## wpiea2023095-print-pdf

## Source details

**Canonical URL:** [wpiea2023095-print-pdf](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023095-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023095-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023095-print-pdf.pdf.json)

---

### Research question and scope
- Would digitalization at firm-level strengthen firms’ resilience to shocks? Could fiscal policy promote firm-level digitalization?
- Methodology: local projection method estimating the impact of aggregate uncertainty shocks on firms’ sales, profit margins, and employment.
- Microdata: Orbis data covering 1.8 million non-financial firms from 53 countries (main analysis after cleaning uses Orbis data that cover about 1.9 million listed and unlisted private non-financial corporations in 53 countries from 2000 to 2017).
  - Sample composition notes: 78 percent Advanced Economies (AE) and 22 percent Emerging Market and Developing Economies (EMDE).
  - Sample restrictions: retain corporations with at least five annual observations; winsorized at top and bottom 1 percentile.
- Aggregate uncertainty shock: World Uncertainty Index (WUI) from Ahir, Bloom, and Furceri (2022) (WUI series referenced from 1996; figure shown for 1990Q1 to 2022Q3).
- Cross-country policy analysis: cross-country panel data on fiscal institutions and digital policy; Executive Summary reports "covering 64 countries" for fiscal interventions; Digital Trade Restrictiveness (DTR) index provides scores for 62 countries.

### Core empirical findings
- Heterogeneous impact of uncertainty shocks by firm digitalization:
  - Digitalized firms weather shocks better, with smaller drops in sales and profits.
  - Less-digitalized firms suffer larger and longer-lasting scars, especially in the service sector.
- Distributional effects on financial performance:
  - Digitalization could create distinctive inequality across firms.
  - Among loss-making firms, lack of digitalization makes negative shocks larger and longer lasting.
  - Some digitalized firms can exhibit positive profits even during crises by taking advantage of market power, especially in EMDEs.
- Employment effects:
  - Short-term: digitalized firms could increase job offering.
  - Long-term: digitalized firms show a decline in employment, possibly reflecting labor substitution with automation.

### Key methodological and data details
- Outcome variables: firm sales, profit margins, employment (converted to real terms using each country’s GDP deflator).
- Industry composition highlights:
  - 15 percent manufacturing firms.
  - 26 percent wholesale and retail business.
- Data cleaning steps:
  - Drop observations with negative assets, tangible assets, negative employees, negative sales.
  - Drop observations missing all of total assets, sales, numbers of employees, total operating revenues.
  - Drop duplicated observations.
  - Sample period up to 2017.
  - Drop observations without industry classification or with negative cost of employees or operating revenue.
  - Drop observations with negative or zero total assets and employees.

### Policy-relevant findings on fiscal interventions
- Fiscal interventions could support promotion of firm-level digitalization.
- Specific fiscal institutional features associated with more effective promotion of firm-level digitalization:
  - Align taxation on digital transactions with general taxation principles (as set in the Ottawa Taxation Framework Conditions).
  - Ensure competitive procurement rules on digital products.
- Policy implications drawn:
  - Governments may need to expect long-lasting scars for non-digitalized firms and consider longer-duration support where appropriate.
  - Blanket supports for affected firms may be less effective when many loss-making, non-digitalized firms are unlikely to recover.
  - Governments should actively encourage firm-level digitalization by establishing fiscal institutions for the digital economy based on international taxation principles and open market competition.

### Cross-country digitalization speed (World Bank’s DAI) and firm-level digitalization index
- DAI (Business) components: simple average of percentage of businesses with websites, number of secure servers, speed of internet, and 3G coverage.
- Distributional findings:
  - Substantial digital gap between ADV, EM, and LICs.
  - Level of business digitalization tends to accelerate as per capita income rises.
  - Government digitalization may progress in LICs while private-sector gap remains deeper.
- Firm-level Digitalization Index construction:
  - Inputs: OECD harmonized Input-Output Tables (IOTs) across 45 sectors; Orbis firm-level data mapped from NACE to OECD IOT two-digit classification.
  - Digital (ICT)-related sectors: manufacturing of computer, electronic and optical equipment; telecommunication; IT and other information services.
  - Measure: sector-level share of intermediate inputs from digital industry × firm’s previous year total outputs, scaled by asset size; assumes same rate of digitalization within sector.

### Firm-level patterns and stylized facts
- Sectoral and size patterns:
  - Index higher in service sectors (ICT, e-commerce, software) and some manufacturing (IT equipment) relative to face-to-face businesses (restaurant and hotel).
  - Index positively correlated with number of employments even after adjusting for assets.
- Binned scatter: sample divided into 100 equally-sized bins; each dot is binned average for 2015.
- Interpretation: positive correlation with employment suggests scale economy—firms with larger labor inputs tend to have more digitalization.

