Methodology

We use industry-level data to study Indian productivity compared to the technological frontier. We use KLEMS (Capital, Labor, Energy, Materials, and Services) databases for Indian and advanced economies prepared by the Reserve Bank of India and the Luiss Lab of European Economics at Luiss University in Rome, respectively.⁴ These databases provide broadly consistent information on inputs and output per industry at the 2-digit industry level. The RBI KLEMS database extends from 1980/81 to 2020/21, with the financial year running from April to March. The EU KLEMS database includes 27 European Union member states, the US, Japan, and the United Kingdom and runs from 1995 to 2020. So, it covers 30 advanced economies, although data is not available for every year and industry, constituting an unbalanced panel.⁵ We focus on the period of overlap between the databases from 1995 to 2019 (1995/96 to 2019/20 for Indian data) and we disregard the most recent observation as it was the first year of the COVID-19 pandemic and it also disproportionately affected the Indian data.⁶ We map 25 industries across the KLEMS databases as described in Annex 1, but we exclude government from the results shown below to focus on the productivity in the marketplace.

Our preferred measure of catch-up to the technological frontier is the change in relative labor productivity. We define labor productivity at the industry level as the value added (measured in constant U.S. dollars at 2011 prices) per employee.⁷ And we define the technological frontier as the 90^th percentile of labor productivity among advanced economies for each industry and at each point in time. The relative labor productivity is then the ratio between Indian labor productivity and the frontier labor productivity for each industry and time. Indian labor productivity relative to the frontier is shown in Figure 5, by industry and over time.

We consider two alternative measures of productivity:

• First, we use labor productivity measured in PPP (purchasing power parity)-adjusted U.S. dollars at the industry level to account for the fact that prices tend to increase with income per capita, especially for services. We use relative prices for 2005 at the industry level for the countries in our sample estimated by Inklaar and Timmer (2014) to express value added in PPP-adjusted U.S. dollars at 2005 prices. In particular, we compute value added measured in PPP-adjusted U.S. dollars at 2005 prices for industry i at time t in country c by:

  ${{VA}^{PPP}}_{i,t,c}=\frac{{VA}_{i,t,c}}{{FX}_{2005,c}{PPP}_{i,2005,c}}\frac{{Deflator}_{i,2005,c}}{{Deflator}_{i,t,c}}$

  Where VA_i,t,c is the nominal value added by industry i at time t in country c measured in local currency, Deflator_i,t,c is the GDP deflator for industry i at time t (we normalize Deflator_i,2005,c = 100), FX_2005,c is the average exchange rate against the U.S. dollar in 2005, and PPP_i,2005,c is the relative price level in industry i in 2005 in country c obtained from Inklaar and Timmer (2014). PPP_i,2005,c is normalized to 1 for the U.S., so that levels lower than 1 imply that it is cheaper to acquire the output from industry i in country c than it is in the U.S.

  It is attractive to use PPP-adjusted labor productivity because it allows to control for prices raising with income per capita in what is known as the Penn-effect (Samuelson, 1994). This is particularly important for services, which tend to be much cheaper in poorer countries than in advanced economies. However, estimates of relative prices of industry output are only available for the year 2005 (Inklaar and Timmer, 2014) and so, require extrapolation based on growth of industry-specific deflators, which have some methodological weaknesses and inconsistencies across countries. In particular, India does not have producer price indexes and so the deflators for some services instead rely on wholesale prices indexes on associated goods. Thus, we present these results as a robustness check, rather than as baseline estimates.

• Second, we measure productivity through an estimate of the Total Factor Productivity (TFP). For consistency across the KLEMS databases, we define TFP for industry i at time t in country c as a residual from the following equation:

  ${TFP}_{i,t,c}=\frac{Y_{i,t,c}}{{\left(L_{i,t,c}\right)}^{{\alpha }_{i,t,c}}{\left(K_{i,t,c}\right)}^{{1-\alpha }_{i,t,c}}}$

  Where T_i,t, L_i,t,c, and K_i,t,c are value added, number of people employed, and capital stock in industry i at time t in country c. The parameter α_i,t,c is computed as the average labor compensation as a share of value added between in years t – 1 and t.⁸ Data on labor and capital composition or number of hours worked are disregarded for this exercise because they are not consistently available across the two KLEMS databases.

  The use of TFP as a measure of productivity is attractive because it allows to abstract from the increase in labor productivity induced by higher capital accumulation. However, it requires more assumptions to be met and more data to compute it, which significantly restricts the number of countries available to define the technological frontier. In balance, we prefer to include these results as a robustness check.

