−626
Electricity, gas, and water−14443
Wholesale and retail trade, restaurants, and1,060−65995
Transport, storage, and communication70417722
Finance, insurance, and real estate5,167−2,8252,341
Community, social, and personal services59784680
Producers of government services3,5634634,027
Source: Author’s calculations using Saudi Arabian National Accounts.Note: SRIs: Saudi Arabian riyals.
Source: Author’s calculations using Saudi Arabian National Accounts.Note: SRIs: Saudi Arabian riyals.

To what extent was structural change achieved? To answer this question, consider a comparison of shares of GDP by sector at the start of the period (1970) with those at the end (1986).

As shown in Table 9.2, the Saudi Arabian economy was significantly more diversified in 1986 than in 1970. During this period, the oil share of GDP was cut almost in half, from 50 percent to 26 percent. Many sectors gained at the expense of the oil sector, notably trade, finance, and government services. Nevertheless, on a per capita basis, GDP was not much greater than in 1970; the implied average annual growth rate over this period was just 1.5 percent, not a high rate compared to that in many other developing countries over the same period.

Table 9.2The Extent of Structural Change in Saudi Arabia, 1970 to 1986
Share in 1970Share in 1986
Agriculture, forestry, and fishing3%5%
Mining and quarrying50%26%
Electricity, gas, and water1%1%
Wholesale and retail trade, restaurants, and hotels2%8%
Transport, storage, and communication4%5%
Finance, insurance, and real estate10%16%
Community, social, and personal services3%4%
Producers of government services15%20%
Per capita GDP (in 1999 SRIs per person)27,06034,330
Average annual GDP growth per capita1.5%
Source: Author’s calculations using Saudi Arabian National Accounts.Note: SRIs: Saudi Arabian riyals.
Source: Author’s calculations using Saudi Arabian National Accounts.Note: SRIs: Saudi Arabian riyals.

To see what has happened over the full time period, consider Table 9.3, which compares the growth rates of the economy by sector over three periods: the boom, the bust, and the period since 1986. The simple growth rate of value added by sector can be a misleading indicator of the drivers of an economy because small sectors may have high growth yet contribute little to the overall increase in GDP because of their small size. To correct for this, the table also shows the contribution of each sectors’ growth to the overall growth rate, calculated as g× s, where g is the average annual sector-specific growth rate and s is the sector’s share in GDP at the beginning of the period. One can see, for example, that during the 1970–80 boom period, the mining sector contributed 5.1 percentage points to the overall 11.5 percent growth in GDP over this period, a contribution to overall growth of approximately 44 percent (5.1 out of 11.5 percent).

Table 9.3Annual Growth and Contribution to Growth by Sector and Time Period, Saudi Arabia
1970–1980 (Boom)1981–1986 (Bust)1986-2009 (Slow Growth)
Growth rate(%)Contribution (%)Growth rate (%)Contribution (%)Growth rate(%)Contribution (%)
Agriculture, forestry, and fishing6.
Mining and quarrying10.15.1-13.5-
Crude petroleum and10.15.0-13.6-
natural gas
Petroleum refining3.
Electricity, gas, and11.
Wholesale retail trade, restaurants, and hotels19.
Transport, storage, and communication10.
Finance, insurance, and real estate16.41.7-3.3-
Ownership of17.11.2-5.0-
Community, social, and personal services9.
Imputed bank service charge18.
Producers of government services9.
Import duties2.70.0-
GDP growth11.511.5-3.7-
Population growth4.25.13.2
Source: Author’s calculations using Saudi Arabian National Accounts.
Source: Author’s calculations using Saudi Arabian National Accounts.

What is noteworthy from the table is that several sectors experienced high growth rates, yet contributed little to overall growth. For example, agriculture, a sector with fairly high growth, never contributed more than 0.3 percentage points to overall economic growth. Significantly, nonpetroleum manufacturing also experienced very high growth, yet also contributed little to overall growth. The second noteworthy fact is that despite the fast growth in several of the non-oil sectors in Saudi Arabia, the overall growth in the economy over the full period, on a per capita basis, has been negative. GDP declined 3.7 percent per year during 1980–86, while the population was rising by 5.1 percent per year; GDP grew 3.0 percent per year during 1986–2009, while the population was rising 3.2 percent per year.

