## Appendix 1. Growth Accounting by KLEMS

## Source details

**Canonical URL:** [Appendix 1. Growth Accounting by KLEMS](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023082.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023082.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023082.pdf.json)

---

### Growth accounting framework and comparison with PWT
- KLEMS growth accounting expression:
  - 푦푡 = 푎푡 + 훼푡 × (푧푡 + 푞푘푡) + (1 − 훼푡) × (푙푡 + 푞푙푡)
  - Definitions:
    - 푎푡 is TFP growth.
    - 푧푡 is capital stock growth.
    - 푞푘푡 is capital quality growth.
    - 푙푡 is employment growth.
    - 푞푙푡 is labor quality growth.
- Capital input:
  - KLEMS considers capital service growth as (푧푡 + 푞푘푡).
  - Capital input series are estimated from investment data in the National Accounts (consistent with PWT and KLEMS approaches).
- Labor input and labor quality differences:
  - KLEMS does not include working hours per worker; PWT includes working hours.
  - Both KLEMS and PWT rely on official labor force surveys to estimate employment levels; PWT uses ILO estimates from 2017 to 2019.
  - KLEMS captures labor quality via earnings of five educational groups; PWT captures human capital via average years of schooling for population aged 25+.
- Aggregation and comparison:
  - KLEMS uses the Tornqvist index to aggregate industry-level real value added, capital, and labor growth rates into an economy-wide series; PWT uses the economy-wide growth rate directly.
  - The KLEMS exercise shows trends similar to PWT-based growth accounting.

### Figure A.1 (value added — growth accounting) — visual components and axis ticks
- Components displayed in the figure:
  - Employment
  - Labour Quality
  - Capital
  - TFP
- Axis ticks shown in the figure:
  - -4.0%
  - -2.0%
  - 0.0%
  - 2.0%
  - 4.0%
  - 6.0%
  - 8.0%
  - 10.0%
  - 12.0%
- Figure source: Reserve Bank of India

### Key interpretive points from growth accounting (as reported in source overview)
- Historical shifts in factor contributions (1971–2019):
  - 1970s–1980s: labor was the major growth driver.
  - Late 1990s–2000s: capital contribution strengthened; capital predominant from 2005 to 2010 while labor contribution fell.
  - Pre-pandemic decade: increasing importance of TFP alongside capital.
- Relationship to reforms and structural change:
  - Increased capital and TFP contributions associated with 1991 market reforms and pro-business reforms in the 1980s; recent TFP gains linked to FDI and services growth.
- Sectoral reallocation and TFP channel:
  - Pandemic-induced labor reallocation from higher-productivity sectors (industry) to lower-productivity sector (agriculture) contributed to labor productivity and TFP declines.

### Appendix 2. Employment Estimation

### Data sources, timing, and challenges
- Data sources used:
  - Quarterly PLFS (urban only).
  - Annual PLFS (July to June reference period; latest available: July 2020 - June 2021).
  - CMIE-CPHS for capturing quarterly fluctuations, especially in rural areas where quarterly PLFS is unavailable.
- Challenges:
  - Quarterly PLFS available only for urban areas; rural quarterly series constructed by combining annual PLFS and CMIE-CPHS.
  - CMIE-CPHS sample size dropped substantially after the pandemic and exhibited composition changes over time.