View Full Size

India’s Labor Productivity by Industry Over Time

(Percent of the frontier)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: Country authorities, EU KLEMS, and IMF staff estimates.Notes: Labor productivity defined as value added (in constant U.S. dollars) per employee at the 2-digit industry level. Frontier is defined as the 90^th percentile among advanced economies.

• Download Figure
• Download figure as PowerPoint slide

Results

The catch up of Indian industries to the technological frontier has been uneven (Figure 6). Agriculture has remained stuck at low relative productivity levels since 1995. In 1995, the productivity of a worker in agriculture in India was only 2.3 percent of that observed for a country at the technological frontier. By 2019, it had increased to only 2.9 percent. This reflects that agriculture in India remains a safety net as was evidenced by the reverse migration wave at the beginning of the COVID-19 pandemic. Excess labor is allocated to agriculture, which keeps productivity depressed in the sector. Catch up to the frontier was also weak for several manufacturing industries (wood and paper, chemicals, electrical and optical equipment, and other manufacturing). Manufacturing of machinery regressed over this period, moving backwards by 2 percentage points in terms of distance to the labor productivity frontier. Productivity in construction improved marginally, catch up to the technological frontier was much faster in mining and petroleum refining, as well as information and business services, health, and other services. Overall, services have been more dynamic in catching up to the technological frontier. Labor productivity in business services is at 18 percent of the frontier.

View Full Size

India’s Labor Productivity

(Percent of the frontier)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: Country authorities, EU KLEMS, and IMF staff estimates.Notes: Labor productivity defined as value added (in constant U.S. dollars) per employee at the 2-digit industry level. Frontier is defined as the 90^th percentile among advanced economies. Industry names detailed in the next figure with A = Agriculture, B= Mining, etc.

• Download Figure
• Download figure as PowerPoint slide

The shift of workers towards more dynamic sectors of the economy has been limited so far (Figure 7). Employment in agriculture declined by 17 million between 1995 and 2019, which is encouraging. But with agriculture still accounting for 43 percent of the workers in the country, significantly more shift of workers towards higher value-added jobs would be needed. Employment grew the most in construction (46 million) and trade (33 million), both relatively low productivity industries that exhibited limited catch up to the technological frontier over the last two and a half decades. Manufacturing added only 15 million jobs over this period, a tenth of the net job creation. Employment in business services did increase significantly by 13 million workers from just 1.6 million in 1995 to 14 million in 2019, while the industry narrowed the distance to the technological frontier by 9 percentage points—a very positive outcome.⁹ Other industries with fast catch up such as mining, petroleum refining, and information services, have added relatively fewer jobs. This is not unexpected given their relatively low labor intensity.

View Full Size

Change in Employment and Productivity: 1995 – 2019

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: Country authorities, EU KLEMS, and IMF staff estimates.Notes: Labor productivity defined as value added (in constant U.S. dollars) per employee at the 2-digit industry level. Frontier is defined as the 90^th percentile among advanced economies.

• Download Figure
• Download figure as PowerPoint slide

The uneven catch up to the technological frontier is robust to the use of alternative measures of productivity:

• Agriculture. The relatively weak performance of agriculture is consistent across metrics. The catch up to the technological frontier in agriculture was 0.6 percentage points based on labor productivity, 0.5 percentage points when labor productivity is adjusted by PPP at the industry level, and -0.9 percentage points in terms of TFP.

• Mining. In mining, the catch up to the frontier in labor productivity (8.6 percentage points) is not just explained by higher capital accumulation because TFP catch-up amounted to 6.7 percentage points, although when adjusted by PPP the catch-up in labor productivity declines to 4.5 percentage points reflecting convergence with international prices.

• Manufacturing. For most of manufacturing, the limited increases in labor productivity appear to be explained mostly by higher capital accumulation with TFP-based catch up being negative or barely positive. On average, TFP-based catch up in manufacturing was just 0.5 percentage points. PPP adjustments make a significant difference only in assessing the catch up to the frontier for the manufacturing of petroleum and of vehicles.

• Utilities. The catch up to the frontier in utilities is larger when measured in PPP-adjusted productivity or in TFP reflecting significant improvements in the sector beyond capital accumulation.

• Construction. The catch up to the frontier in construction is fairly stable, and low, across metrics.

