Extracts from Attempt to measure cross
border VAT fraud
Breakdown within the EU
OCS SPF Finances Belgique
I. INTRODUCTION
The goal of this work is to attempt to value
the actual level of VAT fraud and its distribution within the
Union. In fact, on the basis, of this same paper produced by the
EC for the Council and the European Parliament, the Member States
claim to be very concerned about the size and the recent trend
of this phenomenon.
The majority of valuations carried out before
now give more weight to the macroeconomic loss over the microeconomic.
The results seem variable and not very convincing. Certain countries
have attempted them; others not. What's more, the calculations
done at the European level do not take national disparities into
account.
II. DEFINITIONS
When we speak of VAT fraud, what does that mean?
First, a big distinction must be made between classic and organised
VAT fraud. In its report on the "State of VAT fraud between
the years 2000 to 2003."[1],
the OCS[2]
precisely identified eight types of fraud linked to organised
VAT fraud. These modus operandi are missing trader fraud,
crossed billing, false intra-company deliveries, false exports,
abuses of the margin framework, false invoices, in and outers,
and buffer companies.
The most important in terms of damage is cross
border fraud (MTIC or carousel fraud with the help of defaulting
operators)...
Classic fraud is more diffuse: it means a loss
of earnings for the state, from reasons as diverse as the underground
economy, tax evasion, abuse of deductions, "intentional"
errors, non-payment,...
Carousel fraud is not new. A trace of it can
be seen in the mini-market unique to Benelux in the beginning
of the 1980s but the real level of the amounts involved and their
number are making for some worrying files at the moment. Carousel-type
VAT fraud is notable in that it is closer to swindling than to
true tax fraud. What's more, the level of this fraud is theoretically
limitless. This point will be explored further in Section 2.
The breakdown between classic fraud and organized
fraud is a big unknown without any reliable estimates as to their
sizes. The OCS estimated organized VAT fraud in Belgium for the
year 2001 at 1.1 billion Euros and for the same year, classic
VAT fraud at 2 billion Euros[3]
which gives us a 1/3-2/3 ratio in 2001; organized fraud decreased
to 159 million Euros in 2005. The ratio was thus 1/13-12/13. The
proportion between the two amounts can therefore vary enormously.
One will have understood that organized VAT
fraud and in particular carousel fraud, occupy a unique position
at the heart of VAT fraud.
Between 2001 and 2005, Belgium successfully
implemented a policy against carousel-type fraud. The significant
volume of carousel cases is such that a coordinated action against
the players has had the effect of shattering the breakdown between
classic and organised fraud. It is now clear that in certain cases
organised fraud plays an important role in the total structure
of fraud and that the situation can vary enormously in a short
space of time. The control of one or the other phenomenon is going
to change the balance of power and have an immense impact in terms
of loss. The situation is very different for classic fraud.
The goal of this work is indeed to try to measure
the impact of organised fraud within total VAT fraud and more
precisely on MTIC intra-Community transactions which are a component
of it, made possible due to the introduction of the transitory
regime in January 1993.
The work is divided into three large sections,
the first will encompass the definitions and descriptions necessary
for a good understanding of the operating modes of MTIC fraud.
The following section will dwell on the different approaches possible
to tackle the study and estimate of VAT losses. The third and
last section will utilise the data contained in the EUROCANET
network to evaluate the level and distribution of MTIC fraud within
the European Union. The conclusion will draw up a summary table
of the existing valuations.
We can see the difference between a classic
fiscal fraud and organized VAT fraud whose mechanism is like swindling
through an abuse of the system. In summary, a country will potentially
be at greater risk the bigger its market and the larger its VAT
refunds. It will have a low risk if it has a small domestic market
and low levels of VAT refunds. We will see in the chapter on estimates
where each of the EU countries fit within this profile.
III. ESTIMATES
III.1 The macroeconomic approach
The macroeconomic approach is the one most often
found in the literature. The main reason for this is easier access
to the source data for the estimates. However, they (the estimates)
have disadvantages in terms of accuracy and timing.
GDP
The size of the country and the size of the
domestic markets were able to have an influence on the risk of
VAT fraud. We show now the GDP of the 25 member countries of the
EU.
