Page 11 - Working Paper (Measuring BEPS and Its Countermeasures in Indonesia: A Preliminary Research Guide)

P. 11

```
DDTC Working Paper 1717
11
especially the first three, represent the pool of the Further economic and statistical assumptions
whole financial transaction or financial investment could be made to enable more advance analysis to
flow in which artificial profit shifting take part. The generate more idea about the magnitude of BEPS.
last one – CIT revenue – measures the magnitude Nevertheless, it does not increase the reliability of
of BEPS from the fiscal impact it has produced. In the estimation accuracy.
short, macro approach treats every BEPS activities
are identical, since what matters is to get insight of It gets even more difficult when we go from
the total estimation. measuring BEPS in a group of countries to do
similar purpose in one individual country. The
In most cases, the key variable that are mostly reason is basically two-fold. First, at individual
used as the multiplier of BEPS activities is CIT rate country level, the number of observations are
difference between countries. The reason is that it drastically reduced, which brings both technical
is more directly related to how much tax burden statistic limitation and lower reliability in the
is reduced due to the practices. It also represents result. Second, the nature of the data is changed
mathematical reasoning from MNEs in making from panel data – which comprised of many
business decision regarding profit maximization countries data across a series of time – into time-
choices of action, including the consideration of series data. This change brings us particular
cost for shifting profits (e.g. paying consultant, complication regarding the determination of which
more costly tax division, probability of getting kind of regression to be used.
32
punished). However, the consequence is that it
Econometrically speaking, time-series data
does not distinguish CIT rate difference between
regression requires two stages of important
two countries and between a country with tax
examination before any result could be generated.
haven jurisdiction. This implication is significant,
First, we should test whether the data for each
since BEPS behavior could be different if the
variable is stationary or not. There is a probability
jurisdiction destination is a tax haven.
that stationary data from involved variables
Alternative measurement using FDI could could generate a strong significant relation with
2
be possible without using CIT rate difference, high R . This kind of result would lead us to false
35
although FDI flow is significantly influenced by conclusion, since such correlation shown does not
33
taxation. This method was used by UNCTAD represent any true relationship between variable.
(2015), where identifying suspicious movement is When two unrelated variables have certain
the key here to determine whether a flow can be similar tendency of non-stationary movement,
categorized as BEPS substance or not. There might spurious correlation would appear, while the true
36
be potential to develop this method to measure the correlation between the two is still unknown.
34
BEPS magnitude. The finding firstly identifies the Conclusively, seemingly correlated non-stationary
existence of BEPS by analyzing certain FDI outflow data would lead to a misled interpretation. In
and inflow that are concentrated to tax haven handling such situation, taking first differentiation
countries. Then the research continues by finding – or second differentiation – of each variable is
the correlation between magnitude of utilization of the logical subsequent step. This way, we would
offshore investment hubs for FDI and the level of get stationary form of each variable and we can
rate of return using OLS model. This method could continue to proceed the regression process.
be promising, but still has limitation in bringing
this tool from group of countries level – which uses Second, having the data stationary, we
panel data – to individual-country level – which proceed into selecting which regression method
uses time-series data. to be used. The possible methods include vector
auto regression (VAR), vector error correction
Difficulty arises when we try to differentiate model (VECM), and ordinary least square (OLS).
between part that is real investment flow or real Selecting one of them is quite complicated, since
business transaction and part that is profit shifting each existing regression method has their own
activities. It is plausible, since macro data does boundaries and they are unable to provide ideal
not provide the underlying motivation behind measurement. It also depends on the nature of the
the financial decisions represented in the data. data. For instance, if we are to use VECM, we have
to examine first whether there is co-integration
relationship between related variables.
32. Clemens Fuest, Shafik Hebous, dan Nadine Riedel. “International
Profit Shifting and Multinational Firms in Developing Countries”,
International Growth Centre Working Paper (2011): 5. 35. See Simon P. Burke dan John Hunter, Modelling Non-Stationary
33. OECD, Tax Effects on Foreign Direct Investment: Revent Evidence Time Series: A Multivariate Approach (New York: Palgrave Macmillan,
and Policy Analysis (Paris: OECD Publishing, 2007). 2005), 8-37.
34. UNCTAD, “An FDI-driven approach to measuring the scale and 36. Helmut Lutkepohl, “Univariate Time Series Analysis”, in Applied
economic impact of BEPS”, Technical background paper accompanying Time Series Econometerics, Helmut Lutkepohl and Markus Kratzig, eds.
the UNCTAD Working Paper on “FDI, Tax, and Development” (2015). (Cambridge: Cambridge University Press, 2004): 11.
```