In summary, by some of the conventional measures of diversification, Saudi Arabia is an apparent success. The share of the economy in petroleum production has been cut in half, from 50 percent in 1970 to 24 percent in 2009. Value added in many non-oil sectors has grown, sometimes dramatically. Exports of manufacturers grew on average 13.6 percent per year in dollar terms for a very long period; that is, between 1980 and 2009. Nevertheless, overall economic growth since 1970 on a per capita basis has been only 0.51 percent per year. Growth per capita since 1972 has actually been negative. Growth in the non-oil economy per capita has been only 1.6 percent per year. At the very least, the simple formula—that diversification is sufficient to escape the curse of oil—is not supported by these results. Whether these results say more than that cannot be determined definitively, but they are suggestive. One hypothesis that is consistent with these data is that the drive for diversification has been achieved at the cost of inefficient investments, which although effective in the microgoal of growing certain industries, has come at the expense of overall efficient use of resources for the whole economy, resulting in disappointing growth.

The Case of Azerbaijan

The Azerbaijani data can be compared to those of Saudi Arabia during its boom in the 1970s. There is a close similarity in terms of the sectors affected by the boom. But the data from Azerbaijan offer a better opportunity to assess the inclusive growth issue because employment by sector and some wage data are available. The Azerbaijan data will be examined with two purposes in mind: first, to determine if there is evidence for the kind of diversification and/or long-term productivity improvements that would support sustainably higher incomes, and second, to shed light on the degree to which any productivity gains were reflected in either employment growth or wage growth.

Table 9.4 shows the concentration of the output boom by sector. Most of the real output growth was in only four sectors apart from mining: construction; wholesale and retail trade and repair of motor vehicles; service of hotels and restaurants; and transport, storage, and communication (Table 9.4 gives figures for growth in GDP by sector less population growth).

Table 9.4Composition of GDP Growth in Azerbaijan by Sector, 2000 to 2008
SectorAnnual growth in GDP

at constant prices
Agriculture, hunting and forestry, fishing, fish breeding6%
Mining and quarrying21%
Production and distribution of electricity, gas, and water4%
Wholesale and retail trade; repair of motor vehicles, personal and household13%
Service of hotels and restaurants28%
Transport, storage, and communication15%
Financial intermediation7%
Real estate, renting, and business activities6%
Public administration and defense; social security5%
Health and social work3%
Other community, social and personal service activities4%
GDP (at market prices)15%
Source: State Statistical Committee of the Republic of Azerbaijan.
Source: State Statistical Committee of the Republic of Azerbaijan.

One of the major reasons why resource-driven growth can be disappointing is that demand booms run up against supply constraints, bidding up prices with relatively little real output gain. The data show that in Azerbaijan, at the aggregate level, the boom has not been dissipated entirely by price increases. Over the period in question (2000–08), real output growth in Azerbaijan averaged 12 percent outside of the mining operations (15 percent for the whole economy), while price growth averaged 14 percent per year. This is roughly a 50/50 split between price and real output increases; supply constraints were evidently not a major limiting factor in Azerbaijan. A casual examination of data by sector suggests that if this has occurred in Azerbaijan, it has occurred in only some of the sectors, possibly health, education, manufacturing, and real estate. On the other hand, construction, hotels, and transport have increased output dramatically.

Using Changes in GDP Deflators and Output to Understand Sources of Growth

Before turning to the data on prices and output by sector, an analytical framework would be helpful. A model has been developed and is presented in this chapter’s appendix, drawing from the classic nontraded model but adapted and used to interpret co-movements in prices and quantities under different kinds of driving forces. Critical points from the model are described below.

The model emphasizes the role of four factors that interact with each other. They are, first, whether the forcing variable is demand related (for example, natural resource income) or supply related (for example, a positive productivity improvement in one sector), and second, whether price setting in some of the sectors in the economy is endogenous (as would be the case with nontraded goods and services) or not (traded goods).

A general point illustrated by the model is that changes in output can provide misleading signals about the source of productivity improvements. A productivity improvement in a nontraded sector will not necessarily increase employment in that sector, and it may not increase output, either, because of offsetting forces on the demand and supply side. Hence, the lack of employment growth in one sector, or the lack of fast output growth, does not rule out that the sector was an engine of growth. What is important is to distinguish traded and nontraded sectors. In the case of a nontraded sector, the relative price will decline strongly, and this provides a way to identify such cases in the data. In contrast, if the same productivity improvement occurs in a traded sector, there will be a strong positive increase in employment and even more so in value added. In this case, value added in the other sector will decline, and there will be no impacts on prices. When an increase in demand is the forcing variable, in the case of a nontraded good, both relative prices and employment in that sector will increase together, with the split between the two governed by supply conditions, and employment and value added must decline in the other sector. In the case in which all sectors are traded, the same increase in demand would have no necessary impact on output or employment because any additional demand can be satisfied through imports.

