## Public Administration Digitalisation and Microenterprise Productivity in India, Working Paper No. WP/2026/086

## Source details

**Canonical URL:** [Public Administration Digitalisation and Microenterprise Productivity in India, Working Paper No. WP/2026/086](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026086-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026086-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026086-source-pdf.pdf.json)

---

### Introduction: research question, approach, and main findings
- Research question and approach:
  - Investigates whether public administration digitalisation, which eases regulatory and policy burdens on business, leads to higher productivity growth and lower productivity dispersion among firms in India.
  - Focus: micro-enterprises covered in the Unincorporated Enterprise Surveys of India (two rounds: 2010-11 and 2015-16).
  - Reform identification: Assessment of State Implementation of Business Reforms Report of 2015; assessment of public administration digitalisation undertaken in each state.
  - Empirical strategy (illustrative): For a firm in a more-digitalised state, find a firm with similar characteristics in a less-reformed state and compare total factor productivities (TFPs); compare TFP dispersion across pairwise states; robustness check using firms located in bordering districts of neighbouring state pairs.
- Main findings (as stated):
  - States that undertake more public administration digitalisation experience higher productivity growth and lower productivity dispersion among firms.
  - Digitalisation of public administration scaled up in the mid-2010s, building on institutional initiatives and technological infrastructure developed earlier.
  - The 98-point Action Plan on Ease of Doing Business was finalized in 2014; an evaluation of implementation was done in June 2015.
- Rationale and mechanisms (summary):
  - Digitalisation reduces compliance costs (permits, filing taxes, accessing government schemes), disproportionately benefiting microenterprises.
  - Digitalisation reduces corruption and informal cost variations by automating and increasing transparency.
  - At the macro level, digitalisation can enhance aggregate productivity by better allocating resources and supporting higher-productivity firms.

### Business Environment Reforms 2014-15: design, coverage, and exogeneity
- Reform design and classification:
  - On 29th December 2014, Chief Secretaries finalized a 98-point Action Plan on Ease of Doing Business; digitalisation of public administration was a key part.
  - Implementation assessment completed in June 2015.
  - The 98 points are classified into six reform categories: Tax filing and payments (TR), Construction permits (CPR), Environment and labour standards (ELR), Inspections (IR), Commercial dispute (CDR), Single Window Systems (SWR).
- States implementing reform categories (selected lists as in source):
  - Tax filing and payments: Bihar, Karnataka, Madhya Pradesh, Chhattisgarh, Gujarat, Maharashtra, Odisha, Uttar Pradesh, Punjab, Rajasthan, Uttarakhand, West Bengal.
  - Construction permits: Andhra Pradesh, Chhattisgarh, Gujarat, Haryana, Madhya Pradesh, Maharashtra, Rajasthan, Telangana, Uttarakhand.
  - Environment and labour standards: Andhra Pradesh, Bihar, Chhattisgarh, Gujarat, Haryana, Karnataka, Madhya Pradesh, Maharashtra, Jharkhand, Odisha, Punjab, Rajasthan, Tamil Nadu, Telangana, Uttar Pradesh, Uttarakhand, West Bengal.
  - Inspections: Andhra Pradesh, Chhattisgarh, Gujarat, Haryana, Karnataka, Madhya Pradesh, Odisha, Jharkhand, Punjab, Rajasthan, Punjab, Telangana, Uttar Pradesh, Uttarakhand.
  - Commercial disputes: Andhra Pradesh, Chhattisgarh, Gujarat, Haryana, Madhya Pradesh, Maharashtra, Jharkhand, Rajasthan, Telangana.
  - Single Window System: Andhra Pradesh, Chhattisgarh, Jharkhand, Gujarat, Haryana, Madhya Pradesh, Maharashtra, Rajasthan, Telangana, Punjab, Uttarakhand.
- Exogeneity assessment:
  - Implementation not systematically associated with baseline state development characteristics.
  - OLS regressions of number of reforms on development proxies (N = 19):
    - Average SGDP Growth coefficient: 0.546 (standard error (1.620)); adj. R-sq: 0.083
    - Average Per-Capita SGDP Growth coefficient: 0.234 (standard error (0.700)); adj. R-sq: 0.029
    - Share of Industry in State GVA coefficient: 0.054 (standard error (1.070)); adj. R-sq: 0.008

### Data: surveys, sample scope, and enterprise composition
- Data source and sample restrictions:
  - National Sample Surveys (NSS) on unincorporated microenterprises; restricted to non-agricultural enterprises and excluding construction; further restricted to non-financial enterprises.
  - NSS rounds used:
    - 67th round: July 2010–June 2011
    - 73rd round: July 2015–June 2016
  - NSS 67th round: first round covering the entire unincorporated non-agricultural sector (excluding construction) as follow-up to Economic Census; 73rd round is a repeat.
- Distribution of Active MSMEs by Type (in 1000):
  - Proprietary: Registered 1409; Unregistered 32745; Total 34154; % of Unregistered Enterprises 95.87
  - Partnership: Registered 63; Unregistered 365; Total 428; % of Unregistered Enterprises 85.28
  - Private Company: Registered 43; Unregistered 6; Total 49; % of Unregistered Enterprises 12.24
  - Co-operative: Registered 5; Unregistered 116; Total 121; % of Unregistered Enterprises 95.87
  - Others: Registered 44; Unregistered 765; Total 809; % of Unregistered Enterprises 94.56
  - Not recorded: Registered 0; Unregistered 615; Total 615; % of Unregistered Enterprises 100
  - Total: Registered 1564; Unregistered 34612; Total 36176; % of Unregistered Enterprises 95.68
  - Note: All India Census of MSME, Fourth Round, 2006-2007 reports unregistered enterprises constitute around 96 percent of total MSMEs.
- NSS totals by round:
  - 67th Round (July 2010–June 2011): Total Enterprises 334,475; Non-financial Enterprises 329,280; Manufacturing Enterprises 96,414
  - 73rd Round (July 2015–June 2016): Total Enterprises 288,676; Non-financial Enterprises 285,389; Manufacturing Enterprises 79,966