Five countries alone account for 3/4 of the
EU's GDP: Germany, the UK, France, Italy and Spain. On first glance,
these countries, each with large domestic markets, should be the
most exposed to the risk of type 1 fraud, that is to say, the
case where fraudsters sell their goods on national markets and
receive the VAT from their customers. At the bottom of the list,
eleven countries comprise just 3 per cent of the total wealth.
This cluster of small countries has less risk given the narrowness
of their markets and is less attractive to criminal organizations.
We will clearly see that big disparities exist between Member
States.
Table of 25 Member countries' GDP for the year
2004, and cumulative for the year 2004, also in percentages:
|
| Country | GDP 2004
| Cumulative GDP 2004 | % EU
| Cumulative % |
|
| Germany | 2,740,551,000,000
| 2,740,551,000,000 | 21.4
| 21.1 |
| United Kingdom | 2,124,385,000,000
| 4,864,936,000,000 | 16.6
| 36.0 |
| France | 2,046,646,000,000
| 6,911,582,000,000 | 16.0
| 54.0 |
| Italy | 1,677,834,000,000
| 6,589,416,000,000 | 13.1
| 67.1 |
| Spain | 1,039,927,000,000
| 9,629,343,000,000 | 6.1
| 75.2 |
| Netherlands | 576,979,400,000
| 10,208,322,400,000 | 4.5
| 79.7 |
| Belgium | 352,311,900,000
| 10,560,634,300,000 | 2.8
| 82.5 |
| Sweden | 346,412,400,000
| 10,907,048,700,000 | 2.7
| 85.2 |
| Austria | 292,327,800,000
| 11,199,374,500,000 | 2.3
| 87.5 |
| Poland | 242,292,600,000
| 11,441,667,100,000 | 1.9
| 89.4 |
| Denmark | 241,436,600,000
| 11,683,103,700,000 | 1.9
| 91.2 |
| Greece | 205,215,400,000
| 11,888,319,100,000 | 1.6
| 92.8 |
| Finland | 185,922,500,000
| 12,074,241,600,000 | 1.5
| 94.3 |
| Ireland | 181,622,800,000
| 12,255,864,400,000 | 1.4
| 95.7 |
| Portugal | 167,716,300,000
| 12,423,580,700,000 | 1.3
| 97.0 |
| Czech Republic | 107,014,900,000
| 12,530,595,600,000 | 0.8
| 97.9 |
| Hungary | 100,685,200,000
| 12,631,280,800,000 | 0.8
| 98.6 |
| Slovak Republic | 41,093,970,000
| 12,672,374,770,000 | 0.3
| 99.0 |
| Slovenia | 32,181,750,000
| 12,704,556,520,000 | 0.3
| 99.2 |
| Luxembourg | 31,864,290,000
| 12,736,420,810,000 | 0.2
| 99.5 |
| Lithuania | 22,262,690,000
| 12,758,683,500,000 | 0.2
| 99.6 |
| Cyprus | 15,418,350,000
| 12,774,101,850,000 | 0.1
| 99.8 |
| Latvia | 13,571,250,000
| 12,787,673,100,000 | 0.1
| 99.9 |
| Estonia | 11,238,550,000
| 12,798,911,650,000 | 0.1
| 100 |
| Malta | 5,319,674,000
| 12,804,231,324,000 | 0.0
| 100 |
|
| Source: World Bank |
We will later see, with the help of other indices, if the
large countries are in fact the most affected by fraud.