### Econometric analysis: identification and baseline specification
- Baseline model:
  - Local projection method estimating impulse response functions for horizons h = 0, ...5 (h = 0 is shock year).
  - Outcomes y: real sales, operating profit margins (operating profits divided by revenues), and employment of firm i in industry j, country c at year t.
  - WUI: change in World Uncertainty Index.
  - Controls x: firm’s total asset (in log), current ratio, and lag of real GDP growth.
  - Standard errors clustered by country-industry-year.
  - Fixed effects: country, industry (four-digit NACE4), and year.
- Identification: exploits firm-industry-country-year variation to mitigate reverse causation concerns.

### Baseline results for a standardized 10 percent adverse WUI shock
- Description: standardized uncertainty shock of 10 percent, described as broadly equivalent to shocks experienced after major crises (e.g., 9/11 and the GFC) above the historical median level.
- Impulse response magnitudes after a 10 percent adverse WUI shock at A=0 (total sample, black line):
  - Sales: drop of 2 percent.
  - Profit margins: drop of 0.35 percentage points (pp).
  - Employment: drop of 1.6 percent on average.
- Dynamics:
  - Negative effects peak in year three followed by gradual recovery.
  - For small-sized enterprises (less than 20 employees), scarring effect is larger at the peak for profits and employment.
- Inference: long-lasting scarring effects after an uncertainty shock; robustness confirmed in Annex II.

### State-dependent estimation by firm digitalization (smooth transition)
- Approach: local projection with smooth transition function G(·) to define high (H) vs low (L) digitalization regimes using normalized firm-level digitalization index D*.
- Mitigation effect measured as β_H,h − β_L,h (difference in IR functions between regimes).
- Main results (mitigation effects):
  - High-digitalization firms show significantly better performance after shock in sales and profit margins.
  - Less-digitalized firms experience larger immediate drops and pursue cost cuts to restore profitability in medium term.
  - EMDEs: medium-term difference particularly large—digitalized group shows strong growth in sales, profit margins, and employment after the shock, while less-digitalized firms exhibit negative conditions (interpreted as skewed market structure and high market concentration in EMDEs).
  - SMEs: similar patterns observed.
  - Employment: short-run mitigation of disemployment by digitalization; long-run reversal with digitalization associated with decline in labor share (labor substitution with automation).

### Heterogeneous effects by sector
- Manufacturing vs. services:
  - Mitigation patterns present in both; more salient in the service sector.
- Primary sector: effect not precisely estimated due to limited sample.
- Within-industry distribution:
  - Digitalization creates winners and losers within industries; in services a handful of digitalized firms capture larger gains post-shock while less-digitalized firms face larger losses.

### Robustness checks
- Alternative specifications:
  - Control for country- and industry-specific time trends.
  - Replace country fixed effects with firm-level fixed effects.
- Results:
  - Alternative specifications yield similar or larger effect sizes; profits and employment for digitalized group remain positive; gaps persist between digitalized and less-digitalized firms.
- Additional detail: binned scatter uses 100 bins; each dot is binned average for 2015.

### Zombification, pre-shock profitability, and heterogeneous scarring
- Definition: “loss-makers” = firms consistently making losses for at least three consecutive years and older than ten years.
- Empirical finding:
  - Loss-making firms without digitalization show larger and longer negative scarring in sales and profits than other firms; employment heterogeneity not significant.
  - Non loss-making firms show smaller and shorter negative impacts.
- Policy implications:
  - From firm perspective: loss-making firms, especially “zombies,” urgently need government liquidity support to survive in crisis.
  - From fiscal-risk perspective: less-digitalized loss-making firms show limited recovery prospects; loans or guarantees could accumulate fiscal risks; grants may not improve prospects given large sales impacts.
  - Recommendation: include digitalization in targeting rules of emergency fiscal supports (besides income-based targeting).

### Post-COVID firm dynamics (application and literature synthesis)
- Caveat: COVID-19 introduced unique structural break; out-of-sample forecasting using pre-COVID data is challenging.
- Approach: combine pre-COVID firm-level estimates with size of COVID uncertainty shock (WUI).
- Key implications:
  - Expect larger and longer scars after the pandemic—particularly on less-digitalized firms in service sectors.
  - The size of the COVID uncertainty shock is about four times larger than the standardized 10 percent shock assumed in baseline analysis.
  - Contact-intensive sectors (transportation, accommodation, food services) hit hardest.
- Selected literature findings (examples):
  - Abidi, Herradi, Sakha (2022): Digitalized firms faced smaller sales drop by about 4 p.p than less-digitalized firms.
  - IMF (2022a), Copestake, Estefania-Flores and Furceri (2022): Firms in more digitalized industries recorded 3.4 percent higher sales revenues in a year; hiring rates higher in industries using more digital skills; past recessions benefit on sales 1.4 percent and largest after two years.
  - Cirera et al (2021): 44 percent of businesses increased digital adoption.
  - Kawaguchi, Kitao, Nose (2022): Firms initially adopting remote work faced smaller sales drop by 1.5 percent (1.9 percent for low-contact industries) to 10 percent mobility shock.
  - Bai et al (2021): Firms with high pre-COVID Work-from-Home index had significantly higher sales, net income, and stock returns during COVID-19.
  - Pierri and Timmer (2020): Local ICT adoption reduced increase in unemployment rate by 2 p.p.
  - Cong, Yang, Zhang (2021): Digitalized firms experienced smaller sales decline, sustained cash flow, and quick reopening.