### Estimation of TFPR: method, imputation, and state statistics
- TFPR estimation method:
  - Revenue-based TFPR: TFPR_ijt = p_ijt A_ijt = p_ijt y_ijt / (k_ijt^α_jt l_ijt^(1−α_jt)) (equation (1)).
  - Labour share 1−α_jt computed as mean(labour cost_jt / net value added_jt) (equation (2)); wage payments used as labour costs.
  - Nominal net value added measured as total output minus value of total inputs, net of total fixed asset depreciation during the year.
  - Inflation not adjusted in TFPR values; year-fixed effects used in regressions.
- Labour cost imputation and sample treatment:
  - State-wise minimum daily wage rates for unskilled labour in 2010 and 2015 from Indiastat converted to monthly by assuming 20 workdays per month.
  - Robustness check: actual wage payments used for firms employing 10 or more workers.
  - For self-employed micro and small enterprises with no wage data, the state government's minimum wage is used as imputed wage.
  - For firms where owners are not counted as workers, 1 is added to number of workers.
- State-wise average log TFPR: selected exact figures (2010-11 and 2015-16):
  - Andhra Pradesh: Mean (2010-11) -0.8472.617  Min. -11.891 Max. 18.975 ; Mean (2015-16) 0.32  Std. Dev 1.791  Min. -7.703 Max. 10.101
  - Bihar: Mean (2010-11) 0.4252.369  Min. -11.482 Max. 28.344 ; Mean (2015-16) 1.047  Std. Dev 1.604  Min. -7.611 Max. 12.861
  - Chhattisgarh: Mean (2010-11) -0.2372.219  Min. -10.634 Max. 10.168 ; Mean (2015-16) 0.288  Std. Dev 1.438  Min. -5.788 Max. 5.032
  - Gujarat: Mean (2010-11) -0.6852.443  Min. -15.965 Max. 16.406 ; Mean (2015-16) -0.219  Std. Dev 1.583  Min. -6.726 Max. .595
  - Haryana: Mean (2010-11) -1.3122.573  Min. -12.933 Max. 15.321 ; Mean (2015-16) 0.707  Std. Dev 1.306  Min. -4.883 Max. 4.689
  - Himachal Pradesh: Mean (2010-11) -1.42.672  Min. -13.022 Max. 15.769 ; Mean (2015-16) 0.074  Std. Dev 1.929  Min. -15.345 Max. 6.051
  - Jharkhand: Mean (2010-11) 0.802  Std. Dev 2.473  Min. -7.528 Max. 24.875 ; Mean (2015-16) 0.676  Std. Dev 1.376  Min. -6.437 Max. 6.397
  - Karnataka: Mean (2010-11) -0.262.581  Min. -13.273 Max. 14.146 ; Mean (2015-16) 0.317  Std. Dev 1.525  Min. -6.146 Max. 7.276
  - Kerala: Mean (2010-11) 0.167  Std. Dev 2.865  Min. -11.382 Max. 5.826 ; Mean (2015-16) 0.549  Std. Dev 1.667  Min. -5.868 Max. 8.8
  - Madhya Pradesh: Mean (2010-11) 0.073  Std. Dev 2.681  Min. -10.776 Max. 26.998 ; Mean (2015-16) 0.318  Std. Dev 1.833  Min. -7.412 Max. 12.872
  - Maharashtra: Mean (2010-11) -0.6762.916  Min. -17.654 Max. 29.009 ; Mean (2015-16) 0.041  Std. Dev 1.77  Min. -8.664 Max. 9.336
  - Odisha: Mean (2010-11) 1.522.794  Min. -7.888 Max. 27.815 ; Mean (2015-16) 1.083  Std. Dev 1.43  Min. -4.724 Max. 9.605
  - Punjab: Mean (2010-11) -1.2932.652  Min. -13.933 Max. 24.675 ; Mean (2015-16) 0.43  Std. Dev 1.368  Min. -6.912 Max. 5.298
  - Rajasthan: Mean (2010-11) -1.1582.301  Min. -12.544 Max. 9.632 ; Mean (2015-16) 0.363  Std. Dev 1.472  Min. -6.639 Max. 4.453
  - Tamil Nadu: Mean (2010-11) -0.9162.528  Min. -14.634 Max. 14.3 ; Mean (2015-16) 0.326  Std. Dev 1.652  Min. -7.351 Max. 10.215
  - Uttar Pradesh: Mean (2010-11) -0.4572.717  Min. -12.471 Max. 29.203 ; Mean (2015-16) 0.309  Std. Dev 1.508  Min. -6.448 Max. 12.27
  - Uttarakhand: Mean (2010-11) -1.182.317  Min. -10.027 Max. 8.32 ; Mean (2015-16) 0.369  Std. Dev 1.285  Min. -5.457 Max. 4.26
  - West Bengal: Mean (2010-11) 0.762  Std. Dev 2.944  Min. -13.232 Max. 28.592 ; Mean (2015-16) 1.276  Std. Dev 1.835  Min. -6.233 Max. 12.773
- Aggregate observation: between 2010-11 and 2015-16 there is an improvement in overall TFPR across the states (inflation not adjusted in reported TFPR but handled in regressions).

### Empirical design: matching, DID, and pairwise analysis
- Reform set notation and ordering:
  - Reforms = {TR, CPR, ELR, IR, CDR, SWR} (equation (3)).
  - Reforms_k = {x | x∈Reforms that are carried out in state k} (equation (4)).
  - Partial ordering: Reforms_k ≻ Reforms_l indicates state k strictly more reforms than l (equation (5)).
- Sample construction and matching:
  - Repeated cross-section covering 2010-11 and 2015-16; firms doubly matched across states and years using propensity scores based on Size (total fixed assets), Age (adjusted by subtracting 5 years from 2015-16 ages), sector (rural/urban), and three-digit industry classification.
  - Four-step matching procedure and psmatch2 in STATA 14 with caliper 0.1 used; covariates confirmed well-balanced in matched samples.
- DID specification (equation (6)):
  - lnTFPR_ijkt = μ_j + γ D_k + δ_t + τ (D_k × t) + β1 Size_ijkt + β2 Age_ijkt + β3 Ownership_ijkt + β4 location_ijkt + ε_ijkt
  - D = 1 for the state with strictly greater number of reforms; year t=0 is 2010-11; t=1 is 2015-16.
  - Controls: Size = log total assets; Ownership coded (1 male, 2 female, 3 transgender, 4 partnership same household, 5 partnership different household, 7 trusts, 8 others); Location = rural/urban; Industry fixed-effect μ_j included.
  - Standard errors clustered at the state level; firms are not weighted.

