## _wp04125 - References

## Source details

**Canonical URL:** [_wp04125 - References](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04125.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04125.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04125.pdf.json)

---

### Introduction and context
- 1994 fiscal reforms: 75 percent of VAT revenue assigned to the central government; remaining 25 percent of VAT and business tax assigned to local governments.
- A “revenue-returned mechanism” transferred back 30 percent of the increase in VAT and excise tax revenue to the originating province.
- China’s VAT is production-based (P-VAT); most services are taxed by a non-creditable business tax accruing to local governments.
- Moving to a consumption-based VAT (C-VAT) (excluding capital goods) would reduce distortions but imply overall revenue loss for subnational governments.
- Extending VAT to services would shift revenue from local business-tax receipts to VAT receipts that are shared between center and provinces.

### Method and assumptions for simulations
- Data year: 2001 (simulations use 2001 data and 1997 input-output coefficients for M&E purchases).
- Behavioral responses excluded (first-round effects only).
- VAT revenue defined as: tax base × tax rate × collection efficiency.
- Standard VAT rate used for simulations: 17 percent.
- Collection-efficiency bounds: 100 percent (upper) and 50 percent (lower).
- C-VAT simulations allow VAT credits for machinery and equipment (M&E) only; construction/buildings treated as VAT-exempt.

### C-VAT reform — Base scenario: provincial losses (Table 1; Figure 1)
- Aggregate figures (from Table 1):
  - Loss of VAT Revenue (100 percent CE) = 41.87 (in billions of yuan)
  - Loss of VAT Revenue (50 percent CE) = 20.94 (in billions of yuan)
  - 2001 VAT Revenue (provincial current receipts) = 134.17 (in billions of yuan)
  - Revenue Loss as Percent of Current VAT Revenue = 30.91 (100 percent CE) and 15.45 (50 percent CE)
- Selected provincial examples (Table 1):
  - Beijing: Calculated Purchases of M&E = 19.03; Loss of VAT Revenue (100 percent CE) = 0.81; (50 percent CE) = 0.40; 2001 VAT Revenue = 5.90; Revenue Loss as Percent = 13.71 (100 percent CE), 6.85 (50 percent CE).
  - Hebei: Calculated Purchases of M&E = 55.15; Loss of VAT Revenue (100 percent CE) = 2.34; (50 percent CE) = 1.17; 2001 VAT Revenue = 4.88; Revenue Loss as Percent = 47.98 (100 percent CE), 23.99 (50 percent CE).
  - Guangdong: Calculated Purchases of M&E = 107.58; Loss of VAT Revenue (100 percent CE) = 4.57; (50 percent CE) = 2.29; 2001 VAT Revenue = 17.25; Revenue Loss as Percent = 26.51 (100 percent CE), 13.25 (50 percent CE).
  - Tibet: Calculated Purchases of M&E = 0.36; Loss of VAT Revenue (100 percent CE) = 0.02; (50 percent CE) = 0.01; 2001 VAT Revenue = 0.08; Revenue Loss as Percent = 18.23 (100 percent CE), 9.11 (50 percent CE).
- Distributional patterns:
  - Coefficient of variation of percentage losses across provinces ≈ 0.31.
  - For 100 percent CE: average provincial loss ≈ 30 percent; peaks include Hubei 53.93 percent; Hebei 47.98 percent.
  - For 50 percent CE: average loss ≈ 15 percent; losses below 10 percent for Beijing, Shanghai, and Tibet.
  - Scatter (Figure 1): roughly inverted “U” relationship between provincial GDP per capita and percentage revenue loss (50 percent CE shown); middle-income provinces tend to lose more.
- Central government impact:
  - Estimated reduction in central government’s 75 percent share of VAT revenue due to C-VAT: between Y 62.8 billion and Y 125.6 billion (depending on collection efficiency).
  - 2001 total VAT revenue ≈ Y 536 billion; central government’s share ≈ Y 402 billion.

### Revised VAT revenue-sharing rule scenario (compensate provinces by raising provincial VAT share) (Tables 2–3)
- Objective: increase the share of VAT revenue assigned to local governments so aggregate provincial VAT receipts remain at their initial level (revenue-neutral for provinces in aggregate).
- Required provincial VAT share (Table 2):
  - Revenue-Neutral VAT Share for Provinces = 29.6 (In percent) at 50 percent CE.
  - Revenue-Neutral VAT Share for Provinces = 36.3 (In percent) at 100 percent CE.
- Fiscal cost to central government (Table 2):
  - Revenue Loss of Central Government (50 percent CE) = 114.17 (in billions of yuan) = 28.4 percent of 2001 VAT Revenue.
  - Revenue Loss of Central Government (100 percent CE) = 142.67 (in billions of yuan) = 35.5 percent of 2001 VAT Revenue.
- Province examples under revised sharing rule (selected from Table 3):
  - Beijing: Current VAT Revenue = 5.90; Post-Reform VAT Revenue (50 percent CE) = 6.51; Net Revenue Gain = 0.61; Revenue Gain as Percent = 10.37 (50 percent CE), 25.43 (100 percent CE).
  - Hebei: Current VAT Revenue = 4.88; Post-Reform VAT Revenue (50 percent CE) = 4.40; Net Revenue Gain = -0.49; Revenue Gain as Percent = -9.94 (50 percent CE), -24.39 (100 percent CE).
  - Guangdong: Current VAT Revenue = 17.25; Post-Reform VAT Revenue (50 percent CE) = 17.73; Net Revenue Gain = 0.48; Revenue Gain as Percent = 2.79 (50 percent CE), 6.83 (100 percent CE).
  - Tibet: Current VAT Revenue = 0.08; Post-Reform VAT Revenue (50 percent CE) = 0.09; Net Revenue Gain = 0.01; Revenue Gain as Percent = 7.69 (50 percent CE), 18.87 (100 percent CE).
- Distributional consequence:
  - Correlation between GDP per capita and gains: 0.623 for 50 percent CE; 0.548 for 100 percent CE (richer provinces favored).