Figure 9.4 shows growth in both price1 and output on the horizontal and vertical axes, respectively (figures graphed are average annual percentage changes between 2000 and 2008). This kind of data allows one to narrow the range of possible explanations, even if a definitive conclusion is not possible. Of the sectors shown, financial services shows an unusually high growth in prices for its rather moderate growth in real value added. This is a pattern consistent with a non-traded service with supply constraints—prices rose rapidly with relatively modest output growth. Financial services may also represent a special case in which there was a shift to higher-priced foreign-affiliated financial services as the boom started. Mining is a special case of a traded good with an exogenously determined price rise; it is a prime mover due to the hydrocarbon boom in both discoveries and world prices.

Figure 9.4Percentage Growth in Prices (Vertical) and Real GDP (Horizontal) between 2000 And 2008

Source: State Statistical Committee of the Republic of Azerbaijan.

Of the other sectors, what is noteworthy is that the four rapid-growth sectors also exhibit lower-than-average price growth. This is especially the case for construction and transport, in which price growth averaged only about 5 percent per year between 2000 and 2008. The low price growth suggests some combination of (1) productivity gains, (2) price discipline due to the traded nature of some of the goods and services covered by these categories, and (3) nontraded goods and services with elastic supply conditions that allow output to grow with moderate pressure on prices. To the extent that the boom is real estate related (as seems plausible from the sectors being construction, hotels, transport, etc.), some combination of (1) and (3) seems the more likely explanation. It is also possible that domestic price changes are due to changes in regulation or changes in world prices. Productivity shocks originating outside the country can also be relevant drivers of prices for traded goods, as occurred in regard to computers and cellular telephones. However, there is little evidence that these are dominant influences on the sectors in Figure 9.4, which were probably aggregated sufficiently to mask these effects.

Four other sectors (health and social work, manufacturing, education and real estate, and renting and business activities) exhibited moderate output growth and high price growth. This pattern is consistent with these sectors being nontraded goods and services for which strong demand growth was coupled with supply constraints.

Inclusive Growth Through the Lens of the Labor Market: Azerbaijan

Labor market adjustments are part of the picture on inclusive growth. If demand pressures are high during a boom period and labor supply constraints bind, it is possible that the boom will bid up wages and improve labor’s position at the expense of profits. In this section, we examine the degree of labor productivity improvement by sector and the degree to which this found its way into higher wages. We also examine the degree to which labor gained on the employment or wage side.

The official employment data in Azerbaijan suggest that if labor gained, the gains were primarily on the wage side, not on the employment side. Data on total official employment indicate that employment stood at 3.7 million in 2000 and rose to 4.1 million by 2008, an average annual growth of 1.1 percent. Over the same period, population growth was at 1.2 percent, so the employment-population ratio changed only slightly, remaining at approximately 47 percent of the population. These data provide little evidence that greater demand, better working conditions, or higher wages induced a higher fraction of the population to enter into formal employment. Without additional data, we cannot say what happened to informal employment.

Although the data do not point to a rise in formal employment as a share of the population during the boom period, they imply that labor productivity—value added per employed person by sector—has risen dramatically, setting the stage for substantial wage gains. Table 9.5 shows value added per employee between 2003 and 2008. It shows a huge change in mining that dwarfs that in all the other sectors. There was significant growth in labor productivity in the four sectors that had high output growth, and there was also relatively fast labor productivity growth in manufacturing (7.2 percent per year). Wage gains thus appear to be the main channel for inclusive growth. (Note that Table 9.5 shows GDP divided by population, starting in 2000, and GDP divided by employment, starting in 2003; employment by sector is available only starting in 2003.)