### Fiscal DTR index: definition, empirical strategy, and cross-country findings
- DTR index (ECIPE):
  - Covers 64 countries; measures regulation on digital transactions (0 supportive to 1 restrictive).
- Fiscal DTR disaggregated sub-components:
  - (a) tax regime on digital goods and online services;
  - (b) tariffs on ICT trade (custom policy);
  - (c) procurement rules on digital products.
- Empirical model: two-limit Tobit with DTR quartile dummies (1st and 2nd quartiles as less restrictive groups), controls for log GDP per capita, tertiary education completion, broadband access, financial development; region and year fixed effects.
- Cross-country findings:
  - Stylized fact: negative correlation between a country’s DAI (business) and fiscal restriction on digital activities (controlling for income and region)—more digitalization associated with less fiscal restriction.
  - Tobit results (summary):
    - Top groups (fiscally least restrictive) in taxation and procurement are positive and statistically significant.
    - These top groups increase digital adoption by about 3 percentage points and 6 percentage points, respectively, compared with bottom half groups.
    - Custom regulations on ICT trade less significant.
  - Policy interpretation: aligning tax regimes with international tax standards and competitive bidding on digital goods procurement support firm digitalization.

### Caveats, limitations, and areas for further study
- DTR index limitations:
  - DTR updates lack historical records; measures current state only.
  - Sources based on public legal/policy documents; scope to expand covered fiscal policies.
  - ECIPE scoring involves subjectivity.
- Causality:
  - Findings cannot claim causal relationships between government fiscal intervention and firm-level digitalization; treat policy discussion as suggestive evidence.
- Areas for further study:
  - Use more granular fiscal policy data to strengthen causal evidence.
  - Explore post-COVID firm dynamics as more data become available.
  - A companion paper (Nose, Honda, et al, forthcoming) will further examine tax recovery.

### Quantitative robustness table excerpts (selected)
- Log sales (Years after shock 0 1 2 3 4):
  - Baseline: -0.005 -0.080*** -0.149*** -0.189*** -0.138***
  - w/ industry trend: -0.005 -0.080*** -0.149*** -0.190*** -0.140***
  - w/ country trend: -0.005 -0.090*** -0.189*** -0.239*** -0.172***
  - w/ both trends: -0.005 -0.090*** -0.188*** -0.238*** -0.171***
- Profit Margins (Years after shock 0 1 2 3 4):
  - Baseline: -0.012*** -0.025*** -0.032*** -0.032*** -0.009**
  - w/ industry trend: -0.012*** -0.025*** -0.032*** -0.032*** -0.009**
  - w/ country trend: -0.014*** -0.027*** -0.033*** -0.032*** -0.007
  - w/ both trends: -0.014*** -0.027*** -0.033*** -0.032*** -0.007
- Log employment (Years after shock 0 1 2 3 4):
  - Baseline: -0.047*** -0.064*** -0.106*** -0.143*** -0.110***
  - w/ industry trend: -0.046*** -0.064*** -0.105*** -0.142*** -0.110***
  - w/ country trend: -0.071*** -0.130*** -0.214*** -0.259*** -0.173***
  - w/ both trends: -0.071*** -0.129*** -0.214*** -0.259*** -0.173***
- Regression evidence: selected coefficients on fiscal interventions
  - Taxation:
    - Tax_q1: 0.030 *
    - Tax_q2: 0.004
  - Procurement:
    - Procurement_q1: 0.054 ***
    - Procurement_q2: 0.038 ***
  - Tariffs:
    - Tariff_q1: 0.027 (standard error 0.017)
    - Tariff_q2: 0.008 (standard error 0.014)
  - Controls:
    - Log GDP per capita (in PPP) (lagged): 0.133 ***
    - Broadband subscribers: 0.453 ***
    - GCI-Financial Market Development index: 0.020 *
  - Sample sizes (reported): N 124, 96, 123 (across columns)
  - Region & Year fixed effects: YES

### Sample composition (selected counts and shares)
- Total number of firms: 1,968,918 (100)
- By country (selected):
  - France: 448,408 22.77
  - Spain: 432,447 21.96
  - Bulgaria: 156,429 7.94
  - Italy: 143,109 7.27
  - Russia: 86,097 4.37
  - Czech Republic: 78,919 4.01
- By sector (selected):
  - Wholesale and retail trade: 523,158 26.57
  - Construction: 283,029 14.37
  - Other business sector services: 276,009 14.02
  - Accommodation and food services: 113,970 5.79
  - Real estate activities: 99,140 5.04
  - IT and other information services (ICT-related): 45,358 2.30

### Conclusions and policy takeaways
- Digitalization mitigates adverse impact of uncertainty shocks and creates substantial heterogeneity across firms; benefits are especially salient in services and in EMDEs for digitally advanced firms.
- Key empirical timing: after a "10 percent adverse shock ( broadly equivalent to the magnitude of the shock after major crises in recent history above historical median level)", sales, profit margins, and employment are expected to gradually start recovering "three years after the shock".
- Fiscal policy can facilitate firm-level digitalization through less restrictive tax regimes, tax incentives and subsidies, and open and competitive public procurement (e.g., e-procurement, less stringent tender criteria).
- Policy design should:
  - Anticipate long-lasting scarring and consider duration and targeting of support.
  - Incorporate firm-level digitalization status into targeting rules for emergency fiscal supports.
  - Address fiscal risks from lending/guarantee programs to less-digitalized loss-making firms with weak prospects.