### Raising the standard VAT rate scenario (compensate provinces by higher VAT rate) (Table 4)
- Required revenue-neutral standard VAT rates:
  - Revenue-Neutral Standard VAT Rate = 20.1 (In percent) at 50 percent CE (from 17 percent).
  - Revenue-Neutral Standard VAT Rate = 24.7 (In percent) at 100 percent CE (from 17 percent).
- Fiscal effect to center: Revenue Loss of Central Government = 0 (in billions of yuan); Revenue Loss of Central Government as Percent of 2001 VAT Revenue = 0 (In percent) under both calculated revenue-neutral rates.
- Practical considerations:
  - Required rate increases are sharp and could raise prices and reduce consumption demand.
  - Second-round behavioral effects not modeled; a sharp VAT rate increase would likely reduce the VAT base short run, undermining revenue estimates.

### Extension of VAT to services — method and base scenario (Tables 5–6; Figure 2)
- Finance and insurance excluded from VAT extension (assumed to remain under business tax); finance and insurance accounted for 28 percent of business tax proceeds in 2001.
- Services added to VAT base include: services to farming, forestry and fishery; geological prospecting and water conservancy; transport, storage, post and telecommunications services; real estate, other; and one-third of selected social, education, scientific, culture categories (as classified in the source).
- Revenue accounting (Table 5 columns described in text):
  - Column 1: addition to VAT base from service taxation (excluding finance and insurance).
  - Column 2: loss in VAT base due to removal of cascading (VAT paid on purchases by currently VAT-liable sectors becomes creditable).
  - Column 3: net increase = Column 1 − Column 2.
  - Columns 4 and 5: implied VAT revenue gains for 100 percent and 50 percent collection efficiencies.
- Selected provincial results (Table 5 amounts in hundreds of millions of yuan):
  - Heilongjiang:
    - 2001 Value-Added in Newly Taxable Service Sectors = 720.85
    - Net Increase in VAT Base = 620.88
    - Increase in VAT Revenue (100 percent efficiency) = 26.39
    - Increase in VAT Revenue (50 percent efficiency) = 13.19
    - 2001 Revenue from Business Tax, Excluding Finance and Insurance = 34.59
    - Gain from Reform (100 percent efficiency) = -8.20
    - Gain from Reform (50 percent efficiency) = -21.40
  - Beijing:
    - 2001 Value-Added in Newly Taxable Service Sectors = 771.13
    - Net Increase in VAT Base = 718.93
    - Increase in VAT Revenue (100 percent efficiency) = 30.55
    - Increase in VAT Revenue (50 percent efficiency) = 15.28
    - 2001 Revenue from Business Tax, Excluding Finance and Insurance = 40.91
    - Gain from Reform (100 percent efficiency) = -10.36
    - Gain from Reform (50 percent efficiency) = -25.63
  - Shanghai:
    - 2001 Value-Added in Newly Taxable Service Sectors = 1109.98
    - Net Increase in VAT Base = 979.69
    - Increase in VAT Revenue (100 percent efficiency) = 41.64
    - Increase in VAT Revenue (50 percent efficiency) = 20.82
    - 2001 Revenue from Business Tax, Excluding Finance and Insurance = 71.58
    - Gain from Reform (100 percent efficiency) = -29.95
    - Gain from Reform (50 percent efficiency) = -50.76
  - Guangdong:
    - 2001 Value-Added in Newly Taxable Service Sectors = 2657.72
    - Net Increase in VAT Base = 2381.62
    - Increase in VAT Revenue (100 percent efficiency) = 101.22
    - Increase in VAT Revenue (50 percent efficiency) = 50.61
    - 2001 Revenue from Business Tax, Excluding Finance and Insurance = 263.06
    - Gain from Reform (100 percent efficiency) = -161.84
    - Gain from Reform (50 percent efficiency) = -212.45
- Aggregate totals (Table 5):
  - Aggregate increase in VAT revenue (100 percent efficiency) = 881.38 (hundreds of millions of yuan)
  - Aggregate increase in VAT revenue (50 percent efficiency) = 440.69 (hundreds of millions of yuan)
  - Aggregate calculated loss of business tax revenue = 1324.19 (hundreds of millions of yuan)
  - Aggregate net gain from reform (100 percent efficiency) = -442.81 (hundreds of millions of yuan)
  - Aggregate net gain from reform (50 percent efficiency) = -883.50 (hundreds of millions of yuan)
- Distributional note:
  - Business tax accrues entirely to local governments while VAT is shared; reform could generate substantial center gains (center receives 75 percent of VAT).
  - Center’s additional revenue from VAT services accruing to provinces would be three times between Y 44 billion and Y 88 billion, i.e., between Y 132 billion and Y 264 billion (as presented in source).