Table 9.5Labor Productivity
VA per employee

(Constant 2005 MAN)
Agriculture, hunting, and forestry6357651313.8%
Electricity, gas, and water supply and distribution2,1992,215170.2%
Wholesale and retail trade; repair of motor vehicles,9631,81184913.5%
personal and household goods
Rendering of services by hotels and restaurants3,3317,1373,80716.5%
Transport, storage, and communications4,0068,1354,12915.2%
Financial activity12,03110,632-1,399-2.4%
Real estate, renting, and business activities1,7101,71010.0%
Rendering of health and social services9551,1041492.9%
Other community, social, and personal service activities1,2311,4762453.7%
Source: State Statistical Committee of the Republic of Azerbaijan.Note: VA: value added; MAN: Azerbaijan manat.
Source: State Statistical Committee of the Republic of Azerbaijan.Note: VA: value added; MAN: Azerbaijan manat.

Did these labor productivity changes translate into higher wages? If so, to what extent? To investigate this, we consider the two sectors with data on average wages, mining and manufacturing. In both sectors, average monthly wages are reported in nominal terms. These were multiplied by 12 and deflated by the GDP deflator (2005 = 1) to compare with the labor productivity data, which are expressed as annual value added per employee in 2005 prices. The figures discussed below plot wages against productivity (in logs) on an annual basis between 2003 and 2008.

Figure 9.5 shows that the average wage was not strongly related to average productivity in the mining sector. Over the period 2003 to 2008, a 1 percent increase in labor productivity was associated with only a 0.27 percent rise in average wages. This is the expected result when worker skills are highly substitutable and thus wages are held down by competitive pressures. During this period, productivity in mining soared because of the rise in the world price and the rise in production under conditions of increasing returns to scale. Another way to put this is that wages in mining were a function of national labor market conditions rather than conditions in the mining sector, as the labor was substitutable and the mining sector was a small employer.

Figure 9.5Logarithm Wages and Productivity, Mining

Source: State Statistical Committee of the Republic of Azerbaijan.

In contrast, wages in manufacturing show a strong and positive relation with labor productivity. Over the period 2003 to 2008, labor productivity in manufacturing increased by approximately 35 percent, while the average wage increased by 32 percent—a relationship of slightly less than one-for-one. Note that the data in both Figures 9.5 and 9.6 are measured in logs so that percentage changes can be read directly off the axes. It is also apparent in Figure 9.6 that there was a slight lag in the adjustment of wages to productivity. Initially, between 2003 and 2005, labor productivity (horizontal axis) rose strongly with little positive movement in wages. But wages caught up strongly between 2005 and 2008, rising even more rapidly than labor productivity.

Figure 9.6Logarithm Wages and Productivity, Manufacturing

Source: State Statistical Committee of the Republic of Azerbaijan.

When comprehensive wage data are not available, labor productivity data can serve as a rough proxy for wage data, shedding some light on the inclusive growth issue. This is based on the theoretical link between real wages and the marginal value product of labor, should be subject to the usual cautions that the relation holds better for long periods and large changes, and refers to the marginal, not average, productivity. To check the empirical support for this relation, we can refer to manufacturing in Azerbaijan, a sector that happens to have both wage and labor productivity data. The data shown in Figure 9.6 do provide empirical support for the view that there is a positive association between wages and average labor productivity. This provides empirical support for using productivity trends as a proxy for wage trends.

On the basis of this assumption, the data suggest that rapid wage growth was experienced by a relatively small fraction of the Azerbaijani labor force. Of the officially measured labor force of 3.704 million persons in 2000, the majority were either in agriculture (41 percent) or wholesale and retail trade (17 percent). As shown in Table 9.6, labor productivity growth in these two sectors was 5.5 percent and 12.4 percent, respectively, per year. Although these are fast rates of growth, they fall short of one natural benchmark, namely, the 16 percent real per capita growth reported for the whole economy.

Table 9.6Estimate of Average Wage Growth Based on Labor Productivity Data
Employment in 2000 (thousands)Labor Productivity Growth, 2000–08
Agriculture, hunting and forestry, fishing, fish breeding1,519.25.5%
Mining and quarrying39.619.4%
Production and distributing of electricity, gas, and water40.52.9%
Wholesale and retail trade; repair of motor vehicles, personal and household goods626.112.4%
Service of hotels and restaurants9.814.6%
Transport, storage, and communication167.012.3%
Financial intermediation13.52.7%
Real estate, renting, and business activities98.01.0%
Public administration and defense; social security257.74.1%
Health and social work168.92.2%
Other community, social, and personal service activities123.23.2%
Total Employment3,704.36.8%
Source: State Statistical Committee of the Republic of Azerbaijan.Note: The estimate of overall wage growth is 6.8 percent under the assumption that wage growth was equal to labor productivity growth in all sectors except mining, where it is assumed to have been zero.
Source: State Statistical Committee of the Republic of Azerbaijan.Note: The estimate of overall wage growth is 6.8 percent under the assumption that wage growth was equal to labor productivity growth in all sectors except mining, where it is assumed to have been zero.