The "technology" coefficient summarized in the
table below shows the importance of the electronic market in each
country. It is calculated on the basis of mobile telephone and
PC purchases by head of population. The same trends emerge: a
big difference between small and large countries. Given that we
know fraudsters have a predilection for these business sectors,
this supports the hypothesis of a larger risk in these countries.
|
| Country | Technology
Coefficient
|
|
| Germany | 119.2
|
| United Kingdom | 97
|
| Italy | 89.7
|
| France | 77.5
|
| Spain | 52.3
|
| Poland | 36.6
|
| Netherlands | 29.6
|
| Sweden | 15.3
|
| Czech Republic | 14
|
| Belgium | 13.1
|
| Austria | 12.9
|
| Portugal | 12.8
|
| Greece | 11
|
| Hungary | 10.8
|
| Denmark | 9
|
| Finland | 7.7
|
| Ireland | 6.2
|
| Slovak Republic | 6.1
|
| Lithuania | 4.9
|
| Slovenia | 2.5
|
| Latvia | 2.4
|
| Estonia | 2.1
|
| Luxembourg | 0.9
|
| Malta | 0.5
|
|
| Source: WTI | |
VAT REFUNDS
If the size of the country has a large influence on the amount
of fraud, the level of VAT refunds also obviously increases this
risk, in particular for MTIC fraud. Let's look at the following
table which shows the level of these refunds:
|
| Country | VAT refunds
(Euro billon)
|
|
| United Kingdom | 75.93
|
| France | 34.19
|
| Spain | 19.8
|
| Sweden | 17.3
|
| Poland | 10.2672
|
| Belgium | 7.5
|
| Portugal | 3.3
|
| Ireland | 2.92
|
| Slovak Republic | 2.4648
|
| Greece | 1.24
|
| Luxembourg | 0.84
|
| Lithuania | 0.73
|
|
| Source: Annual reports of national fiscal authorities
|
The United Kingdom is way ahead of the other Member States,
followed by France and Spain. More astonishing is Sweden just
behind, hotly pursued by Poland. However, in order to take into
account the "exporter" nature of the country, the refunds
must be corrected by a coefficient for this factor. In fact, refunds
are very strongly correlated to whether or not (legitimate) exports
are significant in size.
Up until now, we have looked at refunds in terms of their
nominal value. If we compare them in proportion to net VAT receipts,
the results differ a little. The United Kingdom is still ahead
with a rate of almost 55 per cent followed by Spain. By contrast,
the Slovak Republic, Poland and Sweden have an equally high rate.
By grouping together the two variables (refunds and electronics
market) into the same coefficient, we have a clear picture of
the risk for each country.
|
| Country | K-risk
|
|
| United Kingdom | 172.93
|
| France | 111.69
|
| Spain | 71.90
|
| Sweden | 32.60
|
| Poland | 46.87
|
| Belgium | 20.60
|
| Portugal | 16.10
|
| Ireland | 9.12
|
| Slovak Republic | 8.58
|
| Greece | 12.24
|
| Luxembourg | 1.74
|
| Lithuania | 5.63
|
|
Source: VAT refunds: Annual reports from the
national fiscal authorities Electronics market: WTI
|
THE LEVEL
OF INTRA-COMMUNITY
DELIVERIES
The total value of tax exempt goods circulating in the EU
is currently growing by 1,800 billion Euros each year, representing
some 210 billion Euros of VAT. A good number of these exemptions
are justified, but by assuming 10 per cent are fraudulent transactions,
we get 21 billion Euros of potential MTIC fraud.
The Belgian situation is as follows: Belgium acquired 20
billion Euros of tax exemptions in 2004 for 150 billion Euros
of goods in other Member States. That is to say, the loss for
2001, before the action plan against carousels, has been, estimated
at 1.1 billion Euros, equivalent to 5.1 per cent of the intra-community
purchases made by the country. This coefficient of 5.1 per cent
applied to the intra-community purchases across the Union would
give a loss of 10.7 billion Euros for Europe.
CALCULATION OF
THE VAT GAP
(BRITISH MODEL)
Since 2002, the British customs administration has developed
a valuation method for VAT losses, which relies on the calculation
of a VAT gap which attempts to measure the revenue loss to the
Treasury.
The calculation takes place in two steps. The first tries
to estimate the theoretical level of receipts if there were no
loss. This level is called the VAT theoretical tax liability (VTTL),
and is calculated on the basis of the following elements:
— The national accounts and more accurately, the
level of expenses subject to tax;
— The estimate of VAT due on these goods;
— The legitimate deduction of certain duties (like,
for example, tobacco).
The second step consists of subtracting from the VTTL the
actual receipts in the fiscal year concerned. The remaining amount
is assumed to be lost VAT for whatever reason (fraud, tax evasion,
error).
