What Do We Know About Base
Erosion and Profit Shifting? A
Review of the Empirical Literature
ITPF/AEI Conference: “Economic Effects of Territorial Taxation”
Washington, DC
March 31, 2014
Dhammika Dharmapala
University of Illinois at Urbana-Champaign
Background and Motivation
 BEPS initiative
 G-20 communiqué, OECD Report (2013)
 US reform debate
 The magnitude of BEPS is crucial for formulating
policy responses
 This paper provides a guide to the empirical
evidence on the magnitude of BEPS
 Highlights how these estimates have changed as new
and richer datasets have become available
2
Hines and Rice (1994)
$
Affiliate
(Low-tax)
Income-shifting:
Parent
Suppose that the tax rate falls
(High-tax)
by 1 % point; how much more
income will be reported by this
affiliate?
Assume observed pretax income is the sum of:
Semi-elasticity:
 “True” income
% change in reported
 “Shifted” income
income attributable to a 1 %
→ attribute unexplained income to BEPS
point change in the tax rate
(or tax differential)
3
Hines and Rice (1994)
$
Affiliate
(Low-tax)
Parent
(High-tax)
Income-shifting:
Suppose that the tax rate falls
by 1 % point; how much more
income will be reported by this
affiliate?
Used aggregate country-level
Bureau of Economic Analysis (BEA) data
Magnitude: semi-elasticity ≈ 2.25 (i.e. a 10 % point decrease in the
tax rate (e.g. from 35% to 25%) is associated with a 22.5% increase in
reported income (e.g. from $100,000 to $122,500))
4
New Affiliate-Level Datasets
Confidential Firm-Level Panel Data:
 Firm-level BEA data on affiliates of US MNCs
 MiDi data on affiliates of German MNCs and
German affiliates of non-German MNCs
Unconsolidated Panel Datasets (Bureau van Dijk):
 AMADEUS: European affiliates
 ORBIS: Global
Estimates of the magnitude of BEPS are substantially
smaller using these richer and more detailed datasets

Enable researchers to rule out many potential (nontax) alternative
5
explanations
Huizinga and Laeven (2008)
$
Affiliate
(Low-tax)
Parent
(High-tax)
Use AMADEUS data for 1999
Magnitude: semi-elasticity ≈ 1.3
Income-shifting:
Suppose that the tax rate falls
by 1 % point; how much more
income will be reported by this
affiliate?
i.e. a 10 % point decrease in the tax rate (e.g. from 35% to 25%) is
associated with a 13% increase in reported income (e.g. from $100,000 to
$113,000)
6
“Consensus” Estimate
$
Most recent studies (AMADEUS data for 19992009): semi-elasticity of 0.4; BEPS appears to
Affiliate
be decreasing over time
(Low-tax)
(Lohse and Riedel, 2013)
Income-shifting:
Parent
Suppose that the tax rate falls
(High-tax)
by 1 % point; how much more
income will be reported by this
affiliate?
e.g. AMADEUS data for 1995-2005
“Consensus” estimate: semi-elasticity ≈ 0.8
i.e. a 10 % point decrease in the tax rate (e.g. from 35% to 25%) is
associated with an 8% increase in reported income (e.g. from $100,000 to
$108,000)
7
Other Approaches: Dharmapala and Riedel (2013)
How do earnings shocks to the parent firm
propagate differentially to low-tax and high-tax
affiliates?
$
Parent
(High-tax)
AMADEUS data for 1995-2005
Magnitude: 2% of “unexpected” income is
shifted to low-tax affiliates, relative to high-tax
affiliates
Affiliate
(Low-tax)
Income-shifting:
-Tax-motivated?
- Other motivations?
e.g. internal capital markets,
risk-sharing
Affiliate
(High-tax)
8
Summary
A 10% point decrease in a country’s tax rate (e.g.
35% → 25%) → an increase in reported income of:
Study
%
e.g. from
to
Hines and Rice
(1994)
22.5%
$100,000
$122,500
Huizinga and
Laeven (2008)
13%
$100,000
$113,000
“Consensus”
8%
$100,000
$108,000
Lohse and
Riedel (2013)
4%
$100,000
$104,000
Are these effects “large” or “small”?
9
Location of US MNCs’ FDI, 2011
Total
Net
Assets PPE
Cap.
Exp.
Sales
All
20699 1202
countries
190
5969
1115
% in
Havens
8.8
21.8
42.6
32.2
11.1
Net
Value
Income Added
R&D
Empl.
Comp.
No. of
Empl.
1445
46
536
11,785
14.5
10.1
7.3
4.9
“Ipso facto” argument for BEPS being large
How do we reconcile this with the relatively small estimated
(marginal) magnitude of BEPS?
10
Location of US MNCs’ FDI, 2011
Total
Net
Assets PPE
Cap.
Exp.
Sales
Net
Value
Income Added
R&D
Empl.
Comp.
No. of
Empl.
All
20699 1202
countries
190
5969
1115
1445
46
536
11,785
% in
Havens
32.2
11.1
8.8
21.8
42.6
14.5
10.1
7.3
4.9
% in the
Neth.
8.6
1.6
2.1
3.8
13.4
2.4
3.1
3.2
1.9
Alternatively, this pattern may be attributable to the use of
holding companies based in havens, as the fairly similar pattern
for the Netherlands suggests 11
Conclusion
 It is important that international tax reform and
multilateral initiatives are informed by current
academic research
 An extensive empirical literature seeks to estimate
the existence and magnitude of BEPS using a
variety of different approaches
 The more recent empirical literature:
 Uses new and richer sources of data
 Finds an estimated magnitude of BEPS that is much
smaller than that found in earlier studies
12