### Benchmark and robustness results: TFPR gains and dispersion reductions
- Benchmark pairwise DID findings (summary):
  - On average, enterprises in states that carried out a strictly greater number of reforms do better in TFPR.
  - Table 5a and Table 5b (summary counts) indicate states with more reforms generally perform better on TFPR compared to most other states, with exceptions for comparisons involving states with 2 or 3 reforms.
  - Specific states with all six reforms and consistent TFPR outperformance: Gujarat, Madhya Pradesh, and Rajasthan (for overall firms and manufacturing firms).
- Border-district matching (robustness):
  - Firms in neighbouring border districts matched to strengthen parallel-trends plausibility; caveat that business migration could bias estimates downward.
  - Results broadly hold, especially for the manufacturing sector, though sample sizes are smaller.
- Reform differences and TFPR gains (pairwise regression, Table 7 selected coefficients):
  - Reforms Differences: 0.047* (t statistic 1.760) in [1]; 0.233** (t statistic 1.980) in [2]; 0.088* (t statistic 1.940) in [3]; 0.181 (t statistic 0.880) in [4].
  - Reforms Difference Square: -0.043* (t statistic -2.440) reported in [1]; other entries reported as -0.042 (t statistic -1.340) etc.
  - Constant: 0.664*** (t statistic 7.370) in [1]; variation across other specifications.
  - N = 97 for all specifications; Adjusted R-squared: 0.022, 0.07, 0.028, 0.036 respectively.
  - Interpretation: greater number of reforms associated with larger TFPR gains; marginal benefits diminish as reform differences increase (negative square term).
- Impact on TFPR dispersion (Table 8, all pairs; N = 101):
  - [1] All Firms: Reforms Differences: -0.017*  (-1.840); Constant: 0.055*  (1.680); Adjusted R-squared: 0.023
  - [2] All Firms (with quadratic): Reforms Differences: -0.018*  (-1.950); Reforms Difference Square: 0.003*  (2.350); Constant: 0.019  (0.540); Adjusted R-squared: 0.066
  - [3] Manufacturing Firms: Reforms Differences: -0.026**  (-3.030); Constant: -0.031  (-1.060); Adjusted R-squared: 0.076
  - [4] Manufacturing Firms (with quadratic): Reforms Differences: -0.026**  (-3.060); Reforms Difference Square: 0.001  (1.000); Constant: -0.046  (-1.380); Adjusted R-squared: 0.076
  - Interpretation: larger reform differences associated with larger reductions in TFPR dispersion (more negative Dispersion Gain).
- Neighbouring pairs sample (19 pairs):
  - No statistically significant relationship between reform difference and TFPR gains, likely due to much smaller sample.
- Robustness: alternative TFPR estimation and sample restriction (Section 7.3):
  - Alternative wage and capital computations (equations (9)–(13)) and restriction to enterprises with at least 10 workers reduce sample to 141,423 from 473,178.
  - Using restricted sample, enterprises in states with strictly greater number of reforms show more TFP gains (Tables 9a and 9b summarize counts).
  - However, many microenterprises (<10 workers) are dropped and may contain large effects.
- TFPR gain heterogeneity:
  - Median reform effect on TFP is non-negative for any difference in number of reforms except for 1 reform difference (Figure 4, source).
  - Neighboring border-district pairs regressions (Table 10, N = 42) do not show statistically significant relationships; estimates reported as:
    - [1] All Firms: Reforms Differences: -0.006  (-0.080); Reforms Difference Square: -0.031  (-0.540); Constant: 0.298  (1.060); Adjusted R-squared: 0.025
    - Other specifications similarly not significant.
- Dispersion channels (Tables 11a and 11b):
  - Difference-in-difference regressions on dispersion (Theil’s index) of TFP, MPK, and MPL show:
    - All Matched Firms (Table 11a; N = 42): evidence that TFP and MPK dispersions declined more in states with more reforms; some specifications significant at approximately 10 percent.
    - All Matched Manufacturing Firms (Table 11b; N = 42): effects weaker; mixed signs and significance.
  - Example selected entries (Table 11a):
    - [1] TFP: Reforms Differences: -0.008  (-1.450); Reforms Difference Square: 0.005  (1.390); Constant: -0.033*  (-1.710); Adjusted R-squared: 0.012
    - [4] MPK (noted): Reforms Differences: -1.563*  (-2.010); Constant: 2.369*  (2.350); Adjusted R-squared: 0.018
  - Interpretation: primary channel for reduction in TFPR dispersion appears to be reductions in MPK dispersion, implying better capital allocation across firms and suggesting improvements in the functioning of the financial system; evidence for MPL dispersion effects is weaker.

### Key substantive findings and interpretations
- States implementing more public administration digitalisation reforms show:
  - Higher mean TFPR gains for microenterprises between 2010-11 and 2015-16.
  - Lower TFPR dispersion across firms, consistent with reduced misallocation.
  - Reduction in dispersion appears driven mainly by reductions in MPK dispersion (improved capital allocation), especially in all-firm samples; effects weaker in manufacturing-only samples and in small border-district samples.
- Marginal returns to additional reforms diminish as reform differences increase (negative quadratic term in TFPR gain regressions).
- Robustness checks (border matching, alternative TFPR construction, and sample restrictions) largely support primary results, with caveats about sample size and exclusion of many microenterprises in restricted samples.