### Estimation procedure for rural employment (three steps)
- Step 1 — Base-year EPR
  - Set CMIE-CPHS employment-to-population ratio (EPR) in 2019 as the base year.
  - Estimate EPR in pandemic years (2020 and 2021) by applying changes observed in the common sample/respondents to the 2019 level.
- Step 2 — Sample weight adjustment to match projected population distribution
  - Final weight formula:
    - 푓푤푖푗푘 = 푠푗푘 ∗ 푁 / 푁푗푘 ∗ 푝푤푖
    - Definitions:
      - 푓푤푖푗푘 is the final weight for individual (i) in gender (j) and age group (k).
      - 푝푤푖 is the CMIE-CPHS weight adjusted for non-response for individual (i).
      - 푠푗푘 is the fraction of the projected population for category (j, k).
      - 푁 is the sum of all 푝푤푖.
      - 푁푗푘 is the sum of 푝푤푖 for category (j, k).
  - Population projection source: National Commission on Population (2020).
- Step 3 — Level adjustment to align with PLFS
  - Compute the EPR difference between the modified CMIE-CPHS and annual PLFS for the common period (July - June).
  - Apply the difference to adjust the CMIE-CPHS series so levels match official PLFS data.
- Combining rural and urban to produce total employment
  - Combine adjusted rural EPR and urban EPR (urban from quarterly PLFS) using rural/urban population shares from National Commission on Population (2020).

### CMIE-CPHS sample behavior and weighting categories
- Sample size changes:
  - CMIE-CPHS sample size reduced to about one-third of the pre-pandemic level in April 2020.
  - Sample size dropped again during the second wave; by December 2021 it was about 90% of the pre-pandemic level.
- Composition control:
  - Using the same composition of respondents (common sample) is applied to address changing respondent composition over time.
- Weighting categories:
  - Gender (j): two categories — male and female.
  - Age group (k): 11 categories — 15-19, 20-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-59, 55-59, 60-64, and 65+.

### Employment and labor-input-related quantitative findings cited in source
- Pandemic-era macro and sector impacts (FY2020/21 and 2020 observations):
  - Investment contracted about 10 percent in 2020.
  - Private consumption contracted -6 percent.
  - Overall GDP contracted -6.6 percent.
  - Contact-intensive services (about 20 percent of GDP) contracted about 20 percent in 2020.
  - Mining and construction (about 10 percent of GDP) contracted about 8 percent in 2020.
  - Agriculture (about 15 percent of GDP) continued to grow in 2020.
  - By March 2022, all sectors returned to pre-pandemic levels except contact-intensive services.
- Labor-market near-term impacts:
  - Employment-to-population ratio (EPR) declined by about 7 percent in urban areas in FY2020.
  - Hours per worker contracted by 20 percent in 2020Q2 compared with the previous year.
  - Casual worker job losses: about 60% lost employment in the first wave and 30% in the second wave (urban).
- Medium-term labor and human-capital indicators and estimates:
  - Baseline employment growth: 0.6 percent annually versus working-age population growth 2 percent annually (pre-pandemic).
  - Estimated total working hours lower than pre-pandemic trend by about 12 and 4 percent in FY2020 and FY2021, respectively.
  - One additional year of work experience associated with roughly 2.7 percent increase in wage (Das, 2019).
  - Human capital growth decreased by about 0.3 and 0.1 percent in FY2020 and FY2021, respectively.
  - Schools fully open for only 47 days from 16 February 2020 to 31 March 2022 in India (UNESCO).
  - Internet penetration in 2019: 54% urban and 32% rural.
  - About three percent of the working-age population benefits from formal vocational training every year (PLFS).
  - Formal vocational training estimated to enhance wage growth by 4.7% in India (Kumar et al., 2019).

### Potential-growth quantitative conclusions referenced in appendices and main text
- Baseline medium-term potential growth:
  - Potential growth estimated to be about 6 percent in the medium term (2027) under the baseline.
- Upside (reform) scenario:
  - Successful structural reforms could raise medium-term potential growth to about 7 percent.
  - Compared with the baseline, each factor’s contribution (capital, labor, TFP) would increase by around 0.3 to 0.4 percentage points under the upside scenario.
  - RBI comparative range noted: 6.5 percent to 8.5 percent.

*Source: Appendix 1 and Appendix 2 from wpiea2023082 - Appendix 1. Growth Accounting by KLEMS*

### Appendix 1. Growth Accounting by KLEMS ......................................................................... 19

### Appendix 1. Growth Accounting by KLEMS

### Appendix 1: Growth Accounting by KLEMS (page 19)
- Title in source: "Appendix 1. Growth Accounting by KLEMS"
- Location in source PDF: page 19

### Appendix 2: Employment Estimation (page 20)
- Title in source: "Appendix 2. Employment Estimation"
- Location in source PDF: page 20

*Source: wpiea2023082 - Appendix 1. Growth Accounting by KLEMS; PDF pages 19–20.*

### References .............................................................................................................