### Revised sharing rule for services extension (Table 6)
- Objective: increase VAT sharing rate from 25 percent to make the reform revenue-neutral for provinces in aggregate.
- Calculated new sharing rate:
  - Approximately 0.30 at 100 percent VAT efficiency.
  - Approximately 0.37 at 50 percent VAT efficiency.
- Selected outcomes (Table 6 amounts in hundreds of millions of yuan):
  - Heilongjiang:
    - Current Revenue from VAT and Business Tax = 88.86
    - Post-Reform Revenue (100 percent efficiency) = 96.39
    - Post-Reform Revenue (50 percent efficiency) = 100.06
    - Gain from Reform (100 percent) = 7.52
    - Gain from Reform (50 percent) = 11.19
    - Gain percent (100 percent) = 8.46
    - Gain percent (50 percent) = 12.60
  - Beijing:
    - Current Revenue = 240.34
    - Post-Reform (100 percent) = 247.82
    - Post-Reform (50 percent) = 251.53
    - Gain (100 percent) = 7.48
    - Gain (50 percent) = 11.18
    - Gain percent (100 percent) = 3.11
    - Gain percent (50 percent) = 4.65
  - Guangdong:
    - Current Revenue = 490.64
    - Post-Reform (100 percent) = 383.32
    - Post-Reform (50 percent) = 388.77
    - Gain (100 percent) = -107.32
    - Gain (50 percent) = -101.87
    - Gain percent (100 percent) = -21.87
    - Gain percent (50 percent) = -20.76
  - Totals:
    - Current Revenue from VAT and Business Tax (national total) = 3,190.76 (hundreds of millions of yuan)
    - Post-Reform Revenue under revenue-neutral sharing rule (both efficiency assumptions) = 3,190.76 (hundreds of millions of yuan)
    - Aggregate national gain from reform under revenue-neutral sharing rule = 0.00 (hundreds of millions of yuan)

### Revenue-returned (RR) system: formula, properties, and historical data (Tables 7–9)
- RR rule: 30 percent of the increase in VAT and consumption tax collection over their 1994 base is returned to the originating province.
- Incremental mechanics:
  - The RR formula grows one third as fast as VAT and consumption tax revenue, so revenue-returned as a proportion of tax revenue will tend to zero in the long run, but the decline can be slow.
- Historical returned revenue (Table 7; in billions of yuan):
  - Excises: 1996 = 62.0; 1997 = 67.4; 1998 = 81.5; 1999 = 82.0; 2000 = 85.729; 2001 = 92.946
  - VAT: 1996 = 296.3; 1997 = 327.9; 1998 = 362.8; 1999 = 387.8; 2000 = 455.251; 2001 = 535.622
  - Total collection: 1996 = 358.3; 1997 = 395.3; 1998 = 444.3; 1999 = 469.9; 2000 = 541.0; 2001 = 628.6
  - RR actual: 1996 = 194.9; 1997 = 201.2; 1998 = 208.3; 1999 = 216.73; 2000 = 228.2; 2001 = 234
  - RR formula: 2000 = 226.6; 2001 = 239.3
- RR as percent of total collection (selected):
  - RR actual: 1996 = 50.9; 1997 = 46.9; 1998 = 46.1; 1999 = 42.2; 2000 = 37.2
  - RR formula: 1996 = 50.8; 1997 = 47.0; 1998 = 45.1; 1999 = 41.9; 2000 = 38.1
- Incremental tax share (Table 8; examples for 2001, in percent):
  - Average incremental tax share (2001) = 14.44 (percent)
  - Heilongjiang = 8.70
  - Jiangsu = 9.97
  - Zhejiang = 9.07
  - Guangdong = 9.59
  - Shanghai = 12.47
  - Beijing = 11.50
  - Hunan = 20.53
  - Guizhou = 20.74
  - Gansu = 22.54
  - Yunnan = 28.51
- “Half-life” of RR importance (Table 9):
  - Growth Rate of Tax Revenues g = 0.05 → Number of Years = 20.45
  - g = 0.10 → Number of Years = 10.54
  - g = 0.15 → Number of Years = 7.24

### Inclusion of RR in C-VAT reform: scenarios and quantified impacts (Table 10)
- Two RR scenarios when C-VAT reform reduces VAT base:
  1. No change in revenue-returned: reform occurs in weaker growth so VAT + consumption tax show negative growth after reform and the “overall increase” constraint binds; RR not reduced.
  2. Change in revenue-returned: reform occurs in strong growth so VAT + consumption tax still grow and RR falls.
- Quantitative impact when RR falls (summary from Table 10 and text):
  - Applying 25 percent plus incremental tax shares (Table 8) and assuming 50 percent collection efficiency, average percentage VAT revenue losses increase by about 5 percentage points:
    - Average loss rises from 15.40 percent to 20.07 percent (as presented).
  - Table 10 totals and averages (In hundreds of millions of yuan) summarized in source: Totals and averages show higher average percentage losses when RR is adjusted downward.

### Key conclusions and policy options (final findings)
- Main conclusions:
  - Reforming indirect taxation (move to a C-VAT excluding capital goods, and extending VAT to services) would generate considerable revenue losses for local governments and also generate losses for the central government.
  - Extending VAT to services would yield substantial revenue gains to the center but would imply elimination of the business tax on most services—an important local revenue source—making reform politically difficult without compensation.
  - The revenue-returned formula is not transparent and differentiates regions based on past history rather than current revenue capacities or needs; it is only temporarily disequalizing but may take decades to shrink materially in importance.
  - Simple uniform sharing of VAT revenue is transparent but permanently disequalizing.
- Compensation and reform options discussed:
  - Changing parameters of the current system (larger provincial share of VAT or increasing the standard VAT rate) would not adequately target provinces with revenue shortfalls and would be disequalizing.
  - More effective option: extend and redesign equalization grants introduced in 1994, using expenditure needs and revenue capacities to target transfers so that changes in province revenue-raising capacity are picked up in the formula and transfers adjusted accordingly.
  - Complementary measures to facilitate acceptance of a more redistributive system:
    - Joint reassessment of expenditure responsibilities together with revamped revenue assignments.
    - More transparent access to capital markets.
  - Tax assignment and revenue instruments:
    - Provide provinces and counties with margin control over assigned tax rates to improve fiscal accountability, while avoiding excessive distortions and tax competition.
    - Standardize the base for the personal income tax, allowing bounded piggybacking for provinces.
    - Expand use of property taxes at the county level, enhance valuation and recording for leasehold properties, and base annual taxes on the annual lease value equivalent.
  - Spending-side adjustments:
    - Some middle-income counties face pension liabilities and unfunded mandates; recentralization of pension liabilities and unemployment insurance should be considered.
- Implementation note:
  - The recommended reforms form an extensive, interlocking package that will take several years to implement; scope should be comprehensive though pace can be gradual.