*Italic: Source — wpiea2023095-print-pdf*

### Executive Summary ......................................................................................................

### Executive Summary

### Research question and scope
- Would digitalization at firm-level strengthen firms’ resilience to shocks? Could fiscal policy promote firm-level digitalization?
- Methodology: local projection method estimating the impact of aggregate uncertainty shocks on firms’ sales, profit margins, and employment.
- Microdata: Orbis data covering 1.8 million non-financial firms from 53 countries (main analysis after cleaning uses Orbis data that cover about 1.9 million listed and unlisted private non-financial corporations in 53 countries from 2000 to 2017).
  - Sample composition notes: 78 percent Advanced Economies (AE) and 22 percent Emerging Market and Developing Economies (EMDE).
  - Sample restrictions: retain corporations with at least five annual observations; winsorized at top and bottom 1 percentile.
- Aggregate uncertainty shock: World Uncertainty Index (WUI) from Ahir, Bloom, and Furceri (2022) (WUI series referenced from 1996; figure shown for 1990Q1 to 2022Q3).
- Cross-country policy analysis: cross-country panel data on fiscal institutions and digital policy. The Executive Summary reports "covering 64 countries" for fiscal interventions; the paper sets up a Digital Trade Restrictiveness (DTR) index that provides scores for 62 countries.

### Core empirical findings
- Heterogeneous impact of uncertainty shocks by firm digitalization:
  - Digitalized firms weather shocks better, with smaller drops in sales and profits.
  - Less-digitalized firms suffer larger and longer-lasting scars, especially in the service sector.
- Distributional effects on financial performance:
  - Digitalization could create distinctive inequality across firms.
  - Among loss-making firms, lack of digitalization makes negative shocks larger and longer lasting.
  - Some digitalized firms can exhibit positive profits even during crises by taking advantage of market power, especially in emerging markets and developing economies.
- Employment effects:
  - Mixed results: in the short-term, digitalized firms could increase job offering.
  - In the long-run, digitalized firms show a decline in employment, possibly reflecting labor substitution with automation.

### Policy-relevant findings on fiscal interventions
- Fiscal interventions could support promotion of firm-level digitalization.
- Specific fiscal institutional features associated with more effective promotion of firm-level digitalization:
  - Align taxation on digital transactions with general taxation principles (as set in the Ottawa Taxation Framework Conditions).
  - Ensure competitive procurement rules on digital products.
- Policy implications drawn:
  - Governments may need to expect long-lasting scars for non-digitalized firms and consider longer-duration support where appropriate.
  - Blanket supports for affected firms may be less effective when many loss-making, non-digitalized firms are unlikely to recover.
  - Governments should actively encourage firm-level digitalization by establishing fiscal institutions for the digital economy based on international taxation principles and open market competition.

### Contribution to literature
- Provides global evidence using detailed balance sheet and income statement data from Orbis that includes both listed and non-listed firms across AEs and EMDEs.
- Estimates heterogeneous effects of digitalization on resilience to aggregate uncertainty shocks, and quantifies gaps in firm financials associated with digital adoption.
- Examines the role of fiscal institutional quality (taxation, customs policy, procurement) in firm digital adoption using a DTR index.

*Source: wpiea2023095-print-pdf - Executive Summary*

### 9. We overview cross-country differences in digitalization speed using t he World Bank’s Digital

### 9. We overview cross-country differences in digitalization speed using the World Bank’s Digital Adoption Index (DAI)

### Cross-country DAI (Business) overview
- The DAI (Business) is the simple average of four normalized indicators: the percentage of businesses with websites, the number of secure servers, the speed of internet, and 3G (third-generation) coverage.
- Distributional findings:
  - Figure 2 (left) compares the distribution of the DAI by income groups, showing a substantial digital gap between advanced (ADV), emerging markets (EM), and low-income countries (LICs).
  - The level of digitalization in business tends to accelerate as a country’s per capita income rises (Figure 2, right).
  - While government digitalization progresses in LICs, the digital gap in the private sector is much deeper between advanced and developing economies.
- Motivation:
  - Given the deeper private-sector digital gap, the paper examines how fiscal policy could contribute to the promotion of economy-wide digitalization, distinct from government digitalization literature.