• Services. On services, the average catch-up in labor productivity (6.3 percentage points) appears partly explained by capital accumulation given that TFP-based catch-up averaged 3.4 percentage points. PPP-adjusted labor productivity does suggest a much larger catch-up to the technological frontier in services; however, this needs to be interpreted with caution due to the lack of producer price indexes for services in India, with the GDP deflator for those sectors instead relying on other price data, such as wholesale price indexes for goods.

View Full Size

Change in the Distance to Frontier Employment for Alternative Definitions of Productivity: 1995 – 2019

(Percentage points)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: Country authorities, EU KLEMS, and IMF staff estimates.Notes: Labor productivity defined as value added (in constant U.S. dollars) per employee at the 2-digit industry level. Labor productivity adjusted by PPP uses instead value added in PPP-adjusted U.S. dollars at the industry level for the numerator. TFP is total factor productivity and is computed as a residual as discussed in the methodology subsection. In all cases, frontier is defined as the 90^th percentile among advanced economies.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Labor Productivity in Agriculture

(Thousand dollars per worker)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: World Development Indicators, World Economic Outlook, and IMF staff estimates.Notes: Median labor productivity for each country group. Country level observation reflects five years of available data before the COVID-19 pandemic: 2015–2019.

• Download Figure
• Download figure as PowerPoint slide

India’s agricultural productivity is low compared to less advanced peers (Figure 9). Combining the share of agriculture in value added and employment from the World Development Indicators with employment and GDP from the World Economic Outlook, we can estimate labor productivity in agriculture measured in dollars at current prices and at 2017 PPP-adjusted prices. India’s PPP-adjusted labor productivity is only 12.2 percent of the median productivity in advanced economies, but also just 42.6 percent of the median productivity in emerging markets. India’s labor productivity in agriculture is 54.3 percent of the median productivity in Latin America and the Caribbean and even 84.4 percent of the median productivity in emerging and developing Asia. Distance is even larger when measured in dollars at current prices, without the PPP adjustment. Thus, India’s agricultural sector has substantial room to catch up even when compared to peers at closer stages of economic development.

Methodology

We use firm-level data to study Indian productivity compared to the technological frontier at a more disaggregated industry level. We use ORBIS data collected by Bureau van Dijk, which provides comparable financial data for public and private firms around the world, including industry classification following NACE (Nomenclature of Economic Activities) Rev. 2 at the 4-digit level. The data is not limited to publicly-listed firms as data is also retrieved from company registrars.

We clean up the data to mitigate concerns of uneven coverage across countries. We select all firms with sales and employment data at some point in the past 5 years. We broadly follow Diez and others (2019) in cleaning up the data in the case of duplicates.¹⁰ We average sales and employment data per firm across years. Because different countries have different reporting requirements, we focus only on firms with more than 20 employees to ensure that sample coverage is relatively similar and tilted towards medium and large firms.¹¹ We exclude firms with negative sales. We end up with a sample of 731,916 firms, of which 2,071 are Indian (Table 1). Table 1 presents summary statistics. To determine whether a firm is domestically- or foreign-owned, we compare the firm’s country with that of its Global Ultimate Owner. We use sales per employee to proxy for labor productivity. Sales per employee is not an ideal measure of productivity. Unlike value added, it does not exclude materials. And the measure of employment is also imperfect as different countries have different reporting requirements. Nevertheless, sales per employee is an attractive metric because it is simple, can be calculated with little data, and allows us to explore productivity for more finely defined industries and to study heterogeneity in productivity within those finely defined industries.

^{Table 1.}

Number of Firms by Country

Source: ORBIS and IMF staff estimates. Notes: The analysis uses all the firms with sales and employment data for at least one year over the past five years and with more than 20 employees.

^{Table 1.}

Number of Firms by Country

Country	Ranking	Number of Firms	Percent
U.S.	1	264,164	36.1
China	2	108,901	14.9
United Kingdom	3	38,279	5.2
Brazil	4	37,587	5.1
Italy	5	35,998	4.9
Russia	6	30,437	4.2
Germany	7	29,148	4.0
France	8	19,474	2.7
Spain	9	17,978	2.5
Japan	10	15,501	2.1
India	30	2,071	0.3
Other Countries	-	132,378	18.1
Total		731,916	100.0

Source: ORBIS and IMF staff estimates. Notes: The analysis uses all the firms with sales and employment data for at least one year over the past five years and with more than 20 employees.