### References

### I. Introduction
- Paper objective: analyze drivers of India’s growth and potential growth over the past 50 years and through the pandemic, with three main objectives:
  - Examine role of labor, capital, human capital, and TFP in India’s growth over the past 50 years and compare with fast-growing economies.
  - Estimate the impact of the pandemic on potential growth and project medium-term potential growth, accounting for pandemic channels.
  - Consider baseline and upside medium-term potential growth scenarios; upside relies on structural reforms to unleash growth potential.
- Key literature strands referenced: studies on India’s potential growth drivers, pandemic impact on productivity and medium-term growth, and role of structural reforms on output.
- High-level result summary:
  - Pandemic affected labor, human capital, physical capital, and total factor productivity (TFP), producing some medium-term impact under the baseline.
  - Upside scenario: successful structural reforms could more than offset pandemic-induced losses.

### II. Data and Methodology
- Data sources:
  - Historical growth accounting 1971–2019: Penn World Table (PWT) 10.0 (latest observation 2019).
  - Pandemic-period sources: National Accounts (Haver Analytics/Central Statistical Office), Periodic Labor Force Survey (PLFS), United Nations/World Bank World Development Index, Census of India, CMIE Consumer Pyramids Household Survey (CMIE-CPHS).
- Production function used:
  - Augmented Cobb-Douglas in logs: y_t = a_t + α × k_t + (1 − α) × (n_t + h_t)
    - y_t: potential growth rate
    - a_t: TFP growth rate
    - k_t: capital growth rate
    - n_t: labor input (total number of hours worked by all employed persons) growth rate
    - h_t: human capital growth rate
    - α: share of capital
- Capital accumulation:
  - Perpetual inventory: K_{t+1} = (1 − δ_t^K) K_t + I_t^K
  - Investment growth consistent with WEO projections, reflecting initial contraction in 2020 and rebound thereafter.
- Labor input:
  - Employment-to-population ratio for FY2020 and FY2021 estimated using quarterly PLFS for urban, annual PLFS and CMIE-CPHS for rural.
  - Adjustments made for CMIE-CPHS sample size and level differences with PLFS to ensure consistency.
  - Forecast assumption: employment-to-population ratio returns to pre-pandemic (2011-2019) trend; UN population projections used to derive employed persons.
  - Working hours per worker based on average hours by employment status from annual PLFS (till 2Q 2021); hours assumed to gradually recover to pre-pandemic level.
- Human capital:
  - Penn World Table approach for human capital Φ(s):
    - Φ(s) = 0.134*s if s ≤ 4
    - Φ(s) = 0.134*4 + 0.101*(s − 4) if 4 < s ≤ 8
    - Φ(s) = 0.134*4 + 0.101*4 + 0.068*(s − 4) if s > 8
    - s = average years of schooling for adults above 25 (Barro and Lee (2013) data used for India)
  - Pandemic human capital channels considered: lost years of schooling and forgone on-the-job training.
  - Impact of school closures on long-run growth not factored into medium-term estimates in this paper.
- TFP:
  - Consider decline in productivity due to labor reallocation from more productive sectors (industry) to less productive (agriculture).
  - Labor productivity = gross value added / employment by sector.
  - Relation between labor productivity growth and TFP growth estimated from historical 1985–2019 data.
  - Other TFP channels (resource mismatches, firm exit, R&D decline) noted but not quantified due to limited data.