### Firm-level Digitalization Index: construction and data
- Data sources:
  - OECD’s harmonized Input-Output Tables (IOTs) with inter-sectoral flows across 45 sectors.
  - Orbis firm-level data mapped from NACE to OECD IOT two-digit classification.
- Digital (ICT)-related sector definition:
  - Comprises three sectors related to ICT use in production: manufacturing of computer, electronic and optical equipment; telecommunication; and IT and other information services.
- Index construction (conceptual description in source):
  - For each sector j, compute the share of intermediate inputs coming from the digital industry out of sector j’s total intermediate outputs (“digitalization” component).
  - Multiply sector-level digital intermediate input shares by each firm’s previous year’s total outputs to proxy the monetary value of firm-level digital intensity, scaled by asset size.
  - Firm-size adjusted measure of digitalization is defined using firm i’s digital intermediate inputs, sector and firm outputs, and firm total assets from Orbis (previous year).
- Mapping and assumptions:
  - Assumes the rate of digitalization is the same within sector, with diffusion rising for firms that make larger sales in the same market.
  - Orbis lacks consistent firm-level information on broadband access and manager characteristics across countries.

### Firm-level patterns and stylized facts (Figure 3 and notes)
- Sectoral and size patterns:
  - The firm-level digitalization index tends to be higher in service sectors (information and communication technology, e-commerce in retail & wholesale, software) and some manufacturing sectors (production of IT equipment) relative to face-to-face businesses (restaurant and hotel).
  - The index is positively correlated with number of employments even after adjusting for firms’ assets.
- Binned scatter plot specifics:
  - Each dot corresponds to the binned average of firm-level digitalization index in 2015.
  - The sample is divided into 100 equally-sized bins in the computation.
- Interpretation:
  - Positive correlation with employment suggests scale economy: more digitalization for firms operating with larger labor inputs.

### IV. Econometric analysis: identification and specification
- Objective:
  - Estimate impact of aggregate uncertainty shocks on firm-level sales, profit margin, and employment.
- Baseline model:
  - Local projection method estimating impulse response functions for horizons h = 0, ...5 (h = 0 is the year when the uncertainty shock takes place).
  - Outcome variables y: real sales, operating profit margins (operating profits divided by revenues), and employment of firm i in industry j, country c at year t.
  - WUI: change in the World Uncertainty Index (WUI).
  - Controls x: firm’s total asset (in log), current ratio (current assets divided by current liabilities), and a lag of real GDP growth.
  - Standard errors clustered by country-industry-year.
  - Country fixed effects λ_i, industry fixed effects φ_i (four-digit NACE4), and year fixed effects δ_t included.
- Identification strategy:
  - Exploits four-dimensional variation (firm-industry-country-year) to mitigate reverse causation concerns.
  - Argues plausibility that aggregate uncertainty affects firm-level performance rather than vice versa after controlling for fixed effects.

### Baseline results (Figure 4)
- Shock definition:
  - Standardized uncertainty shock of10 percent, described as broadly equivalent to shocks experienced after major crises (e.g., 9/11 and the GFC) above the historical median level.
- Impulse response magnitudes after a 10 percent adverse WUI shock at A=0 (total sample, black line):
  - Sales: drop of 2 percent.
  - Profit margins: drop of 0.35 percentage points (pp).
  - Employment: drop of 1.6 percent on average.
- Dynamics:
  - Negative effects peak in year three followed by gradual recovery.
  - For small-sized enterprises (less than 20 employees), the scarring effect is larger at the peak for profits and employment.
- Inference:
  - Results indicate long-lasting scarring effects after an uncertainty shock.
  - Robustness confirmed in Annex II (as stated in source).

### State-dependent local projection estimation: regime switching by firm digitalization
- Rationale:
  - Use local projection with a smooth transition (regime switching) function to avoid arbitrary binary split and to allow non-linearity by degree of firm digitalization.
- Specification:
  - Smooth transition function G(·) determines high (H) vs low (L) digitalization regimes based on firm-level digitalization index D* normalized against global averages for each year.
  - The mitigation effect is measured as β_H,h − β_L,h, the difference in IR functions between high- and low-digitalization regimes.
- Horizons and controls:
  - Same control variables and fixed effects as baseline; estimate difference in IR responses across regimes.

### Main results: mitigation effects of firm digitalization (Figure 5)
- Aggregate findings:
  - Firms with high digitalization show significantly better corporate performance after a shock in terms of sales and profit margins.
  - Less-digitalized firms experience significantly larger drops in sales and profits immediately after the shock and pursue cost cuts to restore profitability in the medium term.
- EMDEs (Panel B):
  - In the EMDE sample, the medium-term difference in corporate performance is particularly large.
  - Digitalized group shows strong growth in sales, profit margins, and employment after the shock, while less-digitalized firms exhibit negative financial and employment conditions.
  - Interpretation: skewed market structure and high market concentration in EMDEs may allow some digitalized firms with market power to capture gains post-shock.
- SMEs:
  - Similar patterns observed for SME sub-sample.
- Employment effects (Figure 5, right):
  - Short-run: digitalization may mitigate disemployment effects of the shock (digitalized firms offer more jobs; less-digitalized firms save labor costs).
  - Long-run: employment trend reverses—digitalization seems to promote decline in labor share (labor substitution with automation), consistent with cited literature.
- Policy implication:
  - Digitalization before a crisis could be an important insurance strategy to cope with shocks for firms.