*Source: _wp04125 - References..............................................................................................................*

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

### _wp04125 - References..............................................................................................................

### Tables
- 1.   C-VAT Reform Base Scenario ............................................................................................5
- 2.   Revised VAT Revenue-Sharing Rules.................................................................................7
- 3.   C-VAT Reform, Revised Sharing Rule Scenario ................................................................9
- 4.   Revised Standard Rate of VAT..........................................................................................10
- 5.   Extension of VAT to Services, Base Scenario...................................................................12
- 6.   Extension of VAT to Services ...........................................................................................15
- 7.   Returned Revenue, 1996–2001 ..........................................................................................16
- 8.   Incremental Tax Shares by Region, 2001 ..........................................................................18
- 9.   The “Half-Life” of the Revenue-Returned Formula ..........................................................19
- 10.   C-VAT Reform Base Scenario Allowing for Revenue-Returned....................................21

### Figures
- 1.   Provincial Percentage Revenue Losses from C-VAT Reform ............................................7
- 2.   Provincial Losses from Extension of VAT to Services .....................................................13
- 3.   Incremental VAT Shares by Province ...............................................................................19

### Major themes reflected in the table and figure titles
- C-VAT Reform and related scenarios (Base Scenario; Revised Sharing Rule Scenario; Base Scenario Allowing for Revenue-Returned).
- VAT revenue-sharing rules and formulas (Revised VAT Revenue-Sharing Rules; The “Half-Life” of the Revenue-Returned Formula).
- Changes to VAT structure (Revised Standard Rate of VAT; Extension of VAT to Services).
- Geographic distribution of VAT impacts (Provincial Percentage Revenue Losses; Provincial Losses from Extension of VAT to Services; Incremental VAT Shares by Province; Incremental Tax Shares by Region).
- Historical revenue data referenced (Returned Revenue, 1996–2001).

*Source: _wp04125 - References..............................................................................................................*

### 4.   Provincial Revenue Losses from C-VAT Reform .............................................................20

### 4.   Provincial Revenue Losses from C-VAT Reform

### I. Introduction
- 1994 fiscal reforms: 75 percent of VAT revenue assigned to the central government; remaining 25 percent of VAT and business tax assigned to local governments.
- A “revenue-returned mechanism” transferred back 30 percent of the increase in VAT and excise tax revenue to the originating province.
- China’s VAT is production-based (P-VAT) because capital goods are in the VAT base; most services are taxed by a non-creditable business tax accruing to local governments.
- P-VAT and non-creditable business tax create cascading and arbitrary effective tax rates across consumption items; reform toward a consumption-based VAT (C-VAT) would reduce distortions but imply overall revenue loss because capital goods would be excluded from the VAT base.
- Moving to C-VAT and extending VAT to services would change revenue distribution between center and provinces; magnitude of revenue losses varies across provinces and requires compensation mechanisms and transfer-system reform.

### II. Switch to Consumption-Based VAT — Method and Assumptions
- Simulations use 2001 data and exclude behavioral responses (first-round effects only).
- VAT revenue defined as: tax base x tax rate x collection efficiency.
- Standard VAT rate for simulations: 17 percent.
- Collection-efficiency bounds used: 100 percent (upper) and 50 percent (lower). Collection efficiency captures tax evasion, use of lower 13 percent VAT in some sectors, exemptions for smaller enterprises, and administrative noncollection.
- C-VAT simulations focus on allowing VAT credits for machinery and equipment (M&E) only (construction/buildings treated as VAT-exempt because construction is taxed under business tax), implying an underestimate of total provincial revenue losses but approximating relative provincial losses.

### III. Base Scenario — Provincial Loss Estimates (summary of Table 1)
- Calculated purchases of M&E by taxable sectors for each province using 1997 input-output coefficients applied to 2001 provincial sector output (secondary industry excluding construction; wholesale, retail, and catering).
- Aggregate loss figures (from Table 1):
  - Totals: Loss of VAT Revenue (100 percent CE) = 41.87 (in billions of yuan)
  - Totals: Loss of VAT Revenue (50 percent CE) = 20.94 (in billions of yuan)
  - Totals: 2001 VAT Revenue (provincial current receipts) = 134.17 (in billions of yuan)
  - Averages: Revenue Loss as Percent of Current VAT Revenue = 30.91 (100 percent CE) and 15.45 (50 percent CE)
- Selected provincial examples (from Table 1):
  - Beijing: Calculated Purchases of M&E = 19.03; Loss of VAT Revenue (100 percent CE) = 0.81; (50 percent CE) = 0.40; 2001 VAT Revenue = 5.90; Revenue Loss as Percent of Current VAT Revenue = 13.71 (100 percent CE), 6.85 (50 percent CE).
  - Hebei: Calculated Purchases of M&E = 55.15; Loss of VAT Revenue (100 percent CE) = 2.34; (50 percent CE) = 1.17; 2001 VAT Revenue = 4.88; Revenue Loss as Percent = 47.98 (100 percent CE), 23.99 (50 percent CE).
  - Guangdong: Calculated Purchases of M&E = 107.58; Loss of VAT Revenue (100 percent CE) = 4.57; (50 percent CE) = 2.29; 2001 VAT Revenue = 17.25; Revenue Loss as Percent = 26.51 (100 percent CE), 13.25 (50 percent CE).
  - Tibet: Calculated Purchases of M&E = 0.36; Loss of VAT Revenue (100 percent CE) = 0.02; (50 percent CE) = 0.01; 2001 VAT Revenue = 0.08; Revenue Loss as Percent = 18.23 (100 percent CE), 9.11 (50 percent CE).
- Distributional patterns:
  - Coefficient of variation of percentage losses across provinces ≈ 0.31.
  - For 100 percent collection efficiency: average provincial loss ≈ 30 percent, with peaks (e.g., Hubei 53.93 percent; Hebei 47.98 percent).
  - For 50 percent collection efficiency: average loss ≈ 15 percent; losses below 10 percent for Beijing, Shanghai, and Tibet.
  - Scatter (Figure 1): roughly inverted “U” relationship between provincial GDP per capita and percentage revenue loss (50 percent CE shown), with middle-income provinces losing more; three outlying provinces drive the inverted-U shape.

