Transcript Why does China save so much?

The Incentive to Declare Taxes and Tax Revenue:
The Lottery Receipt Experiment in China
Junmin Wan
Fukuoka University
wan＠econ.fukuoka-u.ac.jp
Preparation for the 64th Congress of the IIPF,
the University of Maastricht, the Netherlands,
Aug. 22, 2008
1
Purpose, Findings of Our Paper
• To examine `Lottery Receipt System’ theoretically and
empirically.
Lottery Receipt: public lottery printed on official receipt
• Theoretically, this new system will work well.
• By panel data: the lottery receipt system has raised the
business tax revenue, the real growths of both business tax
and total tax revenues.
• By individual “panel”’ data: the new system promoted
consumer to declare tax.
2
Figure 1: Budget Deficit, Public Income and Tax
Revenues in China, 1951-2004
Ratio of budget
deficit to public
expenditure
0.4
0.3
0.2
Ratio of public
income to GDP
0.1
0
-0.1
2003
1999
1995
1991
1987
1983
1979
1975
1971
1967
1963
1959
1955
1951
-0.2
Ratio of
business tax to
total tax
revenue
Source: China Statistics Yearbook, 1991-2005
3
Reasons for Being Interested in
Taxation in China
• High economic growth, budget deficit in China has been serious (Fig. 1).
• Gini coefficient: 0.39, 0.40, 0.47 in 1995, 2000, 2004, respectively,
the degree of economic inequality is growing.
• Also in other countries, the taxpayer has an incentive to underreport the
amount of taxes due, owing to the asymmetry of information between
government and taxpayer.
• Underground economy in China.
e.g. 14.4% of GDP in 2004 was undervalued
• Mainland China has a new system, ``Lottery Receipt Experiment (LRE).’’
Taiwan and Korea also have this type of system.
4
Lottery Receipt Experiment
(LRE)
• Nationwide, 7.98%, in 2003.
• Beijing: 2002-present.
• Tianjin: 2004-present.
• Reported prize in Beijing:
1,667,700 Yuan in 2002, 11,170,000 Yuan in 2003.
5
Literature on Tax Evasion
• Allingham and Sandmo (1972, J. Pub. E)
underreport, punishment
• Cowell (1990, The MIT Press)
cheat the government
• Cremer and Gahvari (1993, J. Pub. E)
commodity tax, punishment, tax rates
• Fisman and Wei (2004, JPE)
“missing imports” in mainland China
6
Fig.2 Lottery Receipt System
consumers have chance to get prize from government
lottery receipt
government
goods and lottery receipt
firm
tax based on the lottery receipt
consumer
money (transaction)
7
The Benchmark Model
(without lottery receipt)
• From Cremer and Gahvari (1993, J. Pub. E)
• Many homogenous firms, completely
competition, one good,
linear production function, marginal cost =c
• N consumers
• Consumer share the information on transaction
with the firm without any cost
8
Firm’s Expected Profit
•
•
•
•
•
•
•
α: proportion of sales reported by firm, [0, 1]
β: auditing probability, [0, 1]
τ: punishment, [0, ∞)
x: output
p: selling price
t: rate of sales tax
g: cheating cost, [0, ∞)
9
Expected Tax Rate
• Cheating firms’ price: p＝c+g+te,
g: cost for cheating
te : tax rate based on evasion
• Honest firms’ price: ph＝c+t
• p<ph
or
p >ph ?
10
Honest Firms Excluded
• Lemma 1.
p<ph (g+te <t), the tax evasion problem.
• The honest firms are excluded by
cheating firms, and there are only tax
evading firms in the equilibrium.
11
Optimal Policy Under Evasion
• Government’s Problem
•
where,
• ``Social optimal’’
t*>0, β*>0, τ*>0, given that 1 >α*>0
12
Subsidy Mechanism
(e.g. LRE)
• Keep the auditing (β*＞0)
• Give the consumers a subsidy (s, e.g. lottery)
to make them declare tax (e.g. ask for an
official receipt)
13
Pareto-improving
• The consumer’s incentive to declare tax:
ps=c+t*-s ≤ p =c+te*+g(α*)
→ s ≥ t*- te* -g(α*)
• The government’s incentive to introduce the subsidy:
→ s ≤ t*-te*
• Proposition 1. If t*-te* ≥ s ≥ t*- te* -g(α*), this
economy will be Pareto-improving.