View Table

^{Table 1.}

Number of Firms by Country

Country	Ranking	Number of Firms	Percent
U.S.	1	264,164	36.1
China	2	108,901	14.9
United Kingdom	3	38,279	5.2
Brazil	4	37,587	5.1
Italy	5	35,998	4.9
Russia	6	30,437	4.2
Germany	7	29,148	4.0
France	8	19,474	2.7
Spain	9	17,978	2.5
Japan	10	15,501	2.1
India	30	2,071	0.3
Other Countries	-	132,378	18.1
Total		731,916	100.0

Source: ORBIS and IMF staff estimates. Notes: The analysis uses all the firms with sales and employment data for at least one year over the past five years and with more than 20 employees.

We define the technological frontier as the 95^th percentile of sales per employee across non-Indian firms by industry.¹² For the results below comparing Indian firms’ distance to the frontier, we keep industries in which there are at least 5 Indian firms. This restricts the sample to 109 industries.

^{Table 2.}

Summary Statistics

Source: ORBIS and IMF staff estimates. Notes: The analysis uses all the firms with sales and employment data for at least one year over the past five years and with more than 20 employees. Sales and sales per employee are measured in thousands of U.S. dollars. Employees are measured in units.

^{Table 2.}

Summary Statistics

	Sales	Employees	Sales per Employee
5th Percentile	490.0	21.0	9.8
10th Percentile	946.3	23.0	19.6
25th Percentile	2,710.0	32.0	50.0
Median	8,752.4	60.0	137.6
Mean	155,788.7	504.8	447.0
75th Percentile	33,900.0	162.6	274.4
90th Percentile	134,334.8	530.5	622.3
95th Percentile	333,564.3	1,178.6	1,035.8
Standard Deviation	2,217,519	6,553	6,558
Number of observations	731,916	731,916	731,916

Source: ORBIS and IMF staff estimates. Notes: The analysis uses all the firms with sales and employment data for at least one year over the past five years and with more than 20 employees. Sales and sales per employee are measured in thousands of U.S. dollars. Employees are measured in units.

View Table

^{Table 2.}

Summary Statistics

	Sales	Employees	Sales per Employee
5th Percentile	490.0	21.0	9.8
10th Percentile	946.3	23.0	19.6
25th Percentile	2,710.0	32.0	50.0
Median	8,752.4	60.0	137.6
Mean	155,788.7	504.8	447.0
75th Percentile	33,900.0	162.6	274.4
90th Percentile	134,334.8	530.5	622.3
95th Percentile	333,564.3	1,178.6	1,035.8
Standard Deviation	2,217,519	6,553	6,558
Number of observations	731,916	731,916	731,916

Source: ORBIS and IMF staff estimates. Notes: The analysis uses all the firms with sales and employment data for at least one year over the past five years and with more than 20 employees. Sales and sales per employee are measured in thousands of U.S. dollars. Employees are measured in units.

Results

We find that sales per employee among Indian firms are low on average, but with exceptions (Figure 10). In line with the results at the industry level, the median productivity of Indian firms is low, averaging 17 percent of the sales per employee by frontier firms across industries. But there are some sectors with firms exhibiting sales per employee closer or even above the frontier, such as manufacturing of jewelry and soap, printing of newspapers, and other professional services.

There are some strong Indian performers (Figure 11). There are 126 Indian firms already operating at the technological frontier in the sense that their sales per employee are above that of the 95^th percentile for non-Indian firms. Twelve of those Indian firms at the frontier operate in computer programming or other information technology services, and another seven engage in jewelry manufacturing.

These strong Indian performers tend to be large firms (Figure 12). Three quarters of the Indian firms operating at the frontier have assets greater than $100 million whereas only half of the Indian firms not at the frontier have such level of assets. They are also slightly more likely to be foreign owned, with 10 percent of the Indian firms at the frontier being foreign owned compared to 7 percent for non-frontier Indian firms. This correlation is in line with previous literature that has found that FDI can increase productivity (Smarzynska Javorcik, 2004; Zhao and Zhang, 2010: Li and Tanna, 2019). The success of these Indian firms stresses the importance of removing obstacles preventing firms to grow and fostering foreign direct investment. In addition, the presence of these strong performers across industries is consistent with India’s increasing diversification and complexity of exports (Nageswaran and Unnikrishnan, 2023; Harvard’s Growth Lab, 2023).