- Central government impact (aggregate):
  - Estimated reduction in central government’s 75 percent share of VAT revenue due to C-VAT: between Y 62.8 billion and Y 125.6 billion, depending on collection efficiency assumption.
  - For comparison, 2001 total VAT revenue ≈ Y 536 billion; central government’s share ≈ Y 402 billion.

### IV. Revised Sharing-Rule Scenario (compensate provinces by raising provincial VAT share)
- Objective: increase the share of VAT revenue assigned to local governments so aggregate provincial VAT receipts remain at their initial level (revenue-neutrality for provinces in aggregate).
- Required provincial VAT share (Table 2):
  - Low collection efficiency (50 percent): Revenue-Neutral VAT Share for Provinces = 29.6 (In percent)
  - High collection efficiency (100 percent): Revenue-Neutral VAT Share for Provinces = 36.3 (In percent)
- Fiscal cost to central government (Table 2):
  - Revenue Loss of Central Government (50 percent CE) = 114.17 (in billions of yuan) = 28.4 percent of 2001 VAT Revenue.
  - Revenue Loss of Central Government (100 percent CE) = 142.67 (in billions of yuan) = 35.5 percent of 2001 VAT Revenue.
- Province-by-province net impacts under revised sharing rule (selected from Table 3, 50 percent CE and 100 percent CE columns):
  - Beijing: Current VAT Revenue = 5.90; Post-Reform VAT Revenue (50 percent CE) = 6.51; Net Revenue Gain = 0.61; Revenue Gain as Percent = 10.37 (50 percent CE), 25.43 (100 percent CE).
  - Hebei: Current VAT Revenue = 4.88; Post-Reform VAT Revenue (50 percent CE) = 4.40; Net Revenue Gain = -0.49; Revenue Gain as Percent = -9.94 (50 percent CE), -24.39 (100 percent CE).
  - Guangdong: Current VAT Revenue = 17.25; Post-Reform VAT Revenue (50 percent CE) = 17.73; Net Revenue Gain = 0.48; Revenue Gain as Percent = 2.79 (50 percent CE), 6.83 (100 percent CE).
  - Tibet: Current VAT Revenue = 0.08; Post-Reform VAT Revenue (50 percent CE) = 0.09; Net Revenue Gain = 0.01; Revenue Gain as Percent = 7.69 (50 percent CE), 18.87 (100 percent CE).
- Distributional consequence:
  - Correlation between GDP per capita and gains from reform: strong positive (correlation coefficient 0.623 for 50 percent CE; 0.548 for 100 percent CE).
  - A higher VAT sharing rule favors richer provinces and may be disequalizing; it would not appropriately target resource-constrained provinces.

### V. Raising the Standard VAT Rate Scenario (compensate provinces by higher VAT rate)
- Alternative: raise standard VAT rate from 17 percent to achieve aggregate revenue-neutrality.
- Required revenue-neutral standard VAT rates (Table 4):
  - Low collection efficiency (50 percent): Revenue-Neutral Standard VAT Rate for Provinces = 20.1 (In percent) — Revenue Loss of Central Government = 0 (in billions of yuan) — Revenue Loss of Central Government as Percent of 2001 VAT Revenue = 0 (In percent).
  - High collection efficiency (100 percent): Revenue-Neutral Standard VAT Rate for Provinces = 24.7 (In percent) — Revenue Loss of Central Government = 0 — Revenue Loss as Percent = 0.
- Practical considerations:
  - Required rate increases are sharp and could raise prices and reduce consumption demand.
  - “Second-round” behavioral effects are not modeled; a sharp VAT rate increase would likely reduce the VAT base in the short run, undermining revenue estimates.

### VI. Extension of VAT to Services — Method and Base Scenario
- Extending VAT to services would replace the business tax (except finance and insurance, which are assumed to remain under business tax).
- Finance and insurance accounted for 28 percent of business tax proceeds in 2001; taxation of finance and insurance under VAT is excluded because transactions are hard to tax under VAT.
- Services added to the VAT base (Chinese classification): services to farming, forestry and fishery; geological prospecting and water conservancy; transport, storage, post and telecommunications services; real estate, other; and one-third of: social services, health care, sports and social welfare; education, culture, arts, radio, film and television; scientific research and polytechnic services.
- Revenue implications: 
  - Column 1 of Table 5 (not reproduced here) calculates the addition to the VAT base from service taxation (excluding finance and insurance).
  - Column 2 computes the loss in VAT base due to removal of cascading: value of purchases by sectors already liable for VAT (secondary industry and commerce) from sectors newly subject to VAT; VAT paid on those purchases becomes creditable and thus leaves the VAT base.
  - Column 3 shows the net increase as Column 1 minus Column 2.
  - Columns 4 and 5 compute implied VAT revenue gains for 100 percent and 50 percent collection efficiencies, respectively.
- Comparison to provincial losses:
  - The extension to services could largely offset losses from C-VAT, but distributional effects matter because the business tax accrues entirely to local governments while VAT would be shared.
  - The relative size of finance and insurance (which would remain under business tax) differs widely by region; e.g., in 2001 finance and insurance provided about 15 percent of provincial GDP in Beijing but much less elsewhere.