### Policy-relevant implications (as supported by the analysis)
- Scaling public administration digitalisation as part of broader ease-of-doing-business reforms is associated with measurable firm-level productivity improvements and reduced productivity dispersion among microenterprises.
- Reforms that facilitate tax filing and payments, Single Window Systems, and other digitalised administrative procedures are associated with larger mean TFPR gains and dispersion reductions in matched comparisons.
- Improvements in capital allocation (reduced MPK dispersion) suggest complementary benefits through better-functioning financial intermediation and access to finance for firms following reforms.

*Source: Public Administration Digitalisation and Microenterprise Productivity in India, Working Paper No. WP/2026/086*

### 1. Introduction ........................................................................................................

### 1. Introduction

### Research question and approach
- Investigates whether public administration digitalisation, which eases regulatory and policy burdens on business, leads to higher productivity growth and lower productivity dispersion among firms in India.
- Focus: micro-enterprises covered in the Unincorporated Enterprise Surveys of India (two rounds: 2010-11 and 2015-16).
- Reform identification: Assessment of State Implementation of Business Reforms Report of 2015; assessment of public administration digitalisation undertaken in each state.
- Empirical strategy (illustrative): For a firm in a more-digitalised state, find a firm with similar characteristics in a less-reformed state and compare total factor productivities (TFPs); compare TFP dispersion across pairwise states; robustness check using firms located in bordering districts of neighbouring state pairs.

### Main findings (as stated)
- States that undertake more public administration digitalisation experience higher productivity growth and lower productivity dispersion among firms.
- Digitalisation of public administration scaled up in the mid-2010s, building on institutional initiatives and technological infrastructure developed earlier.
- The 98-point Action Plan on Ease of Doing Business was finalized in 2014; an evaluation of implementation was done in June 2015.

### Rationale and mechanisms
- Digitalisation reduces compliance costs (permits, filing taxes, accessing government schemes), disproportionately benefiting microenterprises due to their higher relative administrative cost burden.
- Digitalisation reduces corruption and informal cost variations by automating and increasing transparency, levelling the playing field for microenterprises.
- At the macro level, digitalisation can enhance aggregate productivity by better allocating resources and supporting higher-productivity firms.

### Context and related evidence
- India’s Doing Business ranking improved from 142 (Doing Business 2014) to 130 (Doing Business 2016) during the described reform momentum.
- MSME importance in India:
  - Contribute around 35 percent of manufacturing outputs.
  - Employ around 110 million workers (around 25 percent of the total workforce).
  - Contribute about 45 percent of the overall exports from India.
- Most MSMEs are not formally registered as companies under India’s Companies Act of 1956.

### Literature and theoretical grounding
- Cited findings from other studies:
  - Banerjee et al. (2023): Indonesia — transition to electronic vouchers reduced poverty by increasing program adherence.
  - Lewis-Faupel et al. (2016): e-procurement in India and Indonesia led to quality improvements (road quality in India; reduced delays in Indonesia).
- Cross-country literature: better business regulations associated with faster growth (Djankov et al. 2006); business regulatory reforms associated with GDP growth increases (Haidar 2012); governance problems reduce firm performance (Shleifer and Vishny 1997).
- Informality literature: heavier regulation, labour-market restrictions, and tax burden increase informal sector share (Djankov and Freund (2002), Antunes and Cavalcanti (2007), Perry et al. (2007), Ordonez (2014)); informal sector size decreases with improved credit availability (Straub, 2005; Antunes and Cavalcanti, 2007); obstacles to small firm growth include inadequate enforcement of property rights, inefficient regulation and taxation, political instability, insufficient infrastructure, underdeveloped financial system (Beck and Demirguc-Kunt 2006).

---

### 2. Business Environment Reforms 2014-15

### Reform definition and classification
- On 29th December 2014, Chief Secretaries finalized a 98-point Action Plan on Ease of Doing Business; digitalisation of public administration was a key part.
- Implementation assessment completed in June 2015.
- The 98 points are classified into six reform categories:
  - Tax filing and payments
  - Construction permits
  - Environment and labour standards
  - Inspections
  - Commercial dispute
  - Single Window Systems

### Description of reform categories and states implementing them (until 2015)
- Tax filing and payments:
  - Features: e-registration mandatory for VAT, CST, Professional Tax, Entry Tax; online filing and payments; service centre support and helpline; risk-based inspections.
  - States: Bihar, Karnataka, Madhya Pradesh, Chhattisgarh, Gujarat, Maharashtra, Odisha, Uttar Pradesh, Punjab, Rajasthan, Uttarakhand, West Bengal.
- Construction permits:
  - Features: online applications, upload building plans, software-based automatic plan scans, monitoring compliance with bylaws and codes.
  - States: Andhra Pradesh, Chhattisgarh, Gujarat, Haryana, Madhya Pradesh, Maharashtra, Rajasthan, Telangana, Uttarakhand.
- Environment and labour standards:
  - Features: online application systems; 24/7 access to apply, track applications, file returns, obtain permissions.
  - States: Andhra Pradesh, Bihar, Chhattisgarh, Gujarat, Haryana, Karnataka, Madhya Pradesh, Maharashtra, Jharkhand, Odisha, Punjab, Rajasthan, Tamil Nadu, Telangana, Uttar Pradesh, Uttarakhand, West Bengal.
- Inspections:
  - Features: user-friendly reforms (online self-certification), centralised inspection systems, published procedures and checklists, online inspector allocation.
  - States: Andhra Pradesh, Chhattisgarh, Gujarat, Haryana, Karnataka, Madhya Pradesh, Odisha, Jharkhand, Punjab, Rajasthan, Punjab, Telangana, Uttar Pradesh, Uttarakhand.
- Commercial disputes:
  - Features: passage of the Commercial Courts, Commercial Divisions, and Commercial Appellate Divisions Act (2015); state-level implementation timing varies; e-filing and e-summons introduced in some district courts.
  - States: Andhra Pradesh, Chhattisgarh, Gujarat, Haryana, Madhya Pradesh, Maharashtra, Jharkhand, Rajasthan, Telangana.
- Single Window System:
  - Features: one-stop online/physical window for filing, tracking, approvals, payments, inquiries, verifications.
  - States: Andhra Pradesh, Chhattisgarh, Jharkhand, Gujarat, Haryana, Madhya Pradesh, Maharashtra, Rajasthan, Telangana, Punjab, Uttarakhand.