### III. What Drove India’s Past Growth?
- Statistical phases over 1971–2019:
  - 1970s: low growth with inward-looking policies.
  - 1980–2002: about 5.5 percent average growth with start of liberalization.
  - Early 2000s–pre-pandemic: high growth period.
- Factor contributions (1971–2019 observations):
  - 1970s–1980s: labor major driver.
  - Late 1990s–2000s: capital contribution picked up; capital predominant from 2005 to 2010 while labor contribution fell.
  - Pre-pandemic decade: growing importance of TFP alongside capital.
- Interpretation:
  - Increased capital and TFP contributions associated with 1991 market reforms and pro-business reforms in 1980s; recent TFP gains possibly linked to increased FDI and services sector growth.
  - Despite rapid growth, job creation scope relatively limited; labor force participation declined, especially female participation.

### IV. Baseline Scenario: Impact of the Pandemic
A. Near-term impact
- Macroeconomic contractions in 2020 (FY2020/21):
  - Investment contracted about 10 percent in 2020.
  - Private consumption contracted -6 percent.
  - Overall GDP contracted -6.6 percent.
- Sectoral impacts in 2020:
  - Contact-intensive services (about 20 percent of GDP) contracted about 20 percent.
  - Mining and construction (about 10 percent of GDP) contracted about 8 percent.
  - Agriculture (about 15 percent of GDP) continued to grow in 2020.
  - By March 2022, all sectors returned to pre-pandemic levels except contact-intensive services.
- Labor market near-term impacts:
  - Employment-to-population ratio (EPR) declined by about 7 percent in urban areas in FY2020.
  - Hours per worker contracted by 20 percent in 2020Q2 compared with the previous year.
  - Similar reductions observed in 2021Q2, but smaller magnitude than initial wave; Omicron early 2022 had almost no macro labor-market impact.
- Uneven impact by worker type and demographic:
  - About 60% of casual workers in urban areas lost employment during the first wave and 30% during the second wave (roughly four to five times higher than self-employed and regular wage employees).
  - Vulnerable groups more affected: casual workers, females, youth, lower-skilled.
  - Migrant workers returning to home villages transitioned into agricultural work or became unemployed with lower income.

B. Medium-term impact (four channels)
- Four channels through which pandemic affects potential growth: capital input, labor input, human capital stock, and TFP.
- Capital:
  - Investment rebounded strongly after 2020 contraction and assumed to remain relatively robust in the medium term.
  - Capital stock growth assumed to gradually converge back to rate prior to NBFC crisis.
  - Public infrastructure investment planned in coming years expected to support capital accumulation.
  - Baseline assumes resilient credit growth supporting investment.
- Labor:
  - Baseline assumption: employment will grow consistent with pre-pandemic trend (2011-19), implying a decreasing employment-to-population ratio under the baseline.
  - Structural factors:
    - Declining female employment ratio pre-pandemic; employment growth estimated 0.6 percent annually vs working-age population growth 2 percent annually (pre-pandemic).
    - Limited employment opportunities for youth: over eight years before pandemic, 44 million more young people completed more than upper primary education but only 10 million found employment; 14 million struggled to find a job; 21 million focused on domestic duties.
- Human capital:
  - Estimated total working hours lower than pre-pandemic trend by about 12 and 4 percent in FY2020 and FY2021, respectively.
  - Literature: one additional year of work experience can lead to roughly 2.7 percent increase in wage (Das, 2019).
  - Paper estimates human capital growth decreased by about 0.3 and 0.1 percent in FY2020 and FY2021, respectively.
  - Human capital growth projected to recover as labor returns to pre-pandemic trends, but accumulation level remains lower than pre-pandemic trend without additional vocational training.
  - School closure context:
    - UNESCO: schools were fully open for only 47 days from 16 February 2020 to 31 March 2022 in India.
    - Internet penetration: 54% urban and 32% rural in 2019; uneven access implies uneven learning losses across income groups.
- TFP:
  - Labor productivity declined due to reallocation from higher-productivity sectors to agriculture.
  - Pre-pandemic labor productivity in agriculture is the lowest among sectors.
  - Assumption: TFP growth will gradually recover as labor returns to higher productive sectors; digitalization expected to support TFP improvements.
- Baseline medium-term potential growth estimate:
  - Potential growth estimated to be about 6 percent in the medium term (2027) under the baseline scenario.
  - Baseline potential growth is lower than a counterfactual without the pandemic and subsequent geopolitical shocks.
  - Capital and TFP are the main drivers of medium-term potential growth in the baseline; labor contribution relatively small despite demographic dividend.