III.2 Microeconomic approach
In contrast with the macroeconomic approach, this technique
uses the disaggregated data. Access to the data is more difficult
because only the tax administrations have them to hand, but, to
its advantage, in a rather short time delay. By contrast, this
technique requires greater knowledge of business, domestic economics
and fraud expertise, in particular so that technical problems,
errors and obvious fraud can be excluded. The data used come from
international administrative cooperation efforts such as for example
the VIES system. The most frequently used method is what is known
as mirror flows. Currently another technique is also being used:
profiling. The source data are therefore those from international
databases of VAT declarations.
In the two cases (mirror flows and profiling) we work on
the individual data which are then summed up or extrapolated to
arrive at the desired estimate.
Mirror flows
The mirror flow method is in fact a data cross. By breaking
down the diagram describing the intra-community delivery/purchase
(ICD-ICP), we can better understand this method.
[Diagram between member states 1 and 2 showing the delivery
is the same as the purchase and is reported via VIES quarterly]
On each side of the border, Member State 1 and Member State
2 have concrete information. The matching of this information
will point out fraudulent transactions in Member State 2 (via
missing trader). Border hopping is made possible by cooperative
administrations and more generally by the VIES automated network.
The matching of data furnished by Member State 1 and national
data present (or absent) in VAT declarations provides evidence
of fraud. On one side we learn that MTT carries out some intra-community
purchases in Member State 1 for significant amounts, and at the
same time it is at fault for not filing VAT declarations in its
own country.
Profiling
As for profiling, the technique is the same. The profiles
are detected individually by the models, then aggregated in order
to assess the weight of each fraud profile.
The Belgian method
The development of this organized VAT fraud valuation method
stems from the implementation of automated VAT carousel detection
techniques by a carousel unit of the OCS. The method is simple
and relies on the reasoning explained hereafter.
In summary our technique consists of calculating a ratio
of total fraud to known fraud at any given moment.
A big advantage in terms of organized VAT fraud is due to
the fact a posteriori that the exact amount of fraud can
be determined[4] for several
of the operating modes and particularly in the case of defaulting
operators (MTIC) and crossed billing. What's more, these two modes
alone account for more than 50 per cent of the organized VAT fraud,
and it's thanks to this unique ability to be specific that estimation
is possible. Such an opportunity does not exist for neither classic
VAT fraud nor for direct taxes.
THE COMMISSION
CENSUS
During the second half of the 1990s, the EC carried out a
census of intra-community fraud in the 15 countries of the Union.
The result was a little skinny: 299 cases uncovered for a value
of 500 million Euros. This figure is well below estimates and
opinions. Either the selection criteria were too strict, or the
Member States were a little afraid to state their problems or
the most likely, they couldn't count cases that they had not yet
detected themselves. Whatever the reason, the result was far from
an accurate reflection of the very serious situation at the time.
III.3 EUROCANET
Eurocanet[5] is an
information exchange network based on international administrative
cooperation as stipulated in Rule 1798/2003. The lowest level
statistics used regroup the available data from 2005-06 for a
total amount in excess of 13 billion Euros of deliveries, equivalent
to 2.7 billion Euros of potential VAT fraud. It would appear from
analyses done by the OCS that more than 95 per cent of intra-community
deliveries recorded in the network are fraudulent.
It turns out that one type of fraud falsifies these findings.
The OCS could extract this data corruption which accounts for
a non-negligible share of the transactions: the Dubai[6]
route. This type of fraud is committed with the damage occurring
exclusively to the UK. It involves deliveries for very significant
sums. We also notice that this fraud only shows up in the first
three deciles of the data (that is to say in 30 per cent of gross
transactions).