### Exogeneity of reform implementation
- Reform implementations were not systematically associated with baseline state development characteristics.
- Using average SGDP growth, average per-capita SGDP growth, and industry share in state GVA as proxies, no systematic relationship was found between these indicators and number of reforms.
- OLS regressions of number of reforms on these development proxies yield statistically insignificant coefficients (Table 1 summary):
  - Average SGDP Growth coefficient: 0.546 (standard error (1.620))
  - Average Per-Capita SGDP Growth coefficient: 0.234 (standard error (0.700))
  - Share of Industry in State GVA coefficient: 0.054 (standard error (1.070))
  - Constants reported: -0.453 (-0.190), 2.090 (1.060), 1.410 (0.730) across columns
  - N = 19 in each regression
  - adj. R-sq values: 0.083, 0.029, 0.008

---

### 3. Data (survey and sample)
- Data source: National Sample Surveys (NSS) on unincorporated microenterprises; restricted to non-agricultural enterprises and excluding construction; further restricted to non-financial enterprises.
- NSS rounds used:
  - 67th round: July 2010–June 2011
  - 73rd round: July 2015–June 2016
- NSS 67th round: first round covering the entire unincorporated non-agricultural sector (excluding construction) as follow-up to Economic Census; 73rd round is a repeat.
- Table 2 (Distribution of Active MSMEs by Type, in 1000):
  - Proprietary: Registered 1409; Unregistered 32745; Total 34154; % of Unregistered Enterprises 95.87
  - Partnership: Registered 63; Unregistered 365; Total 428; % of Unregistered Enterprises 85.28
  - Private Company: Registered 43; Unregistered 6; Total 49; % of Unregistered Enterprises 12.24
  - Co-operative: Registered 5; Unregistered 116; Total 121; % of Unregistered Enterprises 95.87
  - Others: Registered 44; Unregistered 765; Total 809; % of Unregistered Enterprises 94.56
  - Not recorded: Registered 0; Unregistered 615; Total 615; % of Unregistered Enterprises 100
  - Total: Registered 1564; Unregistered 34612; Total 36176; % of Unregistered Enterprises 95.68
- Note: According to All India Census of MSME, Fourth Round, 2006-2007, unregistered enterprises constitute around 96 percent of total MSMEs.
- Definitions:
  - Unincorporated enterprises include proprietaries, partnerships (excluding limited liability partnerships), Self Help Groups (SHGs), and Non-Profit Institutions (NPIs).
  - Data used for regressions excludes NPIs and non-financial SHGs (non-financial SHGs are treated as partnerships when engaged in non-financial activities).
- Table 3 (NSS data on unincorporated enterprises):
  - 67th Round (July 2010–June 2011): Total Enterprises 334,475; Non-financial Enterprises 329,280; Manufacturing Enterprises 96,414
  - 73rd Round (July 2015–June 2016): Total Enterprises 288,676; Non-financial Enterprises 285,389; Manufacturing Enterprises 79,966

---

*Source: wpiea2026086-source-pdf - 1. Introduction ........................................................................................................*

### 4.1 Estimation of TFPR

### 4.1 Estimation of TFPR

### Estimation method
- Revenue-based total factor productivity (TFPR) at the firm level is computed as the product of price p_ijt times physical productivity A_ijt, assuming a Cobb-Douglas production function:
  - 푇퐹푃푅_푖푗푡 = 푝_푖푗푡 퐴_푖푗푡 = 푝_푖푗푡 푦_푖푗푡 / (푘_푖푗푡^훼_푗푡 푙_푖푗푡^(1−훼_푗푡))  (equation (1) in source)
- 푘_푖푗푡 is firm capital, 푙_푖푗푡 is labour, 퐴_푖푗푡 is physical total productivity, and 훼_푗푡 is capital share (one minus the labour share).
- Labour share (1−훼_푗푡) is estimated by industry-year averages:
  - 1−훼_푗푡 = mean(labour cost_푗푡 / net value added_푗푡)  (equation (2) in source)
- Wage payments are used as labour costs. Nominal net value added is measured as total output minus value of total inputs, net of total fixed asset depreciation during the year.
- Inflation is not adjusted in TFPR values, but is accounted for by year-fixed effects in regressions.

### Labour cost imputation and sample treatment
- State-wise minimum daily wage rates for unskilled labour in 2010 and 2015 from Indiastat are used, converted to monthly by assuming 20 workdays per month.
- Robustness check: actual wage payments are used for firms employing 10 or more workers.
- For micro and small enterprises with no wage data (self-employed units), the state government's minimum wage is used as imputed wage.
- For firms where owners are not counted as workers, 1 is added to the number of workers.
- Appendix (source) contains more details on TFPR estimation.