• The net increase of government revenue: g(α*)
14
Cheating Firms Excluded,
Pareto-efficient Policy
• Proposition2.
The government can find a s** to enable the
practice of the Pareto-efficient taxation (t**)
without evasion.
• The cheating firms will be excluded.
• The cheating cost g(α*) will be totally saved, and
the monitor cost d(β*)will be partially saved.
15
Consumers’ Reporting Cost (ζ)
• Proposition 3.
If g(α*)- ζ>0, t*-te* ≥ s ≥ t*- te* -g(α*) +ζ,
this economy will be Pareto-improving.
• To decrease the ζ Pareto-improves the
economy.
16
Evidence from Macro Panel Data
• Panel estimation based on 37 districts in Beijing and Tianjin,
1998-2003
• Data Sources:
Business tax revenue, Total tax revenues, CPI, Population, etc.
Beijing Statistics Yearbook 1999-2004;
Beijing Public Finance Statistics Yearbook 2002-2004;
Tianjin Statistics Yearbook 1999-2004;
China Statistics Yearbook 1999-2004.
• Information on LRE and prize:
Beijing Local Taxation Bureau;
Tianjin Local Taxation Bureau;
China Taxation Bureau; Newspapers, etc.
17
Treatment and Control Group
18
Random Trend (Growth) Model
• Heckman and Hotz (1989), Papke (1994)
• yit: level or log value of tax revenue at time t
• Zit: other control variables
• LREit: Lottery Receipt Experiment
（Dummy=after dummy * experiment dummy)
• ci ：specific effect at district i
• θi： trend, or growth trend
• t : time trend
19
Empirical Results from Macro Panel
• Random Trend Model
(effect on the level of tax revenue)
Business tax revenue: + 17.1%
Total tax revenues: +, but not significant
• Random Growth Model
(Effect on the growth of tax revenue)
Growth of business tax revenue: + 21.5%
Growth of total revenues: + 10.04%, but weakly
significant (at 10%)
20
Evidence from Individual “Panel” (1)
• Chinese Household Survey, Feb. 2006
• Osaka University performed the survey
• Shanghai, Beijing, Chengdu, Guangzhou, Shenyang,
Wuhan (LRE in 2005)
• 1,500 households, randomly sampled
• Questions on the consumers’ behavior before and
after the LRE
21
Evidence from Individual ``Panel” (2)
• For the 1,021 respondents who did not ask for
and receive official receipts before the LRE:
384 respondents (384/1,021=37.6%) of
them asked for and received lottery receipts per
2.039 times shopping or eating out after the LRE,
and 331 respondents (331/384=86.2%) answered
that the reason is ``because the lottery was
printed on the receipt.’’
22
Evidence from Individual “Panel” (3)
• For the 479 respondents who did ask for and
receive official receipts before the LRE:
They asked for and received an official
receipt per 2.532 times shopping or eating out,
while after the LRE 450 respondents of them
did ask for and receive official receipts per
1.620 times shopping or eating out, and 332
respondents (332/450=73.7%) answered that
the reason is ``because the lottery was printed
on the receipt.’’
23
Empirical Results by Individual “Panel”
• The frequency of receiving receipt was significantly and
largely increased by the LRE
(p-value = 0.000 by difference tests).
• When a consumer asks for and receives an official receipt, the
government can simultaneously know the transaction between
a buyer and a seller, then easily collect the business (sales) tax.
• For a consumer, the LRE is given, thus the results are reliable
(robust) because we control the “self-selection problem” well.
• It strongly supports that LRE significantly promotes the
consumer to declare tax.
24
Findings and Implications (1)
1) Theoretically, LRE will work well.
2) LRE worked well based on macro panel data.
3) Individual `panel’ data strongly supports that LRE promote
consumers to declare tax.
(a natural experiment, control self-selection)
4) The LRE may contribute to national accounting system
(NSA) by mitigating the issue of underground economy.
25
Findings and Implications (2)
5) To make LRE work better, we should decrease the
consumer’s reporting cost.
6) This system may work well in other countries.
7) China’s government had better go on this new system.
8) Just like LRE, institutional innovations may contribute
to explain “China Miracle.”
26
For Future Research
• To clarify why the consumer prefers lottery.
• More comprehensive data, nationwide, 2004
－now.
• Micro “panel” data analysis in rural sectors.
(the authors is doing this work with rural
individual “panel”)
27
Thank you very much!
28