You are on page 1of 51

Regulation and State Intervention in Non-Renewable Natural

Resources: the cases of the oil and gas sectors in Argentina


+
.
Enrique Kawamura
Universidad de San Andrs
April 1, 2014
Abstract
This paper develops a theoretical model based on stylized facts regarding regulation and perfor-
mance of the oil and gas sectors in Argentina after the fall of Convertibility. Regarding performance,
facts include an important drop in domestic production and investment, as well as a reversal in the
energy trade balance (from surplus to decit). Regulations include a direct government intervention
in upstream and downstream prices (below international ones), together with strong restrictions
to export. The model uses a self-enforcing equilibrium concept derived from Yared (2010) to endo-
geneize such policy variables such as upstream domestic prices and downstream energy prices. The
model can explain some of the regulation policies observed in the facts, including very low energy
prices and strictly positive subsidies.
1 Introduction
Beyond its well-known comparative advantage in agricultural resources, Argentina is also a relatively
important producer of some non-renewable resources, mainly, oil, gas and some metallic minerals (espe-
cially copper). Indeed, Argentina has been the third largest oil producer in South America in the early
1990s (after Venezuela and Brazil), although with a declining trend after 2000
1
. Argentina became a net

This paper is part of the IDB-funded research project "Fiscal Revenues and Eciency in LAC Non-Renewable Natural
Resource Sectors". I deeply thank the comments that Fernando Navajas gave to the dierent stages of this paper. I also
thank the coordinators, Andy Powell and Osmel Manzano, for their comments. The usual disclaimer applies.
1
For example, by 1996 Argentinas oil production represented a 16.4% of total oil production of South American coun-
tries (and 27.2% when excluding Venezuela from the sample), while ten years later Argentinas oil production represented
only 10.4% of total South-American oil production, according to the site of BP, http://www.bp.com.
1
exporter of crude-oil. Even though the net export of oil also decreased in the last decade, the country
maintains its position in this sector. Similarly, natural gas exploitation also evidenced an important
increase in Argentina. In 1990 Argentinas natural gas production was 23.02 :
3
billions, while in 2004
the production was 52.38 :
3
billions. Argentina was also an important net exporter of gas until 2006.
The two non-renewable resources referred above belong to the group of those related to energy. This
feature may make these two sectors strong candidates to suer distorting interventions by the State
given that energy plays a key role in electoral competitions. For example, although Argentina was an
important net exporter of gas, beginning 2006 this feature reverted to negative foreign trade balances.
The main conjecture of why this happened is linked to a complex regulation system implemented by
the Administration who took power in 2003 in Argentina. Also, during the 2002 crisis the Argentine
government decided to introduce a tax on oil exports, a tax that still exists nowadays. Note that this
year coincides with the starting point of the increase in the international price of crude oil. These
two cases are just examples of the type of interaction between production and investment decisions by
the producers and the government as the major regulation authority. These interventions impact on
production and investment decisions. But, more importantly, such tax and regulation changes are likely
to be endogenous to the changes in the international and macroeconomic environment, changes that
impacts on the policy-maker incentives, which typically leads to time-inconsistency problems regarding
pre-committed tax or other regulations set ex-ante.
This paper presents a brief summary of major stylized facts about, not only production and in-
vestment in the oil and gas sectors in Argentina after the crisis, but also a description on changes in
regulation introduced by the National government during the period 2002-2012. In this period oil pro-
duction decreased about 25%, gas production decreased 15% between 2004 and 2012 and exploration
activity decreased until 2010. Resource (oil and gas) prices as well as energy prices were completely
controlled by the government. Both were kept way below their corresponding international prices. At
the same time, natural gas consumption heavily increased (about 48%). This combination of decline
in production, increase in domestic energy demand and domestic low prices implied an explosion in
the trade decit, especially in natural gas, whose physical volume reached values even higher than the
maximum trade surplus reached in 2004. As Navajas (2006) and other authors state, such regulation can
hardly be reconciled with standard theories regarding dynamic regulations or taxation in non-renewable
resources.
Given such facts, the paper proceeds to construct and develop a stochastic, intertemporal model of
a non-exhaustible resource with price regulations and quantity restrictions imposed by the state, whose