### State-wise average log TFPR: 2010-11 and 2015-16 (key statistics)
- Table 4 reports state-wise firm-level average TFPR (in logarithms) for 2010-11 and 2015-16. Selected rows (preserving numbers exactly as in source):
  - Andhra Pradesh: Mean (2010-11) -0.8472.617  Min. -11.891 Max. 18.975 ; Mean (2015-16) 0.32  Std. Dev 1.791  Min. -7.703 Max. 10.101
  - Bihar: Mean (2010-11) 0.4252.369  Min. -11.482 Max. 28.344 ; Mean (2015-16) 1.047  Std. Dev 1.604  Min. -7.611 Max. 12.861
  - Chhattisgarh: Mean (2010-11) -0.2372.219  Min. -10.634 Max. 10.168 ; Mean (2015-16) 0.288  Std. Dev 1.438  Min. -5.788 Max. 5.032
  - Gujarat: Mean (2010-11) -0.6852.443  Min. -15.965 Max. 16.406 ; Mean (2015-16) -0.219  Std. Dev 1.583  Min. -6.726 Max. .595
  - Haryana: Mean (2010-11) -1.3122.573  Min. -12.933 Max. 15.321 ; Mean (2015-16) 0.707  Std. Dev 1.306  Min. -4.883 Max. 4.689
  - Himachal Pradesh: Mean (2010-11) -1.42.672  Min. -13.022 Max. 15.769 ; Mean (2015-16) 0.074  Std. Dev 1.929  Min. -15.345 Max. 6.051
  - Jharkhand: Mean (2010-11) 0.802  Std. Dev 2.473  Min. -7.528 Max. 24.875 ; Mean (2015-16) 0.676  Std. Dev 1.376  Min. -6.437 Max. 6.397
  - Karnataka: Mean (2010-11) -0.262.581  Min. -13.273 Max. 14.146 ; Mean (2015-16) 0.317  Std. Dev 1.525  Min. -6.146 Max. 7.276
  - Kerala: Mean (2010-11) 0.167  Std. Dev 2.865  Min. -11.382 Max. 5.826 ; Mean (2015-16) 0.549  Std. Dev 1.667  Min. -5.868 Max. 8.8
  - Madhya Pradesh: Mean (2010-11) 0.073  Std. Dev 2.681  Min. -10.776 Max. 26.998 ; Mean (2015-16) 0.318  Std. Dev 1.833  Min. -7.412 Max. 12.872
  - Maharashtra: Mean (2010-11) -0.6762.916  Min. -17.654 Max. 29.009 ; Mean (2015-16) 0.041  Std. Dev 1.77  Min. -8.664 Max. 9.336
  - Odisha: Mean (2010-11) 1.522.794  Min. -7.888 Max. 27.815 ; Mean (2015-16) 1.083  Std. Dev 1.43  Min. -4.724 Max. 9.605
  - Punjab: Mean (2010-11) -1.2932.652  Min. -13.933 Max. 24.675 ; Mean (2015-16) 0.43  Std. Dev 1.368  Min. -6.912 Max. 5.298
  - Rajasthan: Mean (2010-11) -1.1582.301  Min. -12.544 Max. 9.632 ; Mean (2015-16) 0.363  Std. Dev 1.472  Min. -6.639 Max. 4.453
  - Tamil Nadu: Mean (2010-11) -0.9162.528  Min. -14.634 Max. 14.3 ; Mean (2015-16) 0.326  Std. Dev 1.652  Min. -7.351 Max. 10.215
  - Uttar Pradesh: Mean (2010-11) -0.4572.717  Min. -12.471 Max. 29.203 ; Mean (2015-16) 0.309  Std. Dev 1.508  Min. -6.448 Max. 12.27
  - Uttarakhand: Mean (2010-11) -1.182.317  Min. -10.027 Max. 8.32 ; Mean (2015-16) 0.369  Std. Dev 1.285  Min. -5.457 Max. 4.26
  - West Bengal: Mean (2010-11) 0.762  Std. Dev 2.944  Min. -13.232 Max. 28.592 ; Mean (2015-16) 1.276  Std. Dev 1.835  Min. -6.233 Max. 12.773
- Finding: between 2010-11 and 2015-16, there is an improvement in overall TFPR across the states (inflation not adjusted in reported TFPR but handled in regressions).

### Illustrative example: Single Window Systems and tax-related reforms (kernel-density observations)
- Hsieh and Klenow (2009) motivate comparing TFPR dispersion; greater dispersion implies greater misallocation.
- Kernel density comparisons for matched firms (propensity score matching) show:
  - Gujarat (all six reforms) vs Chhattisgarh (five reforms; no Single Window System): both saw reduced dispersion from 2010-11 to 2015-16; mean TFPR shift is larger in Gujarat.
  - Madhya Pradesh (all six reforms) vs Chhattisgarh (five reforms): both improved; mean TFPR improvement larger in Madhya Pradesh; dispersion in Madhya Pradesh decreased more.
  - Gujarat/Madhya Pradesh vs Andhra Pradesh/Haryana (did not implement tax-related reforms): states implementing tax-related reforms show increases in mean TFPR and reductions in dispersion.
- Interpretation: public administration and tax-related reforms are associated with both higher mean TFPR and reduced dispersion (less misallocation) in states that implemented reforms.

### Reform ordering and categorization
- Reform set defined as:
  - 푅푒푓표푟푚푠 = {푇푅, 퐶푃푅, 퐸퐿푅, 퐼푅, 퐶퐷푅, 푆푊푅}  (equation (3))
  - TR = Tax Related Reforms, CPR = Construction Permit Reforms, ELR = Environment and Labour Standard Reforms, IR = Inspection Reforms, CDR = Commercial Dispute Reforms, SWR = Single Window System Reform.
- For state k, carried-out reforms expressed as:
  - 푅푒푓표푟푚푠_푘 = {x | x∈푅푒푓표푟푚푠 that are carried out in state k}  (equation (4))
- Partial ordering: state k strictly more reforms than l written as 푅푒푓표푟푚푠_푘 ≻ 푅푒푓표푟푚푠_푙 (equation (5)).

### Matching firms across states and years (sample construction)
- Repeated cross-section covering 2010-11 and 2015-16; firms doubly matched across states and years using propensity scores based on:
  - Size (total fixed assets), age (adjusted by subtracting 5 years from 2015-16 ages), sector (rural/urban), and three-digit industry classification.
- Four-step matching procedure to obtain doubly matched firms (steps 1–4 listed in source).
- Confirmed covariates are well-balanced in matched samples.
- STATA 14 command psmatch2 with caliper 0.1 used.

### Difference-in-difference (DID) estimation
- Dependent variable: lnTFPR_ijkt (natural log of TFPR).
- Treatment: additional reforms for state A paired with state B when 푅푒푓표푟푚푠_𝐴 ≻ 푅푒푓표푟푚푠_𝐵.
- Regression specification (equation (6) in source):
  - lnTFPR_ijkt = μ_j + γ D_k + δ_t + τ (D_k × t) + β1 Size_ijkt + β2 Age_ijkt + β3 Ownership_ijkt + β4 location_ijkt + ε_ijkt
  - i, j, k, t refer to firm, industry, state, year. Year t=0 is 2010-11; t=1 is 2015-16.
  - D = 1 for the state with strictly greater number of reforms.
  - Controls: Size = log of total assets; Ownership coded (1 male, 2 female, 3 transgender, 4 partnership same household, 5 partnership different household, 7 trusts, 8 others). Location = rural/urban. Industry fixed-effect μ_j included.
  - Standard errors clustered at the state level. Firms are not weighted.