We may also estimate wage growth for the entire economy (outside of mining). The table shows that the four rapid-growth sectors (construction, trade, hotels, and transport) accounted for 26 percent of total employment in 2000 (956,000 out of 3.704 million). Using employment shares, we can derive estimates for overall wage growth during the period 2000 to 2008 in Azerbaijan. Assume, for illustration, that labor productivity gains in mining did not accrue to domestic labor (this is backed up by the lack of statistically significant association between wages and labor productivity shown in Figure 9.5). Assume further that wage growth in all other sectors tracked labor productivity growth one for one (backed up by the observed wage behavior in manufacturing). These assumptions imply that on a weighted-average basis, wage growth was 6.8 percent, as shown in the last row of Table 9.6 (with weights equal to employment shares). Alternatively, if wage growth were instead assumed to be 80 percent of labor productivity growth, the weighted average would fall slightly to 5.5 percent per year.

Hence, this estimate indicates that labor productivity growth was 6.8 percent per year during 2000–08, using as weights employment shares at the beginning of the boom period. If this is a good proxy for actual wage growth, it would represent rapid wage growth by almost all absolute standards. But in the context of Azerbaijan, with GDP per person growth in excess of 15 percent, it means that wages lagged growth in value added by a substantial margin during the boom period.

An additional manner in which labor could have improved its position was by migrating to higher-than-average-productivity sectors after 2000. Is there evidence that this took place? The Azerbaijani data on employment do not suggest that there was strong migration of labor from certain sectors toward other sectors during the boom period, because there are no sectors with negative net employment growth. Instead, it appears that new employment simply found its way to a select group of sectors. Table 9.7 shows that total employment increased by 351.5 thousand persons between 2000 and 2008. Of this, the largest increase was in construction (72.5), followed by transport (41.5) and real estate and business activities (41.4).

Table 9.7Employment Shifts across Sectors during the Boom Period
Employment (thousands)20002008Change
Agriculture, hunting and forestry1,517.21,553.135.9
Fishing, fish breeding2.04.32.3
Electricity, gas, and water supply and distribution40.545.55.0
Wholesale and retail trade; repair of motor vehicles, personal and626.1654.228.1
household goods
Rendering of services by hotels and restaurants9.823.313.5
Transport, storage, and communications167.0208.541.5
Financial activity13.519.05.5
Real estate, renting, and business activities98.0139.441.4
Public administration and defense; social security257.7274.216.5
Rendering of health and social services168.9183.114.2
Other community, social, and personal service activities123.2135.412.2
Extraterritorial organizations activity0.20.60.4
Total economy3,704.54,056.0351.5
Source: State Statistical Committee of the Republic of Azerbaijan.
Source: State Statistical Committee of the Republic of Azerbaijan.

Did workers tend move to higher-productivity and higher-wage sectors? To some extent, yes, but this was not a dominant fact. In Figure 9.7, it is shown that the simple association between net employment changes and labor productivity is not particularly strong. Labor productivity in 2000 is measured on the vertical axis, and the increase in employment is measured on the horizontal axis. The sector with the largest employment gain, construction, was not an unusually high-labor-productivity sector in 2000. Two sectors that were, electricity and financial services, experienced some of the smallest employment gains. Overall, there is neither a strong positive nor a strong negative relation between initial productivity and subsequent employment increases. Note that to avoid possible problems with reverse causality because employment changes can affect labor productivity, Figure 9.7 shows productivity in 2000 before any employment changes took place. Nevertheless, would use of a later year make a difference? The answer turns out to be no; the evidence in the figure is not altered greatly if labor productivity in the final year, 2008, is used instead of labor productivity in 2000.

Figure 9.7Labor Productivity in 2000 (Vertical) versus Net Employment Increase, 2000–08 (Horizontal)

Source: State Statistical Committee of the Republic of Azerbaijan.