COMPLETE DISTRIBUTION
OF NETWORK
DATA
By considering as a hypothesis[7]
an amount of 14.8 billion Euros for annual classic carousel fraud
(therefore without the Dubai route), we obtain the following results:
|
| Country | Breakdown
(w/o Dubai)
| 15 billion
Euros for the EU
|
|
| GB | 25.4%
| 3,756,464,827 |
| ES | 17.3 |
2,583,335,071 |
| IT | 15.7 |
2,324,173,825 |
| DE | 13.2 |
1,953,740,782 |
| FR | 10.2 |
1,515,000,000 |
| NL | 4.6 |
680,001,613 |
| DK | 3.0 |
445,428,207 |
| BE | 1.5 |
221,483,000 |
| PT | 2.4 |
361,934,897 |
| LU | 1.4 |
203,865,264 |
| AT | 1.0 |
143,771,179 |
| IE | 0.8 |
120,205,178 |
| CY | 0.7 |
105,901,856 |
| PL | 0.7 |
102,889,729 |
| SE | 0.6 |
90,101,126 |
| EL | 0.4 |
66,485,645 |
| CZ | 0.3 |
45,943,138 |
| SK | 0.3 |
33,641,868 |
| MT | 0.3 |
25,392,354 |
| EE | 0.3 |
22,483,286 |
| OTHERS | 0.1
| 17,757,355 |
| 100.00%
| 14,800,000,000 |
|
| Source: Aggregate data from Eurocanet
|
In the network, the total fraud for the UK breaks down between:
— 30 per cent for classic carousel.
— 70 per cent for Dubai route.
As for damages from the Dubai route, used exclusively with
goods from the electronics sector, we can establish an 8.85 billion[8]
Euro loss to the British Treasury.
The result gives a breakdown which appeared completely valid,
especially if we compare it to the macroeconomic criteria submitted
as evidence in the section devoted to that approach (GDP, refunds...).
Nonetheless, in view of the results, the hypothesis of 14.8
billion Euros must be treated as a "Upper limit" valuation.
IV. CONCLUSION
For the first time in this type of analysis, we have used
microeconomic data for all of the Union. This reinforces the accuracy
of our results on the relative breakdown of MTIC fraud at the
heart of the Union. We can say these results are credible thanks
to the cross checks with national studies to which we had access.
A first observation is imperative: the inequality in the breakdown
at the community level; one country accounts for 50 per cent of
the fraud and five countries suffer 85 per cent of the total.
By contrast, the results in absolute terms (a total of 24 billion
Euros) are much lower than recently cited figures of 50-60 billion.
Even if our study is only treated as an upper limit estimate,
these are important. What is especially worrying is the parallel
between the results of the Eurocanet breakdown and the distribution
of national wealth (See the graph of GDP), even though the network
is only quasi-exclusively reporting fraudulent flows. This would
indicate a strong correlation between potential risk and known
fraud.
We must note as well that the Dubai route itself is a significant
problem which must be treated with the closest attention given
the enormous amounts at play in the UK.
The size of the country as well as repayment policies and
the clampdown measures implemented strongly influence the level
of fraud. As a function of these factors, it is interesting to
note that MTIC fraud can vary significantly from year to year.
Fraud is going to gorge itself on the weaknesses of the system
as soon as it identifies them, thus creating a snowball effect
and an exponential growth in losses to the States.
In the future, the solution for a damage estimate will be
the development of an accurate valuation tool for fraud, one single
model implemented at national level, working on national data,
which will then allow relevant comparisons between States.
August 2006
1
Yannic HULOT, OCS-multidisciplinary unit of SPF Finances Belgique,
"The state of organized VAT fraud-Years 2000 to 2003",
2004. Back
2
The OCS is a multidisciplinary unit specializing in the fight
against organized VAT fraud. Back
3
By extrapolation of fiscal checks. Back
4
After four months for defaulters and after one month for crossed
bills. Back
5
Stands for European Carousel Network. Back
6
By Dubai route, we mean fraud involving a third country in the
billing chain, which in 90 per cent of cases is Dubai, but also
Hong Kong, Switzerland, the USA or other emirate states. Back
7
Arithmetic average of our four results obtained by the macro
approach: 21 (10 per cent intra-community deliveries), 10.7 (5.1
per cent intra-community deliveries), 11.45 (UK method extrapolated
to the EU on the basis of GDP), 16 (mirror flows for the EU). Back
8
15 billion x 27 per cent (electronics share) x 68.6/31.4 = 8.85
billion. Back
|