### Benchmark results (pairwise DID findings)
- Overall result: on average, enterprises in states that carried out a strictly greater number of reforms do better in TFPR.
- Table 5a (non-financial enterprises) and Table 5b (manufacturing enterprises) summarize counts of pairwise estimations where DID coefficients (τ) are positive and statistically significant at 5 percent or 10 percent.
  - Example interpretation from source: row 6 and column 5 in Table 5a shows 10/15, meaning 15 pairs of 6-reform states vs 5-reform states and 10 such pairs show significant TFPR gains by states with more reforms.
- Key finding: states with more reforms generally perform better on TFPR compared to most other states, except possibly for comparisons involving states with 2 or 3 reforms.
- Specific states with all six reforms and consistent TFPR outperformance: Gujarat, Madhya Pradesh, and Rajasthan (for overall firms and manufacturing firms).

### Border-district matching: robustness
- To strengthen parallel-trends plausibility, firms in neighbouring border districts of paired states (one with reforms, one without) are matched using the same propensity-score procedure.
- Caveat: business migration could bias border-district estimates downward (underestimate reform impacts).
- Results (Tables 6a and 6b): sample size smaller; results broadly hold, especially for the manufacturing sector—states with strictly greater number of reforms do better on TFPR even in border-district samples.

### Reform differences and TFPR gains: cross-pair analysis
- From benchmark regression results, pairwise TFPR effect coefficients (τ̂_kl) are regressed on pairwise reform differences using:
  - τ̂_kl = α1 + α2 ∆_kl ∆_t Reform_P + α3 (∆_kl ∆_t Reform_P)^2 + ε_kl  (equation (7))
  - ∆_kl denotes state k minus state l; ∆_t denotes difference after minus before.
- Table 7 (selected coefficients exactly as reported):
  - Reforms Differences: [1] 0.047* (t statistic 1.760), [2] 0.233** (t statistic 1.980), [3] 0.088* (t statistic 1.940), [4] 0.181 (t statistic 0.880)
  - Reforms Difference Square: -0.043* (t statistic -2.440) in [1]; -0.042 (t statistic -1.340) in [2]/[3]/[4] layout per source table formatting.
  - Constant: 0.664*** (t statistic 7.370) in [1]; 0.319* (t statistic 1.920) in [2]; 1.025*** (t statistic 6.610) in [3]; 0.693** (t statistic 2.370) in [4].
  - N = 97 for all specifications; Adjusted R-squared: 0.022, 0.07, 0.028, 0.036 respectively.
  - Note: Dependent variables are the pairwise differences of log TFPR estimated using DID. Regressors are Reform Difference of pair of states. t statistics in parentheses.
- Interpretation (source): greater number of reforms is associated with larger TFPR gains; however, marginal benefits diminish as reform differences increase (evidence from negative square term).

*Source: Authors’ estimations based on survey data.*

### 7.2 Impact of Reform Difference on TFPR Dispersion Reduction

### 7.2 Impact of Reform Difference on TFPR Dispersion Reduction

### Framework and measurement
- Theory: Following Hsieh and Klenow (2009), dispersion in TFPR reflects distortions that misallocate factor inputs; business environment reforms are expected to reduce these distortions and lower TFPR dispersion.
- Dispersion measure: Theil’s index computed for matched firms (same procedure as in Section 5). The dependent variable in regressions is the log(TFPR) Dispersion Gain, measured as the difference of the change in the Theil’s Index in the pair of states pre-and post-reforms.
- Empirical strategy: Difference-in-difference regressions across pairs of states (k,l) with the Reform Difference of the pair as the main regressor. Also include a quadratic term (Reforms Difference Square) in some specifications.

### Key regression results — All pairs of states (Table 8)
- Dependent variable: log(TFPR) Dispersion Gain; sample: All pairs of states; N = 101.
- Columns and estimates:
  - [1] All Firms
    - Reforms Differences: -0.017*  (-1.840)
    - Constant: 0.055*  (1.680)
    - Adjusted R-squared: 0.023
  - [2] All Firms (with quadratic)
    - Reforms Differences: -0.018*  (-1.950)
    - Reforms Difference Square: 0.003*  (2.350)
    - Constant: 0.019  (0.540)
    - Adjusted R-squared: 0.066
  - [3] Manufacturing Firms
    - Reforms Differences: -0.026**  (-3.030)
    - Constant: -0.031  (-1.060)
    - Adjusted R-squared: 0.076
  - [4] Manufacturing Firms (with quadratic)
    - Reforms Differences: -0.026**  (-3.060)
    - Reforms Difference Square: 0.001  (1.000)
    - Constant: -0.046  (-1.380)
    - Adjusted R-squared: 0.076
- Interpretation: Table 8 shows results consistent with the prediction that reforms lower productivity dispersion across firms (more negative Dispersion Gain when reform differences are larger).

### Neighbouring pairs sample
- For neighbouring pairs of states (19 pairs), no statistically significant relationship between reform difference and TFPR gains was found (report omitted), likely due to much smaller sample numbers.

### Robustness: alternative TFPR estimation and sample restrictions (Section 7.3)
- Alternative firm-level wage: estimate w_ijt as MPL_ijt = Compensation to employees_ijt / Total number employees_ijt (equation (9)).
- Capital cost: Capital Cost_ijt = GVA_ijt − compensation to employees_ijt (equation (10)).
- Total capital constructed as Owned Asset_ijt + Leased Asset_ijt + Added Asset_ijt / 2 (equation (11)); Owned asset is beginning-of-period stock value, Leased and Added are flows, half counted as average total assets.
- Marginal product of capital: r_ijt = MPK_ijt = capital cost_ijt / Total capital_ijt (equation (12)).
- TFPR from MPK and MPL following Hsieh and Klenow (2009) and using industry average capital share α_jt (equation (13)).
- Estimation sample restriction: Compensation to Employees data missing for self-employed and misreported for small family-run businesses → restrict to enterprises that employ at least 10 workers. Sample reduces to 141,423 from 473,178.
- Using the restricted sample, regressions like equation (6) show enterprises in states with a strictly greater number of reforms show more TFP gains (Tables 9a and 9b summarize counts of positive and statistically significant DID coefficients across state-pairs).