*Sources: China Statistical Yearbook; and authors’ calculations.*

### 0.9 percent of GDP in Heilongjiang.

### _wp04125 - 0.9 percent of GDP in Heilongjiang.

### Methodology for estimating provincial revenue effects
- Assumption: each province retains a share a_i of its business tax revenue, where (i) a_i is proportional to the share of finance and insurance in regional GDP; and (ii) overall, 28 percent of business tax revenue is retained.
- The formula for a_i is derived from: let R_i denote the business tax revenue in province i; a_i = λ s_i where s_i is the share of insurance and finance in province i; and ∑_i a_i R_i = 28.0. Solving yields a_i = (∑_i R_i s_i)^{-1} (0.28 ∑_i R_i s_i) expressed in the text as ∑_i (R_i s_i) a / 28.0 (textual derivation present).
- Provincial revenue loss from abolishing the business tax (except on finance and insurance) = (1 - a_i) × actual business tax revenue in province i.
- Overall gain from the reform = increase in VAT revenue (calculated at 100 percent or 50 percent VAT efficiency) minus calculated loss of business tax revenue.

### Provincial pattern of gains and losses (base scenario)
- Reform assessed under two VAT collection-efficiency assumptions: 100 percent efficiency and 50 percent efficiency.
- General observations from the base scenario:
  - Provinces with relatively large service sectors but where finance and insurance are unimportant tend to lose most (they lose all business tax revenue on these activities).
  - Provinces where finance and insurance are important tend to lose less or gain (because business tax revenue on these activities is preserved).
  - Visual analysis (Figure 2) shows no clear pattern by GDP per capita; ignoring Beijing and Shanghai, the reform appears mildly equalizing as middle-income provinces do a little worse on average than lower-income ones.
  - Note: Figure 2 assumes 50 percent collection efficiency.

- Selected provincial results from Table 5 (Extension of VAT to Services, Base Scenario; amounts in hundreds of millions of yuan):
  - Heilongjiang (Table 5 row):
    - 2001 Value-Added in Newly Taxable Service Sectors: 720.85
    - 2001 Total Purchases from Currently Exempt Sectors: 99.96
    - Net Increase in VAT Base: 620.88
    - Increase in VAT Revenue (100 percent efficiency): 26.39
    - Increase in VAT Revenue (50 percent efficiency): 13.19
    - 2001 Revenue from Business Tax, Excluding Finance and Insurance: 34.59
    - Gain from Reform (100 percent efficiency): -8.20
    - Gain from Reform (50 percent efficiency): -21.40
  - Beijing (Table 5 row):
    - 2001 Value-Added in Newly Taxable Service Sectors: 771.13
    - 2001 Total Purchases from Currently Exempt Sectors: 52.21
    - Net Increase in VAT Base: 718.93
    - Increase in VAT Revenue (100 percent efficiency): 30.55
    - Increase in VAT Revenue (50 percent efficiency): 15.28
    - 2001 Revenue from Business Tax, Excluding Finance and Insurance: 40.91
    - Gain from Reform (100 percent efficiency): -10.36
    - Gain from Reform (50 percent efficiency): -25.63
  - Shanghai (Table 5 row):
    - 2001 Value-Added in Newly Taxable Service Sectors: 1109.98
    - 2001 Total Purchases from Currently Exempt Sectors: 130.29
    - Net Increase in VAT Base: 979.69
    - Increase in VAT Revenue (100 percent efficiency): 41.64
    - Increase in VAT Revenue (50 percent efficiency): 20.82
    - 2001 Revenue from Business Tax, Excluding Finance and Insurance: 71.58
    - Gain from Reform (100 percent efficiency): -29.95
    - Gain from Reform (50 percent efficiency): -50.76
  - Guangdong (Table 5 row):
    - 2001 Value-Added in Newly Taxable Service Sectors: 2657.72
    - 2001 Total Purchases from Currently Exempt Sectors: 276.09
    - Net Increase in VAT Base: 2381.62
    - Increase in VAT Revenue (100 percent efficiency): 101.22
    - Increase in VAT Revenue (50 percent efficiency): 50.61
    - 2001 Revenue from Business Tax, Excluding Finance and Insurance: 263.06
    - Gain from Reform (100 percent efficiency): -161.84
    - Gain from Reform (50 percent efficiency): -212.45
  - Totals (Table 5):
    - Aggregate increase in VAT revenue (100 percent efficiency): 881.38 (hundreds of millions of yuan)
    - Aggregate increase in VAT revenue (50 percent efficiency): 440.69 (hundreds of millions of yuan)
    - Aggregate calculated loss of business tax revenue: 1324.19 (hundreds of millions of yuan)
    - Aggregate net gain from reform (100 percent efficiency): -442.81 (hundreds of millions of yuan)
    - Aggregate net gain from reform (50 percent efficiency): -883.50 (hundreds of millions of yuan)

### Central government gains and redistribution capacity
- Central government receives 75 percent of VAT revenue.
- Center’s additional revenue from VAT services accruing to provinces (three times provincial accrual): three times between Y 44 billion and Y 88 billion, or Y 132 billion and Y 264 billion.
- Implication: the center can redistribute these additional sums to provinces for equalization, or even increase the revenue share in favour of local governments.