View Full Size

Sales per Employee for Indian Firms

(Percent of the frontier)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: ORBIS and IMF staff estimates.Notes: The analysis uses all the firms with sales and employment data for at least one year over the past five years and with more than 20 employees. Frontier is defined at the industry level as the 95^th percentile for non-Indian firms.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Number of Indian Firms at the Technological Frontier by Industry

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: ORBIS and IMF staff estimates.Notes: The analysis uses all the firms with sales and employment data for at least one year over the past five years and with more than 20 employees. Frontier is defined at the industry level as the 95^th percentile for non-Indian firms.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Frontier and Non-Frontier Indian Firms by Characteristic

(Share of firms, in percent)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: ORBIS and IMF staff estimates.Notes: The analysis uses all the firms with sales and employment data for at least one year over the past five years and with more than 20 employees. Frontier is defined at the industry level as the 95^th percentile for non-Indian firms.

• Download Figure
• Download figure as PowerPoint slide

Methodology

India’s growing population implies that it needs to create a large number of jobs in the coming years. Estimates of the number of jobs needed vary according to the underlying assumptions. For example, McKinsey (2020) puts the number of non-farm jobs needed by 2030 at 90–145 million while Mody (2023) estimate 200 million by 2033. In this section we calculate the number of jobs needed from 2023 to 2050 accounting for rises in population, potential changes in labor force participation rates, and government job schemes.

We use data from India’s Periodic Labour Force Survey (PLFS) and the United Nations population projections as the basis for our estimation. According to the PLFS, the working age population (defined as 15 years and older) stood at just over 1 billion in 2022, of which 50 percent were working (the labor force participation rate was 53 percent, and the unemployment rate was 5.7 percent).¹³ The UN projects that by 2050 India’s working age population will reach 1.337 billion. With this underlying data, we estimate the number of jobs needed as follows:

• Estimate the lower bound of the number of jobs needed in 2022: We assume that India was at full employment at its 2022 unemployment rate, meaning that the number of individuals who are unemployed are not counted in the missing jobs total.¹⁴

• Estimate the upper bound of the number of jobs needed in 2022: For our upper bound estimate, we consider the number of jobs needed if the labor force participation was higher. Specifically, we consider a LFPR of 60 percent which was the LFPR in 2005 and is approximately the average LFPR across the G20 emerging market economies.¹⁵ We believe these two benchmarks are reasonable assumptions for approximately where India’s LFPR should be and assume the difference between the current LFPR and this target are discouraged workers. Under these assumptions, we estimate that there were 80 million missing jobs in 2022.

• Estimate the number of jobs needed in 2030: We assume that working age population grows at the rate projected by the UN population, who estimate there will be 1,169 million Indians 15 years and older in 2030. We further assume that the unemployment rate of 5.7 percent remains constant, given our earlier assumption that this is the natural rate of unemployment. We estimate an upper and lower bound of jobs needed based on the current and targeted LFPR, as described above. Wth these assumptions, the lower bound estimate of jobs needed is 60 million and the upper bound estimate is 148 million.

• Estimate the number of jobs needed in 2050: Under the same assumptions as our 2030 projections, with an underlying working age population assumption of 1,337 million, we estimate again a lower and upper bound of jobs needed based on the two LFPR measures. The lower bound estimate of jobs needed is 83 million and the upper bound estimate is 96 million.

• Finally, we sum the estimates from 2022–2050. This gives us a lower bound estimate of 143 million jobs needed and an upper bound estimate of 324 million.¹⁶

Finally, we conduct a simple counterfactual exercise on India’s aggregate growth decomposition to see what would happen to growth if labor became a more important contributor. To do this, we start with our baseline growth decomposition, which is based on the production-function approach using India KLEMS data at the industry level (with a total of 27 industries). We assume this baseline growth and its decomposition is the average contribution of each component over the 2010–19 period. This implies a growth rate of 6.2 percent with the contribution from employment 1.1 percentage points, from labor quality 0.2 percentage points, from capital stock 3.8 percentage points, from capital composition 0.4 percentage points, and from total factor productivity 0.7 percentage points. We then conduct two counterfactual scenarios where labor plays an increasing role in output. In the first scenario, we fix the total number of workers in the economy and assume that 5 percent of the workforce shifts from agriculture to construction and the ten services sub-sectors in the KLEMS data, in equal proportions (we assume all workers find jobs). We then estimate the impact that this increase in employment in services and construction sectors would have on gross value added (GVA) based on the historical elasticity of the growth of GVA to employment growth, which we estimate using ordinary least squares and data from 2010–19 for each aggregate sector. We assume agriculture workers are replaced by machinery and therefore value added does not decline in the sector and labor productivity rises. We then calculate the labor income for each sub-sector based on the estimated GVA growth rate and the 2010–19 average labor share of income for each sub-sector. Finally, we follow the methodology of India KLEMS using the Tornquist aggregation to calculate the new aggregate employment contribution to growth from the rise in employment in the construction and services sub-sectors (see Chattopadhyay, Nath, Sengupta and Joshi (2023) for details). In the second scenario we take the same approach, but instead assume 5 percent of the workforce shifts from agriculture to the 13 manufacturing sub-sectors, and again all workers find jobs.