### TFPR gain heterogeneity and DID coefficients
- Regression of logTFP gain on Reform differences (τ̂_kl) finds that the median reform effect on TFP is non-negative for any difference in the number of reforms except for 1 reform difference (Figure 4).
- However, a regression of number of reforms on TFPR gains (Table 10) for Pairs of Neighboring States along Bordering Districts does not find a statistically significant relationship between the number of reforms and TFP gains; small-sample issues and dropped microenterprises (<10 workers) likely cause lack of significance.

Table 10 (Pairs of Neighboring States; N = 42)
- [1] All Firms
  - Reforms Differences: -0.006  (-0.080)
  - Reforms Difference Square: -0.031  (-0.540)
  - Constant: 0.298  (1.060)
  - Adjusted R-squared: 0.025
- [2] All Firms (other specification)
  - Reforms Differences: 0.187  (0.510)
  - Reforms Difference Square: -0.027  (-0.440)
  - Constant: 0.055  (0.100)
  - Adjusted R-squared: 0.043
- [3] Manufacturing Firms
  - Reforms Differences: 0.035  (0.350)
  - Constant: -0.049  (-0.150)
  - Adjusted R-squared: 0.022
- [4] Manufacturing Firms (other specification)
  - Reforms Differences: 0.21  (0.510)
  - Constant: -0.276  (-0.450)
  - Adjusted R-squared: 0.043

### Dispersion in MPK and MPL and channels of TFPR dispersion reduction (Tables 11a and 11b)
- Method: Difference-in-difference regression like equation (7) applied to dispersion (Theil’s index) of TFP, MPK, and MPL.
- Main finding: TFP and MPK dispersions declined more in states with more reforms (Table 11a) at approximately 10 percent significance in some specifications; effects weaker for manufacturing-only sample (Table 11b). The reduction in TFPR dispersion appears mainly via reductions in MPK dispersion, suggesting improved allocation of capital across firms and a better-functioning financial system.

Table 11a (All Matched Firms; N = 42)
- Columns: [1] TFP, [2] TFP, [3] MPK, [4] MPK, [5] MPL, [6] MPL
- [1]
  - Reforms Differences: -0.008  (-1.450)
  - Reforms Difference Square: 0.005  (1.390)
  - Constant: -0.033*  (-1.710)
  - Adjusted R-squared: 0.012
- [2]
  - Reforms Differences: -0.038*  (-1.690)
  - Reforms Difference Square: 0.243*  (2.000)
  - Constant: 0.004  (0.140)
  - Adjusted R-squared: 0.02
- [3]
  - Reforms Differences: -0.053  (-0.280)
  - Reforms Difference Square: -0.063  (-0.290)
  - Constant: 0.506  (0.810)
  - Adjusted R-squared: 0.024
- [4]
  - Reforms Differences: -1.563*  (-2.010)
  - Reforms Difference Square: 0.243*  (2.000) [reported in column 2 but relevant quadratic noted]
  - Constant: 2.369*  (2.350)
  - Adjusted R-squared: 0.018
- [5]
  - Reforms Differences: -0.022  (-0.060)
  - Reforms Difference Square: -0.063  (-0.290)
  - Constant: -0.793  (-0.620)
  - Adjusted R-squared: 0.025
- [6]
  - Reforms Differences: 0.369  (0.230)
  - Reforms Difference Square: -0.063  (-0.290)
  - Constant: -1.275  (-0.520)
  - Adjusted R-squared: 0.05

Table 11b (All Matched Manufacturing Firms; N = 42)
- Columns: [1] TFP, [2] TFP, [3] MPK, [4] MPK, [5] MPL, [6] MPL
- [1]
  - Reforms Differences: 0.021  (-1.350)
  - Reforms Difference Square: 0.005  (0.450)
  - Constant: -0.012  (-0.260)
  - Adjusted R-squared: 0.029
- [2]
  - Reforms Differences: -0.007  (-0.120)
  - Reforms Difference Square: -0.008  (-1.510)
  - Constant: 0.022  (0.280)
  - Adjusted R-squared: 0.01
- [3]
  - Reforms Differences: 0.003  (0.370)
  - Reforms Difference Square: -0.016  (-1.020)
  - Constant: -0.033  (-1.140)
  - Adjusted R-squared: 0.023
- [4]
  - Reforms Differences: 0.056  (1.560)
  - Reforms Difference Square: -0.008  (-1.510)
  - Constant: -0.098*  (-2.200)
  - Adjusted R-squared: 0.014
- [5]
  - Reforms Differences: -0.016  (-0.870)
  - Reforms Difference Square: -0.016  (-1.020)
  - Constant: 0.009  (0.120)
  - Adjusted R-squared: 0.018
- [6]
  - Reforms Differences: 0.082  (0.860)
  - Reforms Difference Square: -0.016  (-1.020)
  - Constant: -0.111  (-0.990)
  - Adjusted R-squared: 0.027

### Interpretation of channels
- The primary channel for reduction in TFPR dispersion appears to be reductions in MPK dispersion, implying better capital allocation across firms after reforms and suggesting improvements in the functioning of the financial system.
- Evidence for MPL dispersion effects is weaker.

### Summary statements from the section
- Table 8 shows results consistent with the prediction that reforms lower productivity dispersion across firms.
- For neighbouring pairs, power is lost (small samples): "Although the power is lost for border districts samples, perhaps due to small samples (N = 22), for all pairs of states, the results are as predicted, that is, more reforms reduces TFP dispersion across firms (report omitted)."
- With the restricted sample (>=10 employees), enterprises in states with strictly greater number of reforms show more TFP gains; many microenterprises (<10 workers) are dropped and may contain large effects.

*Source: Public Administration Digitalisation and Microenterprise Productivity in India, Working Paper No. WP/2026/086*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026086-source-pdf.pdf_