### Heterogeneous effects by sector (Figure 6)
- Manufacturing vs. services:
  - Similar mitigation patterns in both sectors, though effect is more salient in the service sector.
- Primary sector:
  - Effect could not be precisely estimated due to limited sample size.
- Within-industry distribution:
  - Digitalization creates winners and losers even within industries.
  - In services, a handful of digitalized firms capture larger gains post-shock while less-digitalized firms face larger losses.
  - Larger digital gaps within services indicate ICT adoption is essential for firms to compete and perform after shocks.

### Robustness checks (specifications and findings)
- Endogeneity concerns addressed via two specification checks:
  - 1) Control for country- and industry-specific time trends simultaneously to reduce bias from selective digital adoption.
  - 2) Replace country fixed effects with firm-level fixed effects to account for unobserved firm heterogeneity (e.g., broadband access, manager education) that may correlate with digital adoption and performance.
- Results:
  - Both alternative specifications present similar trajectories of IR functions to the baseline, often with larger effect sizes.
  - With country- and industry-specific trends controlled, the IR function shows similar or slightly larger short-term impact on sales (t = 0 or 1) than baseline.
  - Firm-FE specification yields similar trajectories; profits and employment for the digitalized group remain positive and sizable gaps between digitalized and less-digitalized firms persist.
  - Overall, robustness checks confirm main findings on mitigation effects of firm digitalization.

- Additional methodological detail:
  - For the binned scatter in Figure 3, sample divided into 100 equally-sized bins; each dot is the binned average for 2015.

*Italic: Source — Excerpt from IMF Working Paper chapter/section titled "9. We overview cross-country differences in digitalization speed using the World Bank’s Digital Adoption Index (DAI)"*

### 22. Building on these results, we further examine if pre-shock corporate performance (profitability)

### 22. Building on these results, we further examine if pre-shock corporate performance (profitability)

### Zombification, digitalization, and heterogeneous scarring
- Definition used: “loss-makers” = firms consistently making losses for at least three consecutive years and older firms (older than ten years of firm age) (often called “zombies” in literature).
- Hypothesis: digital and disruptive technologies (e.g., artificial intelligence (AI), machine learning) help avoid zombification of small-medium enterprises by facilitating productivity expansion, market access, and cost reduction through production innovation.
- Empirical finding (firm-level micro data, tracking profitability from three years prior to the shock):
  - For loss-making firms without digitalization, the impulse response (IR) function exhibits larger and longer negative scarring effects on sales and profits than other firms and remains negative for a longer period.
  - Heterogeneity in terms of employment is not significant for loss-making firms.
- Comparative note: non loss-making firms show smaller and shorter negative impacts on sales and profit margins relative to loss-makers.

### Policy implications for fiscal support to firms hit by shocks
- From an individual firm perspective:
  - Fiscal support is needed to keep businesses alive during the crisis; loss-making firms, especially “zombies,” urgently need government liquidity support to survive.
- From a government/fiscal-risk perspective:
  - Given the magnitude of scarring on sales and profits for less-digitalized loss-making firms (larger and longer), prospects of recovery may be limited.
  - Providing loans or guarantees to such firms could result in accumulation of fiscal risks.
  - Grants or other support forms may not improve business prospects for these firms in light of significant sales impacts.
- Policy recommendation:
  - Digitalization should be a core aspect in designing the targeting rule (besides income-based targeting) of emergency fiscal supports during crises.

### Post-COVID firm dynamics — application of results and case-study synthesis (Box 1)
- Caveat: COVID-19 introduced a unique structural break; out-of-sample forecasting using pre-COVID data is challenging.
- Approach: use pre-COVID firm-level estimates together with available data on the size of the COVID uncertainty shock (WUI index assumed to capture common features of historical global shocks).
- Key implications and empirical validation:
  - Expect larger and longer scars on firms after the pandemic—particularly on less-digitalized firms in service sectors—than in past crises.
  - The size of the COVID uncertainty shock is about four times larger than the standardized 10 percent shock assumed in the baseline analysis.
  - The pandemic limits face-to-face interaction; contact-intensive sectors (transportation, accommodation, food services) are hit hardest.
  - Micro-level evidence from recent literature supports:
    - (a) deeper reduction in firm sales and employment after COVID-19 relative to past crises; and
    - (b) significant role of firm digitalization in mitigating adverse shocks, particularly after COVID-19.
- Selected findings from Box Table 1 (literature synthesis):
  - Abidi, Herradi, Sakha (2022), Middle East and Central Asia (13 countries): DID — Digitalized firms faced smaller sales drop by about 4 p.p than less-digitalized firms.
  - IMF (2022a), Copestake, Estefania-Flores and Furceri (2022), Asia and Pacific: DID — Firms in more digitalized industries recorded 3.4 percent higher sales revenues in a year; hiring rates are higher in industries using more digital skills; for past recessions the benefit on sales is smaller at 1.4 percent and largest after two years.
  - Cirera et al (2021), 38 mainly developing countries: Descriptive — COVID-19 accelerated digitalization unevenly; 44 percent of businesses increased digital adoption.
  - Kawaguchi, Kitao, Nose (2022), Japan: Bartik IV — Firms initially adopting remote work faced smaller sales drop by 1.5 percent (1.9 percent for low-contact industries) to 10 percent mobility shock.
  - Bai et al (2021), US: DID — Firms with high pre-COVID Work-from-Home index had significantly higher sales, net income, and stock returns during COVID-19.
  - Pierri and Timmer (2020), US: LPM — Local ICT adoption reduced an increase in unemployment rate by 2 p.p.
  - Cong, Yang, Zhang (2021), China: DID — Digitalized firms experienced smaller sales decline, sustained cash flow, and quick reopening.