Results

As previously described, we assume India was at fully employment in 2022 and thus there is no need to create additional jobs. This is our lower-bound estimate. We add to that the number of workers estimated to be discouraged, which we measure as the difference in LFPR between 2005 and 2022 (or, equivalently the difference in LFPR between the G20 emerging market average and India). This is an additional 80 million to the missing jobs in 2022, giving us a range of 0 to 80 million missing jobs in 2022 (Figure 13).

View Full Size

Missing Jobs – Upper Bound Estimate

(millions)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: MoSPI, UN and IMF staff calculations.Note: Numbers cited are rounded to nearest million. Data is based on Current Weekly Status reported in PLFS/NSS.

• Download Figure
• Download figure as PowerPoint slide

Using the UN population projections and the upper bound of our assumptions on LFPR (60 percent), we calculate that India will need to create 148 million jobs by 2030 and 324 million by 2050, cumulatively (that is, 80 million today + 148 million by 2031 + 96 million from 2031–2050, or on average 11.6 million jobs per year between 2022 and 2050). If we instead assume a constant LFPR – which would leave a large share of the population, especially women and youth, without meaningful employment – India would need to instead create 60 million jobs by 2030 and 143 million jobs by 2050, cumulatively, or about 5 million per year. Note that these estimates represent the total number of jobs to be created in the economy, not the additional number of jobs to be created over and above the baseline expectation for job creation that is expected to take place with economic growth. Indeed, India created 12 million jobs annually on average between 2012–23, suggesting that the upper bound estimate, which would bring significantly more working age individuals into the labor force, is within reach if India can sustain the GDP growth rates it saw during this period (averaging around 6 percent).

Underemployment is an important problem in India as it is in many emerging market economies, but it is difficult to quantify. Our estimate of missing jobs may be a lower bound if one also considers underemployment.

The latter includes, for example, those working part time but desiring full-time work, those over skilled for their jobs (see the policy option section), those working in government job guarantee schemes (MNREGA) and those working as unpaid family labor. It would also be important to take into consideration the number of non-farm jobs needed to be created. As the Indian economy continues to shift from farm to non-farm work, employment in the categories of unpaid family worker, own account workers, and casual workers will decline, and more regular wage jobs in urban areas will be needed to absorb this extra labor. Finally, another important determinant of the number of jobs India needs to create also depends on outward migration of workers. Skilled migration from India and the contribution of those workers to the India economy through remittances has been rising over time. If a greater share of the working age population emigrates than was projected in the UN population projections used here, then less jobs will need to be created domestically.

View Full Size

Growth Decomposition: Other Emerging Markets and Advanced Economies

(Percent)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: LA KLEMS; Asia KLEMS; EU KLEMS; and IMF staff estimates.Note: The different KLEMS databases use different measures which attempt to capture labor quality. In the U.S. Mexico, UK, and Peru it is labor composition which is defined as the ratio of labor input to total hours worked. In Korea, and Colombia, it is labor compensation which is based on wage data.

• Download Figure
• Download figure as PowerPoint slide

We next put India’s employment needs in a macro and international context. To do so, we start by conducting a growth decomposition for emerging market peer countries (Mexico, Colombia, and Peru) and three advanced economies (United States, Korea, and United Kingdom) using EU and World KLEMS data and assuming a Cobb-Douglas production function.¹⁷ This will give us perspective for the role that labor plays in contributing to economic output in India versus other countries.