The lack of a strong relation between labor productivity and migration across sectors means that labor migration induced by the resource boom did not exert a huge influence on average wages. Nevertheless, there was some effect, and it was not trivial because a sizable number of workers moved to construction and transport, both of which have higher-than-average productivity. To see this, consider average labor productivity in 2000 (outside of the mining sector) and assess how the average would have changed if labor productivity in each sector stayed the same but the 2008 employment distribution prevailed rather than the 2000 employment distribution. Measured in 2000 prices, average nonmining labor productivity in 2000 was 86 Azerbaijan manat per employee. If the 2008 employment distribution prevailed in 2000, average nonmining labor productivity would instead have been 102 manat per employee. Hence, the changed employment distribution can account for a rise from 86 to 102 in average labor productivity. Over the eight years of 2000–08, this rise represents an annual growth rate of 2.2 percent. This calculation indicates roughly that structural changes in employment can account for 2.2 percent of the wage growth during the boom period.

In sum, this section has considered evidence for inclusive growth by looking at labor market developments during Azerbaijan’s boom period of 2000 to 2008. Determining the inclusive nature of resource-driven growth is not a simple matter of determining who gets the resource-related rents or whether the resource sector itself is or is not labor intensive, because demand effects can cause the boom to have large impacts on sectors that are not closely related on the production side. Also to be considered is whether labor is shifting to labor-intensive sectors or migrating to high-wage sectors.

The evidence suggests that mining does have low employment, and there is evidence that little of the productivity gains in mining found their way into increased average wages for production workers. Outside of the mining sector, what evidence we have is in manufacturing, which does show a strong empirical relation between wage growth and labor productivity growth. Using this relationship as justification for basing estimates of wage trends on labor productivity data, our estimates suggest that employment-weighted wage growth was 6.8 percent per year, far below GDP per capita growth over this period and far below nonmining GDP per capita growth. On the employment side, there is little evidence of labor incomes improving through increases in labor force participation.

Appendix: A Guide to Interpreting Changes in Price and Output Data at the Sector Level


This model offers a framework for keeping track of several factors that are important in interpreting changes over time in data on value added, employment, and prices by sector. The two major factors highlighted interact with each other. They are, first, whether the forcing variable is demand related (for example, natural resource income) or supply related (for example, a positive productivity improvement in one sector), and second, whether price setting in some of the sectors in the economy is endogenous (as would be the case with nontraded goods and services) or not (traded goods).

The conclusions that will be demonstrated are the following:

  • When there is a positive productivity improvement in one sector and if the output is not traded, then employment will not necessarily increase in that sector. The relative price of that sector will decline strongly. Value added in that sector will also not rise necessarily, but will rise to a greater extent than employment.
  • If the same productivity improvement occurs in a traded sector, there will be a strong positive increase in employment and even more so in value added. Value added in the other sector will decline, and, of course, there will be no impacts on prices from the productivity improvement.
  • When an increase in demand is the forcing variable, in the case of non-traded goods, both relative prices and employment in that sector will increase together, and employment and value added must decline in the other sector.
  • When all goods and services are traded, the increase in demand has no necessary impact on output or employment because any additional demand can be satisfied through imports.

This is intended as a guide to empirical work. Examination of how value added, prices, and employment move over time and move relative to each other, together with a knowledge of the economy, can help in interpreting whether the structural changes observed in the data are likely to signal demand-side or supply-side causes. The latter are more likely to underpin sustainable development, improvements in labor productivity, wages, and inclusive growth.

Description of the Model

The model has two sectors and a third natural resource sector, which, to avoid unnecessary complexity, does not employ domestic resources. One of the sectors will be called agriculture; it produces output according to

The second sector, denoted manufacturing, produces output according to

The price of manufacturing relative to agriculture is p = pmpa. At the optimum, employment is distributed across sectors according to

and total labor supply must be employed in one of the two sectors:

On the demand side, consumers maximize log utility U = βln(Ca) + ln(Cm) so that at the optimum, consumption of the two goods follows

It is important to consider the demand side and not just production when analyzing structural change because demand will play a critical role when some of the goods cannot be traded. When goods are not traded internationally, local demand must equal supply:

Alternatively, when goods can be traded internationally, total international payments must balance (accumulation of foreign assets is not examined in this model). In the following equation, R stands for production of the natural resource. To simplify without losing the ability to address important points, natural resource production is assumed to be sold entirely internationally without using any domestic labor or other resources:

Distinguishing Demand and Productivity Shocks When Some Goods Are Not Traded

The essential equations of the model when the output of one sector is not traded internationally are the four equations below, which determine Lm, p, Cm, and Cα:

After substitution to eliminate the consumption variables Cm and Ca, the model reduces to two equations in two unknowns, simultaneously determining the relative price and employment in manufacturing:

The graphical solution to the model is presented in Figure 9A.1.