### Role of government and fiscal policy in promoting firm-level digitalization
- Government roles highlighted:
  - Adoption of digital technologies in fiscal operations (“GovTech”) can improve public service efficiency.
  - GovTech may facilitate economy-wide digital adoption by encouraging broader use of digital solutions by citizens and firms.
  - Examples include fintech application to government payments leading to formalization of small business and female formal labor force participation, widening digital use in the economy.
  - Governments can promote digitalization through upskilling, adoption of disruptive technologies in business transactions, and support to B2B start-ups with innovation digital solutions.
- Empirical exploration uses the fiscal DTR index as a comparable cross-country measure of fiscal institutions supporting adoption of modern technology.

### Fiscal DTR index: definition and empirical strategy
- The DTR index (compiled by the European Centre for International Policy Economy (ECIPE)):
  - Covers 64 countries (in both AEs and EMDEs).
  - Measures the level of regulations a country imposes on digital transactions (lower values indicate “digital friendliness”); takes value from 0 (supportive) to 1 (restrictive).
- The fiscal DTR (fiscal restriction) index is disaggregated into three sub-components:
  - (a) tax regime on digital goods and online services;
  - (b) tariffs on ICT trade (custom policy); and
  - (c) procurement rules on digital products.
- Empirical model:
  - Due to censoring of the Digital Adoption Index (DAI), a two-limit Tobit model is used with DTR quartile dummies (1st and 2nd quartiles as less restrictive groups), controlling for macrovariables (log GDP per capita, tertiary education completion rate, broadband access, financial development), region and year fixed effects.

### Cross-country empirical findings on fiscal institutions and digital adoption
- Stylized fact: strong cross-country negative correlation between a country’s digitalization level (World Bank’s business Digital Adoption Index) and fiscal restriction on digital activities, controlling for income level and regional differences.
  - Interpretation: more digitalization is associated with less fiscal restriction on digital activities.
- Tobit model results (summary):
  - Coefficients for the top groups (fiscally least restrictive) in terms of taxation scheme and procurement are positive and statistically significant.
  - These top groups tend to significantly increase digital adoption by about 3 percentage points and 6 percentage points, respectively, compared with the groups of the bottom half.
  - Custom regulations on ICT trade appear less significant.
- Policy interpretation:
  - Aligning tax regimes with international tax standards and having competitive bidding on digital goods procurement (in line with Fiscal Transparency Evaluation standard, removing barriers for cross-border digital investment, facilitating e-commerce, reducing corruption) seem effective in supporting firm digitalization.

### Caveats and limitations
- DTR index limitations:
  - The DTR index continues to be updated but historical records are not available; data availability is limited to measuring the current state of government policy.
  - Sources for DTR come from publicly available legal and policy documents; there could be scope to expand the fiscal policies and regulations covered.
  - Cross-country DTR scores are assigned by ECIPE on the basis of policy alignment with international standards and best practices; the rating process inevitably involves some level of subjectivity.
- Causal inference:
  - Given the above caveats, the finding cannot claim causal relationships between government fiscal intervention and firm-level digitalization; the policy discussion should be treated as suggestive evidence.

### Conclusions (section VI)
- The paper uses global firm data and new firm-level digitalization measures to empirically examine:
  - The benefit of digitalization in strengthening firm-level resilience to global crises while quantifying the size of the digital divide.
  - What fiscal policies could effectively drive firm-level digitalization.

*IMF Working Paper excerpt: Firm-level Digitalization and Resilience to Shocks: Role of Fiscal Policy*

### 32. Our findings point the benefit of digitalization to mitigate the adverse impact of uncertainty

### 32. Our findings point the benefit of digitalization to mitigate the adverse impact of uncertainty

### Main findings on digitalization and uncertainty shocks
- Digitalization mitigates the adverse impact of uncertainty shocks, creating significant heterogeneity between digitalized and less-digitalized firms.
- Digitalized firms may improve profitability following the shock while less-digitalized ones are worse off; this heterogeneity is particularly salient in the service sector.
- Among loss-making firms, lack of digitalization makes the negative shock even larger and longer-lasting.
- After a "10 percent adverse shock ( broadly equivalent to the magnitude of the shock after major crises in recent history above historical median level)", on average, firm-level sales, profit margins, and employment are expected to gradually start recovering "three years after the shock".
- COVID-19 had prolonged impacts on firms’ profitability, particularly in less-digitalized firms, affecting the pace of their tax recovery.