Because of the potentially large number of missing jobs (Figure 13) and large number of low-productivity jobs (see Figure 3), labor’s contribution to growth in India is small relative to other emerging market economies. Emerging market economies tend to have labor inputs as an important driver of growth, as shown in our growth decomposition of peer emerging markets and advanced economies (Figure 14). India’s growth recently has however been more capital driven (see Figure 4 and Kotera and Xu, 2023). Like many other emerging markets, India has a large population of low skilled workers. Unlike other emerging markets, however, growth in India is driven in large part by capital and TFP, making it more comparable to advanced economies. By supporting the growth of large, productive sectors like manufacturing much of India’s labor could be absorbed—and thereby making labor a more important contributor to growth. Furthermore, stronger education and training would help boost the contribution from labor quality, as has been observed in other emerging markets. Our simple counterfactual exercise based on growth decompositions attempts to show the impact of an employment rebalancing on growth (Figure 15). As described above, our baseline scenario is the historical average growth decomposition in India. Under the alternative scenarios, we find that polices that allow a shift of labor out of agriculture boost growth by 0.2–0.5 percentage points. In scenario 2, where 5 percent of the workforce currently in the agriculture sector shifts to the construction and services sectors), we find growth would increase modestly to 6.3 percent. This estimate relies on a few important, but probably unrealistic assumptions. The first is that we assume no increase in unemployment. We also do not assume any endogenous impact on TFP growth from this labor shift. The direction of error from this second assumption is ambiguous – on the one hand, more employed workers in sectors with already higher productivity may help boost TFP further; on the other hand, if there is some initial skill mismatch between jobs and workers, even if workers are employed there may be a drag on productivity (at least temporarily). Scenario 3 assumes the manufacturing sector absorbs the full 5 percent of shifting agriculture workers. Under this assumption, because manufacturing tends to be both more labor intensive and productive, growth would rise further to 6.7 percent. However, the caveats on unemployment and TFP estimates remain.

View Full Size

Growth Decomposition – Counterfactual Scenarios

(percent)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: India KLEMS (RBI); UN; and IMF staff calculations.Notes: Baseline is the 2010–19 average. All scenarios are relative to the baseline. It is assumed that in scenarios 2 and 3 the shift of workers towards construction, services, and manufacturing maintains the 2010–19 average distribution of workers across sub-sectors. Scenarios assume capital, labor quality, and TFP remain constant, and notably do ‘not include the possible endogenous impacts on TFP from labor shifting.

• Download Figure
• Download figure as PowerPoint slide

To realize these gains would require an increase in supply of jobs in more productive sectors like manufacturing, which itself would require additional reforms. This could include reforms to land use, for promoting larger firm sizes and for hiring of more female workers, among others. Furthermore, because workers cannot shift seamlessly between most occupations, this may involve some extent of re-skilling, which could further boost growth via increases in human capital.

Strengthening education and skilling

Despite significant investments in education and some progress in educational enrollment, India’s labor force still has fewer years of formal education than peers and the quality of its education remains low (ASER, 2023).¹⁹ Only 61 percent of the population aged 25 and older have at least completed primary education, compared to an average of 71 percent among lower-middle income peers and 89 percent among upper middle-income countries (Figure 16). This translates into skill shortages in the labor force.

View Full Size

Educational Attainment

(Percent, population aged 25 years or more, most recent observation since 2017)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: World Development Indicators and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Skill mismatch is large in India, with most of the labor force having less formal education than what it would be desirable for the current occupation (Figure 17). We use the PLFS microdata and compute the level of skill mismatch in 2021/22 by comparing worker’s education with the level of formal education that would be desirable given their primary occupation as proposed by ILO’s Education and Mismatch Indicators methodology (ILO, 2023). We find that three quarters of India’s skilled agricultural workers have fewer years of formal education than desirable because they have not completed primary education. Over 60 percent of craft and related workers and plant and machine operators and assemblers and 52 percent of services and sales workers have similarly had too few years of formal schooling. There are also some pockets of overeducation with workers having more years of formal education than what would be desirable for their occupation. This is the case for clerks and technicians, and appears particularly prevalent in Chandigarh, Puduchery, Andaman & N. Island, Haryana, and Delhi.

The Indian authorities are working on addressing these challenges and strengthening education and skilling. The 2020 National Education Policy aims to revamp the education system towards meeting their aspirational goals for the 21^st century (GoI, 2020). The authorities are also working on a dynamic, industry-centric, and forward-looking skilling ecosystem, including through technologically-equipped Industrial Training Institutes with a combined annual enrollment of 1.24 million students, Technical and Vocational Education and Training integration in schools and higher education institutes, apprenticeship for 740 thousand trainees, industry-aligned short-term training, and catalyzing entrepreneurship. They have launched Skill India Digital as a platform for skilling and job matching and have designed courses for new skills such as coding, AI, robotics, mechatronics, IOT, 3D printing, and drones piloting and maintenance. Further efforts in this area can ensure that Indian workers are well equipped to meet the demands of the labor market. Facilitating migration would also help address skill mismatch around the country.