### V. Upside Scenario: Reform Dividends
- Upside scenario assumptions and reform channels:
  - Investment/Capital: investment-friendly policies, continued infrastructure investment, easing FDI regulation to raise capital accumulation and potential growth.
  - Labor:
    - Reforms to improve female labor force participation and reduce youth unemployment can slow EPR decline and unlock demographic dividend.
    - Implementation examples: swift implementation of past reforms (e.g., new labor codes by states), ease administrative bottlenecks, support formalization, improve targeting of social benefits.
    - Enhancing non-agricultural job opportunities in rural areas critical to increase female labor force participation.
  - Human capital:
    - Strengthen vocational training and education; currently about three percent of working-age population benefits from formal vocational training every year (PLFS).
    - Empirical finding: formal vocational training can enhance wage growth by 4.7% in India (Kumar et al., 2019).
  - TFP:
    - Improve business environment to shift labor from agriculture to industry/services; advance agriculture and land reforms; progress in formalization and digitalization; reduce digital divide.
    - Digitalization cited as having potential to increase productivity across sectors (MeitY (2019) identified 30 digital themes).
    - Climate-related sectoral adaptation policies could support productivity in long run.
- Upside scenario quantitative result:
  - Successful implementation of wide-ranging structural reforms could raise medium-term potential growth to about 7 percent, more than offsetting persistent pandemic impacts.
  - Compared with baseline, each factor’s contribution (capital, labor, TFP) would increase by around 0.3 to 0.4 percentage points under the upside scenario.
  - Note: estimate range comparable with RBI’s estimates (6.5 percent to 8.5 percent) and contrasts with other higher estimates cited.

### VI. Conclusion and Policy Implications
- Historical summary:
  - 1970s–80s: growth mainly driven by labor.
  - 1990s–2000s: capital key driver.
  - Past decade pre-slowdown: TFP growth picked up.
  - India has accumulated productive physical capital, supporting growth and structural transformation.
- Main findings:
  - Pandemic had multi-channel adverse effects, producing some medium-term impact on potential growth under baseline.
  - Structural reforms can more than offset pandemic impact and support medium-term potential growth.
- Policy considerations:
  - Support investment-friendly policies to sustain capital accumulation and potential growth.
  - Implement reforms to improve female labor force participation and reduce youth unemployment to unleash labor market potential.
  - Strengthen vocational and education policies to address pandemic-related learning losses, especially for poorer households.
  - Pursue wide-ranging structural reforms (agriculture, land, business environment), continued digitalization, and measures to improve productivity in medium and long run.

*International Monetary Fund — Unleashing India’s Growth Potential (content unit: References).*

### Appendix 1. Growth Accounting by KLEMS

### Appendix 1. Growth Accounting by KLEMS

### Growth accounting framework
- The KLEMS growth accounting exercise is expressed as:
  - 푦푡 = 푎푡 + 훼푡 × (푧푡 + 푞푘푡) + (1 − 훼푡) × (푙푡 + 푞푙푡)
  - Definitions:
    - 푎푡 is TFP growth.
    - 푧푡 is capital stock growth.
    - 푞푘푡 is capital quality growth.
    - 푙푡 is employment growth.
    - 푞푙푡 is labor quality growth.
- Capital input:
  - KLEMS, similar to Penn World Tables (PWT), considers capital service growth expressed as (푧푡 + 푞푘푡).
  - Capital input series are estimated based on investment data from the National Accounts for both PWT and KLEMS.
- Labor input differences between KLEMS and PWT:
  - KLEMS does not include working hours per worker.
  - Both PWT and KLEMS rely on the official labor force survey to estimate employment level; PWT uses ILO estimates from 2017 to 2019.
  - Labor quality:
    - KLEMS captures labor quality by considering earnings of five educational groups.
    - PWT captures human capital based on average years of schooling for those aged 25 and above.
- Aggregation:
  - KLEMS applies the Tornqvist index to aggregate the real value added, capital, and labor growth rates of each industry to compute the economy-wide growth rate.
  - PWT uses the growth rate of the entire economy.
- Comparison:
  - The KLEMS growth accounting exercise (presented in Figure A.1) suggested similar trends as those based on PWT.