### Revised sharing-rule (revenue-neutral for provinces) scenario
- Objective: increase the VAT sharing rate from 25 percent to make the reform revenue-neutral for provinces in the aggregate.
- Calculated new sharing rate:
  - Approximately 0.30 if the VAT is collected very efficiently (100 percent).
  - Approximately 0.37 if the VAT is collected less efficiently (50 percent).
- Results summarized in Table 6 (Extension of VAT to Services, revenue-neutral sharing rule; amounts in hundreds of millions of yuan):
  - Heilongjiang (Table 6 row):
    - Current Revenue from VAT and Business Tax: 88.86
    - Post-Reform Revenue from VAT and Business Tax (100 percent efficiency): 96.39
    - Post-Reform Revenue from VAT and Business Tax (50 percent efficiency): 100.06
    - Gain from Reform (100 percent efficiency): 7.52
    - Gain from Reform (50 percent efficiency): 11.19
    - Gain from Reform (100 percent efficiency), as Percent of Initial VAT and BT Revenue: 8.46
    - Gain from Reform (50 percent efficiency), as Percent of Initial VAT and BT Revenue: 12.60
  - Beijing (Table 6 row):
    - Current Revenue from VAT and Business Tax: 240.34
    - Post-Reform Revenue (100 percent efficiency): 247.82
    - Post-Reform Revenue (50 percent efficiency): 251.53
    - Gain (100 percent): 7.48
    - Gain (50 percent): 11.18
    - Gain percent (100 percent): 3.11
    - Gain percent (50 percent): 4.65
  - Guangdong (Table 6 row):
    - Current Revenue from VAT and Business Tax: 490.64
    - Post-Reform Revenue (100 percent efficiency): 383.32
    - Post-Reform Revenue (50 percent efficiency): 388.77
    - Gain from Reform (100 percent efficiency): -107.32
    - Gain from Reform (50 percent efficiency): -101.87
    - Gain percent (100 percent): -21.87
    - Gain percent (50 percent): -20.76
  - Totals (Table 6):
    - Current Revenue from VAT and Business Tax (national total): 3,190.76 (hundreds of millions of yuan)
    - Post-Reform Revenue under revenue-neutral sharing rule (100 percent efficiency): 3,190.76 (hundreds of millions of yuan)
    - Post-Reform Revenue under revenue-neutral sharing rule (50 percent efficiency): 3,190.76 (hundreds of millions of yuan)
    - Aggregate national gain from reform under revenue-neutral sharing rule: 0.00 (hundreds of millions of yuan) for both efficiency assumptions

- Observation: even with a revenue-neutral increase in the general sharing rule, it is not clear that benefits or losses are targeted toward poorer provinces; an increased general sharing rate would not appropriately target provinces facing revenue shortfalls.

### The revenue-returned (RR) system and distributional properties
- Mechanism: the revenue-returned formula returns an additional portion of VAT and consumption tax revenues to provinces.
- Rule: 30 percent of the increase in VAT and consumption tax collection over their 1994 base is returned to the originating province.
- Distributional issues:
  - The RR mechanism does not operate uniformly across provinces; the amount returned for an identical incremental revenue gain differs across provinces.
  - Differences in treatment across provinces appear largely arbitrary rather than targeted toward poorer provinces or those with larger expenditure–revenue gaps.
- Revenue-returned formula (as presented in text):
  - ti_RR,i = [expression in the source; formula shown in full in source text where ti_RR,i depends on ti_VAT and ti_CT and previous-period returns]

*Sources: China Statistical Yearbook; and authors’ calculations.*

### 0.3 times the growth rate of the sum of VAT and consumption tax revenues for that province.

### _wp04125 - 0.3 times the growth rate of the sum of VAT and consumption tax revenues for that province.

### Revenue-returned formula and mechanics
- The revenue-returned (RR) formula returns to province i in year t an amount that tracks actual returned revenue well. The formula implies that an increment ∆ in either VAT or consumption tax revenue in year t will change revenue-returned to province i in that year according to equation (2):
  - ∆RRti = ∆CTVATti + 0.3 × (∆CTVATti − RRti−1 − CTVATti−1)  (presented in the source as (2))
- The share of any increase in VAT or consumption tax revenue in year t returned to the province via the revenue-returned formula (the incremental tax share) is given by equation (3):
  - incremental tax share = 0.3 × RRti−1 / (CTVATti−1 + consumption_taxti−1)  (presented in the source as (3))
- Implications from (3):
  - A province’s incremental tax share will be higher (i) the higher last year’s revenue-returned; (ii) the lower last year’s VAT and consumption taxes collected.
- When the VAT share of 25 percent is included, a province receives 25 percent plus the incremental tax share.

### Empirical evidence and historical returned revenue (Table 7 highlights)
- Returned-Revenue, 1996–2001 (In billions of yuan):
  - Excises: 1996 = 62.0; 1997 = 67.4; 1998 = 81.5; 1999 = 82.0; 2000 = 85.729; 2001 = 92.946
  - VAT: 1996 = 296.3; 1997 = 327.9; 1998 = 362.8; 1999 = 387.8; 2000 = 455.251; 2001 = 535.622
  - Total collection: 1996 = 358.3; 1997 = 395.3; 1998 = 444.3; 1999 = 469.9; 2000 = 541.0; 2001 = 628.6
  - RR actual: 1996 = 194.9; 1997 = 201.2; 1998 = 208.3; 1999 = 216.73; 2000 = 228.2; 2001 = 234
  - RR formula (selected years shown in source): 2000 = 226.6; 2001 = 239.3 (presented in table)
- RR as percent of total collection (selected values shown):
  - RR actual (selected): 1996 = 50.9; 1997 = 46.9; 1998 = 46.1; 1999 = 42.2; 2000 = 37.2
  - RR formula (selected): 1996 = 50.8; 1997 = 47.0; 1998 = 45.1; 1999 = 41.9; 2000 = 38.1
- Interpretation:
  - The RR formula grows one third as fast as VAT and consumption tax revenue, so revenue-returned as a proportion of tax revenue will tend to zero in the long run, but decline can be slow depending on tax growth.