View Full Size

Worker’s Skill Mismatch by Occupation

(percent)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: PLFS and IMF staff estimates.Notes: A worker is defined to have few/more years of formal education than desirable if their education level is lower/greater than usually required for their primary occupation as per the ILO’s Education and Mismatch Indicators methodology. Skill mismatch is determined by using ILO’s Education and Mismatch Indicators methodology (ILO, 2023) on employment and occupation data by worker from the PLFS. For instance, a worker employed in an elementary occupation is considered to have adequate years of formal education if s/he has completed primary education and is considered to have more/fewer years than desirable if s/he has more/fewer years of formal education than completed primary.

• Download Figure
• Download figure as PowerPoint slide

Advancing labor market reforms

India approved four new labor codes at the federal level, which streamline significantly labor regulations in the country. However, implementation of these reforms has not been completed yet as notification of the rules by the Centre and state governments is still pending. Evidence shows that in India, these strict labor market regulations have contributed to a misallocation of labor (Mohommad and others, 2021; Besley and Burguess, 2004; Dougherty and others, 2014). Furthermore, loosening restrictive employment protection legislation would lead to an increase in total employment over time (including female employment), though it may briefly cause a decline in employment during the initial shift of labor across sectors or jobs (Agarwal and others, 2022). As such, working with states to accelerate the reform momentum in this area would be essential to further enhance labor market flexibility while still providing adequate protection for workers.

Fostering trade integration

Removing trade barriers remains a key priority to further advance trade integration and boost productivity, labor market outcomes, and ultimately, growth (Salgado and Anand, 2022). The authorities are making steady progress in negotiating new bilateral trade agreements that would open new markets for Indian exports, but tariff and non-tariff import and export restrictions remain high and extensive. Removing trade restrictions would subject Indian producers to healthy competition that would help them move closer to the technological frontier (Dabla-Norris and others, 2023, Topalova and Khandelwal, 2020) and provide for better allocation of resources.

Removing red tape and other obstacles to private sector growth

The authorities’ efforts to remove bureaucratic inefficiencies and improve the business climate are supported by their digitalization push. For example, the launch of the Udyam Portal aims to streamline registration for MSMEs. In addition, their Production Linked Incentive schemes (PLIs) are being perceived by market participants as a signal of the government’s intention to work with the private sector to remove obstacles holding back investments and employment. PLIs provide subsidies to firms that meet investment and employment targets in specific sectors. The multiyear cost of the existing PLIs assuming full implementation is relatively low at 0.72 percent of FY 2022/23 GDP and the authorities estimate they may help create up to 6 million jobs (Figure 18; GoI, 2022). While it is too early to assess these schemes and the data is not available yet, targeted and temporary industrial policy could help address market failures. At the same time, however, it would be important to subject these schemes to rigorous cost benefit analysis, to keep them time bound, and to ensure competition is fostered. Importantly, PLIs do not replace broader market reforms, which should continue to be prioritized.

View Full Size

Fiscal Cost of Production-Linked Schemes

(Percent of 2022/23 GDP)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: Indian authorities and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Continuing the public investment push

Efforts by the Union government are expected to raise public investment by the center to 3.3 percent of GDP in 2023/24, twice as much as in 2018/19 (Figure 19). Stronger physical public infrastructure, together with India’s world class digital public infrastructure, will help increase the productivity of the private sector.

View Full Size

Central Government Public Investment

(Percent of GDP)

Citation: IMF Working Papers 2024, 138; 10.5089/9798400281617.001.A001

Source: Indian authorities and IMF staff estimates.Notes: Net acquisition of nonfinancial assets by the central government. 2023/24(P) reflects IMF staff projections.

• Download Figure
• Download figure as PowerPoint slide

Strengthening the social safety net

A more flexible and targeted social safety net can help support the structural transformation of India’s economy and facilitate migration of workers from rural to urban areas in search of better paying jobs (Alonso and others, 2023). For example, the rollout of “One Nation, One Ration Card” during the COVID 19 pandemic allowed migrant workers to benefit from their food subsidies wherever they were at the time.

Facilitating access to credit

Channeling credit to the most productive firms and sectors of the economy is key to foster growth and employment. The health of the financial and corporate sector has improved significantly over the last few years, but continued nimble and vigilant supervision would be needed to ensure healthy credit growth, supporting a rise in private investment (Schipke and others. 2023). In addition, facilitating the exit of non-viable firms and promoting asset resolution remain core priorities.