### Figure A.1: Value Added — Growth Accounting (Total Economy, 2011-12 Price)
- Components shown:
  - Employment
  - Labour Quality
  - Capital
  - TFP
- Axis ticks presented in the figure:
  - -4.0%
  - -2.0%
  - 0.0%
  - 2.0%
  - 4.0%
  - 6.0%
  - 8.0%
  - 10.0%
  - 12.0%
- Source noted in figure: Reserve Bank of India

### Appendix 2. Employment Estimation

### Data sources and challenge
- Quarterly PLFS is available only for urban areas.
- To estimate rural employment, both annual Periodic Labor Force Survey (PLFS) and Centre for Monitoring Indian Economy - Consumer Pyramids Household Survey (CMIE-CPHS) are used.
- Annual PLFS survey period: July to June (does not match fiscal year April to March).
- Latest available annual PLFS at time of writing: July 2020 - June 2021.
- Approach: rely on PLFS for annual employment level and CMIE-CPHS for capturing quarterly fluctuations.
- Adjustment: sample weights of CMIE-CPHS are adjusted to mitigate sample size fluctuations and ensure representativeness after the pandemic.

### Estimation steps for rural employment
- Step 1: Base-year EPR
  - Set the employment-to-population ratio (EPR) of CMIE-CPHS in 2019 as the base year.
  - Estimate EPR in pandemic years (2020 and 2021) by using common samples/respondents:
    - Apply the change in EPR in the common sample to the level of the employment ratio in 2019 to estimate the EPR in the pandemic years.
- Step 2: Sample weight adjustment to match projected population distribution
  - Final weight formula:
    - 푓푤푖푗푘 = 푠푗푘 ∗ 푁 / 푁푗푘 ∗ 푝푤푖
    - Definitions:
      - 푓푤푖푗푘 is the final weight used for individual (i) in gender (j) and age group (k).
      - 푝푤푖 is the CMIE-CPHS weight adjusted for the non-response factor for individual (i).
      - 푠푗푘 is the fraction of the projected population for the category (j, k).
      - 푁 is the sum of all 푝푤푖.
      - 푁푗푘 is the sum of 푝푤푖 for the category (j, k).
  - Population projection source: National Commission on Population (2020).
- Step 3: Level adjustment to align with PLFS
  - Calculate the difference in EPR between the modified CMIE-CPHS and annual PLFS for the common period (July - June).
  - Apply the difference to the adjusted CMIE-CPHS series so CMIE-CPHS levels are consistent with official PLFS data.
- Combining rural and urban for total employment:
  - Adjusted EPR of rural areas and EPR of urban areas (taken from quarterly PLFS) are combined using the rural and urban population ratio from National Commission on Population (2020).

### Notes on CMIE-CPHS sample behavior and categories
- Sample size behavior:
  - Available sample size of CMIE-CPHS reduced significantly after the pandemic (about one-third of the pre-pandemic level in April 2020).
  - Samples dropped again during the second wave; size was about 90% of the pre-pandemic level in December 2021.
  - Using the same composition of respondents aims to address concerns from differing compositions of individuals over time.
- Gender and age categories used in weighting:
  - Gender (j): two categories (male and female).
  - Age group (k): 11 categories — 15-19, 20-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-59, 55-59, 60-64, and 65+.

*Source: Appendix 1 and Appendix 2 from wpiea2023082 - Appendix 1. Growth Accounting by KLEMS*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023082.pdf_