### Role of fiscal policy in promoting digitalization
- Fiscal policy measures can play an important role in facilitating firm-level digitalization.
- Cross-country analyses suggest that fairness and even-handedness of fiscal institutions related to taxation and procurement rules are critical drivers of firm-level digitalization.
- Specific fiscal levers discussed:
  - Relaxing barriers for firms’ digital investment through less restrictive tax regime, tax incentives and subsidies.
  - Open and competitive public procurement (e.g., through e-procurement and less stringent regulations on tender criteria for procuring digital goods) to promote digitalization in private business.

### Policy implications (drawn directly from the findings)
- First: Policymakers should anticipate that scarring effects after uncertainty shocks can be long lasting.
  - Key empirical timing: gradual recovery of sales, profit margins, and employment begins "three years after the shock" following a "10 percent adverse shock ( broadly equivalent to the magnitude of the shock after major crises in recent history above historical median level)".
- Second: Firm-level digitalization strengthens resilience to uncertainty shocks.
  - Micro: digitalized firms are more resilient.
  - Macro: broader digitalization across firms strengthens economy-wide resilience.
  - Recommendation: closely monitor and properly measure the status of firm-level digitalization to assess resilience.
- Third: The government’s role to promote firm-level digitalization should be further explored.
  - Fiscal levers: less restrictive tax regimes, tax incentives, subsidies.
  - Procurement: open and competitive public procurement (e.g., e-procurement, less stringent tender criteria).
- Fourth: Account for sizable heterogeneity when designing firm support in response to uncertainty shocks.
  - Some digitalized firms show better performance post-shock while some less-digitalized loss-making firms suffer larger and longer-lasting scarring effects with little prospect for recovery.
  - Policy implication: target government support carefully; manage fiscal risks where ex-post fiscal support (e.g., loans and/or guarantees) is delivered to less-digitalized firms with weak business prospects.

### Robustness checks (baseline vs. alternative trend specifications)
- Log sales (Years after shock 0 1 2 3 4):
  - Baseline: -0.005 -0.080*** -0.149*** -0.189*** -0.138***
  - w/ industry trend: -0.005 -0.080*** -0.149*** -0.190*** -0.140***
  - w/ country trend: -0.005 -0.090*** -0.189*** -0.239*** -0.172***
  - w/ both trends: -0.005 -0.090*** -0.188*** -0.238*** -0.171***
- Profit Margins (Years after shock 0 1 2 3 4):
  - Baseline: -0.012*** -0.025*** -0.032*** -0.032*** -0.009**
  - w/ industry trend: -0.012*** -0.025*** -0.032*** -0.032*** -0.009**
  - w/ country trend: -0.014*** -0.027*** -0.033*** -0.032*** -0.007
  - w/ both trends: -0.014*** -0.027*** -0.033*** -0.032*** -0.007
- Log employment (Years after shock 0 1 2 3 4):
  - Baseline: -0.047*** -0.064*** -0.106*** -0.143*** -0.110***
  - w/ industry trend: -0.046*** -0.064*** -0.105*** -0.142*** -0.110***
  - w/ country trend: -0.071*** -0.130*** -0.214*** -0.259*** -0.173***
  - w/ both trends: -0.071*** -0.129*** -0.214*** -0.259*** -0.173***

### Regression evidence: impact of fiscal interventions on digitalization (selected coefficients)
- Taxation:
  - Tax_q1: 0.030 *
  - Tax_q2: 0.004 (not significant in shown column)
- Procurement:
  - Procurement_q1: 0.054 ***
  - Procurement_q2: 0.038 ***
- Tariffs:
  - Tariff_q1: 0.027 (standard error 0.017)
  - Tariff_q2: 0.008 (standard error 0.014)
- Controls and notable coefficients:
  - Log GDP per capita (in PPP) (lagged): 0.133 ***
  - Broadband subscribers: 0.453 ***
  - GCI-Financial Market Development index: 0.020 *
- Sample sizes reported in table: N 124, 96, 123 (across columns)
- Region & Year fixed effects: YES

### Sample composition (selected counts and shares)
- Total number of firms: 1,968,918 (100)
- By country (selected):
  - France: 448,408 22.77
  - Spain: 432,447 21.96
  - Bulgaria: 156,429 7.94
  - Italy: 143,109 7.27
  - Russia: 86,097 4.37
  - Czech Republic: 78,919 4.01
- By sector (selected):
  - Wholesale and retail trade: 523,158 26.57
  - Construction: 283,029 14.37
  - Other business sector services: 276,009 14.02
  - Accommodation and food services: 113,970 5.79
  - Real estate activities: 99,140 5.04
  - IT and other information services (ICT-related): 45,358 2.30

### Areas for further study
- Further exploration of the role of fiscal policy in digitalization with more granular data to obtain more robust empirical evidence.
- Exploration of post-COVID firm dynamics as more post-COVID data become available.
- A companion paper (Nose, Honda, et al, forthcoming) will further examine tax recovery.

*IMF Working Paper: Firm-level Digitalization and Resilience to Shocks: Role of Fiscal Policy (Working Paper No. WP/2023/095)*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023095-print-pdf.pdf_