### Incremental tax shares across provinces (Table 8 key figures)
- Methodological notes:
  - Provincial revenues from VAT available for 2000; provincial distribution of aggregate revenue-returned based on 1998 data scaled up by growth in actual revenue-returned; consumption tax provincial distribution inferred from VAT collection.
- Totals and average:
  - Totals in Table 8: Revenue-returned, 2000 = 228.20 (billions of yuan); Total VAT Collected, 2000 = 455.99 (billions of yuan); Total Consumption Tax Collected, 2000 = 85.70 (billions of yuan)
  - Average incremental tax share (2001) = 14.44 (percent)
- Examples of incremental tax shares (Incremental Share, 2001 in percent):
  - Heilongjiang = 8.70
  - Jiangsu = 9.97
  - Zhejiang = 9.07
  - Guangdong = 9.59
  - Shanghai = 12.47
  - Beijing = 11.50
  - Hunan = 20.53
  - Guizhou = 20.74
  - Gansu = 22.54
  - Yunnan = 28.51
- Summary finding:
  - Considerable variation: incremental tax share of about 8 percent in Heilongjiang, about 28 percent in Yunnan.

### Speed of decline in importance of revenue-returned (Table 9 “half-life”)
- Table 9: Number of Years for Revenue-returned as a Proportion of Tax Revenue to Fall by Half
  - Growth Rate of Tax Revenues, g = 0.05 → Number of Years = 20.45
  - g = 0.10 → Number of Years = 10.54
  - g = 0.15 → Number of Years = 7.24
- Interpretation:
  - If all tax revenues grow at 10 percent a year, it would take over a decade for revenue-returned to become half as important for tax revenue as it was in 2001. If revenue growth slows, the RR would take decades to fall by half.

### Inclusion of RR in C-VAT reform: scenarios and quantified impacts
- Two scenarios when C-VAT reform reduces VAT base:
  1. No change in revenue-returned: If reform occurs in a period of weaker economic growth so VAT + consumption tax show negative growth after reform, the “overall increase” constraint binds and RR is not reduced.
  2. Change in revenue-returned: If reform occurs in strong growth so VAT + consumption tax still show positive growth despite reform, the “overall increase” constraint does not bind and RR falls.
- Quantitative impact when RR falls (summary from Table 10 and discussion):
  - When applying 25 percent plus the incremental tax shares of Table 8 to calculate VAT loss, and assuming 50 percent collection efficiency, on average percentage VAT revenue losses increase by about 5 percentage points:
    - Average loss rises from 15.40 percent to 20.07 percent (presented in source).
  - Table 10 totals and averages (In hundreds of millions of yuan):
    - Totals (loss columns): 31.48 and 134.17 (as presented)
    - Averages: 24.34 (presented)
- Distributional effect:
  - Including RR produces a more pronounced inverted “U” shape relationship between percentage losses and per capita income of provinces; extending analysis to include RR would reinforce earlier conclusions that losses are concentrated in industrial provinces and that distributional impacts are nonuniform.

### Key conclusions and policy options
- Main conclusions:
  - Reforming indirect taxation (move to a C-VAT excluding capital goods, and extending VAT to services) would generate considerable revenue losses for local governments and also generate losses for the central government.
  - Extending VAT to services would yield substantial revenue gains to the center but would imply elimination of the business tax on most services—an important local revenue source—making reform politically difficult without compensation.
  - The revenue-returned formula is not transparent and differentiates regions based on past history rather than current revenue capacities or needs; it is only temporarily disequalizing but may take decades to shrink materially in importance.
  - Simple sharing of VAT revenue is transparent and uniform but permanently disequalizing.
- Compensation and reform options discussed:
  - Compensating provinces by changing parameters of the current system (larger share in VAT revenue sharing or increasing the standard VAT rate) would not adequately target provinces with revenue shortfalls and would add to regressivity.
  - A more effective option: extend the equalization grants introduced in 1994, using expenditure needs and revenue capacities to target transfers so that changes in province revenue-raising capacity are picked up in the formula and transfers adjusted accordingly.
    - This redesigned transfer system could achieve more effective redistribution, but political buy-in from richer provinces would be needed.
  - Complementary measures to facilitate acceptance of a more redistributive system:
    - Joint reassessment of expenditure responsibilities together with revamped revenue assignments.
    - More transparent access to capital markets.
  - Tax assignment and revenue instruments:
    - Provide provinces and counties with some control over rates of assigned taxes at the margin to improve fiscal accountability; choices should avoid excessive distortions and tax competition.
    - Strong arguments for standardizing the base for the personal income tax, while allowing bounded control (piggybacking) for provinces.
    - Expand use of property taxes at the county level, including enhanced valuation and recording for leasehold properties, and base annual taxes on the annual lease value equivalent.
  - Spending-side adjustments:
    - Sub-national expenditure assignments are not matched to resources; some middle-income counties struggle with pension liabilities and unfunded mandates.
    - Recentralization of pension liabilities and unemployment insurance should be considered.
- Implementation note:
  - The recommended reforms form an extensive, interlocking package that will take several years to implement; scope should be comprehensive though pace can be gradual.

*Sources: Ministry of finance; China Statistical Yearbook; and authors’ calculations as presented in the source document.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04125.pdf_