Figure 9A.1

Source: Author’s calculations.

Once p and Lm are determined, other variables in the model can be solved for by substitution back into the equations above. Note that impacts on GDP can be determined from the model since non-natural-resource GDP, measured in units of the agricultural good, is given by ψ(LLm) + pθLma,, and total GDP is given by ψ(LLm) + pθLmα,+ R.

A natural resource boom will cause national income and demand to rise for both goods. The increased demand in the nontraded sector must be satisfied domestically. This is achieved in part by labor flowing into that sector from agriculture and in part by relative prices for the sector rising to dampen demand as shown in Figure 9A.2.

Figure 9A.2

Source: Author’s calculations.

This first kind of impact from a natural resource boom is well known: Resource sector output rises, output shifts to nontraded sectors, and relative prices rise in nontraded sectors. The impact on nonresource GDP is composed of both an impact on p and impacts on Lm. When GDP is measured at constant, preboom prices, the impact through p is of course held constant. In this case, the observed impact on constant price nonresource GDP depends only on the impact through Lm. Due to the envelope theorem, the impact through Lm will be approximately zero for small resource booms. But it will be significant if the resource boom is large, and in particular, nonresource GDP will decline. This is from the fact that the resource boom shifts more output toward the nontraded sector than would be efficient in the absence of a boom. In effect, the boom temporarily distorts incentives in the nonresource economy and causes it to operate inefficiently from the perspective of the prices that prevailed before the boom or will prevail after the boom is over. When nonresource GDP is measured at constant prices of a period before the boom, it will show a decline. Thus, for large changes in R, provided there are no productivity-raising investments in nonresource sectors and if prices are constant, the model predicts a decline in nonresource GDP.

This tells us what we should expect to observe when the demand boost from a resource boom is the dominant event. What should we expect to observe if instead a productivity boost in the nontraded manufacturing sector is the dominant event? A rise in q, the productivity term in the nontraded sector, would shift both schedules down, as shown in Figure 9A.3.

Figure 9A.3

Source: Author’s calculations.

The upward-sloping schedule shifts to the right because the higher productivity draws labor into the manufacturing sector. The downward-sloping schedule shifts downward because a price decline is necessary to induce consumers to consume more of the nontraded good. The net impact on employment in the nontraded sector happens to be zero in this model, as the two effects exactly offset one another, but the impact on the relative price of the nontraded product would be a decline. To summarize, determining the full impact of a productivity increase on nonresource GDP would in general require attention to three effects—the direct productivity effect, the induced price change, and the indirect effect through employment shifts, Lm. Note that even though the net impact on employment is zero in this model, in a more general setting, it would be ambiguous.

Demand and Productivity Shocks When All Goods Are Traded

When both sectors produce traded goods, the equations of the model are as follows:

Because prices are set internationally, the distribution of employment is determined uniquely by the first equation, as illustrated in Figure 9A.4.

Figure 9A.4

Source: Author’s calculations.

Here, a rise in productivity has a large increase in employment in the sector in which it occurs, unlike in the nontraded case, where it is ambiguous. This also implies that employment and output must decline in the other sector, again unlike in the nontraded case.

Furthermore, once employment is determined as illustrated above, consumption in the two sectors is determined by the latter two equations, as illustrated in Figure 9A.5.

Now what is noteworthy is that a demand increase from a natural resource discovery (increase in R) serves to increase consumption in both sectors (Figure 9A.6). However, by the fact that R does not enter into the equation in Figure 9A.4, the resource boom does not have any impact on the distribution of employment or value added across sectors.

Figure 9A.5

Source: Author’s calculations.

Figure 9A.6

Source: Author’s calculations.

This research was funded through the multidonor Diagnostic Facility for Shared Growth (DFSG), established to support the development and dissemination of methodological tools and approaches for better determining the binding constraints to shared (inclusive) growth in different country contexts. The findings reflect those of the author and do not represent the views of the World Bank or any of the countries contributing to the DFSG.

Andrew Warner is Resident Scholar at the International Monetary Fund.


The prices are GDP deflators at the sector level, but will be referred to as prices.

    Other Resources Citing This Publication