values arise endogenously in equilibrium. Embedded in a well-known model of non-renewable-resource
taxation, such as Deacon (1993), the framework considers explicit (incumbent) politicians who have
2
the capacity to set the values of such policies, but that could be removed from oce by some electoral
process. The equilibrium concept used for this purpose is an adaptation of the self-enforcing equilibrium
developed in the macro political economy papers by Yared (2010) and Acemoglu et al (2010a). This
concept internalizes self-enforcement problems faced by the regulators by requiring a non-deviation
condition by all relevant decision makers in equilibrium. Thus, the sequence of policy variables arising
in such equilibrium is by construction time consistent.
In the model, upstream rms produce a non-renewable resource to be used as the only input for the
production of a nal consumption good called energy, produced by a downstream rm. These rms face
two types of regulations: the prohibition to export as long as domestic demand is not satised (which
occurs in the equilibrium analyzed in this paper) and prices completely set by the government. Energy
is one of the three goods consumed by households, the other two being a numeraire good and the third
an international commodity that is unrelated to energy uses. Households receive an endowment of labor
time (normalized to one) and a stochastic endowment of the international energy-unrelated commodity.
The government is one of many possible (identical) bureaucrats who consume from the unused scal
revenues coming from the taxes on the energy-unrelated commodity paid by households. On the other
hand, the government also sets the price of energy received by the downstream rm, and may potentially
subsidize the downstream rm to cover losses. Also, part of the input necessary to produce energy may
have to be imported by the government in case that upstream rms do not produce (extract) sucient
quantities of the resource to cover the total demand for energy inputs.
The model also includes a political game between consumers and bureaucrats. The main assumption
is the existence of elections at the beginning of each period, where consumers may choose to re-elect
the incumbent government or else to replace it with another bureaucrat. The model assumes that
consumers enjoy some additional utility coming from the replacement of the incumbent (preference for
turnover). Also, any incumbent bureaucrat faces a utility cost of being expelled from power. Thus,
these two preference features shape the bureaucrats incentives to set low prices for energy to allow for
re-election to occur.
The main results of the theoretical model are as follows. The model predicts that domestic energy
prices are way below the international price of the resource as long as the cost of the incumbent govern-
ment to be expelled from power and the voters preference for political turnover are both high enough
(together with a high enough lower bound in the realization of the endowment of the energy-unrelated
commodity). The intuition for this result is simple. Under the last condition tax revenues are high
to get enough consumption for the incumbent in case that his re-election take place. High utility cost
for the incumbent after loosing elections imply lower benets from setting high energy prices (and low
subsidies) rather than low. The high value of preference for turnover make the incentives to reduce the
3
energy price even stronger, to ensure that voters have enough incentives to re-elect the incumbent. Note
that such conditions also ensure strictly positive subsidies paid to the energy producer (downstream
rm), a fact that characterize the natural gas sector in Argentina especially after 2008.
Other results from the theoretical model include conditions under which resource imports occur in
equilibrium. Such conditions are those that imply a very low price-to-marginal-cost relation in the
upstream sector. In turn, conditions for such low domestic price are characterized by a rst-order non-
linear Euler equation. Its (stochastic) terminal condition (given at the time of exhaustion of the resource)
implies that the domestic price in that terminal period is a constant fraction of the international price,
where this fraction reects the share of variable inputs in the resource-extraction technology. This Euler
equation may also be consistent with declining expected domestic prices, although such specic result
needs a much sharper quantitative characterization that is out of the scope of this paper.
It is important to stress that, to our knowledge, this is the rst paper in presenting a dynamic,
endogenous, self-enforcing regulation model in the non-renewable resource sector. Indeed, the now
very extensive
2
literature on taxation and regulation of non-renewable resources do not include such an
endogenous taxation analysis
3
. In particular, such papers do not take into account the endogeneity of
the risk of expropriation. However, for Latin American countries such as Argentina this type of risk
is clearly of a rst-order importance, provided the regular instability of governments and, especially,
public policies in the region. The relevance of this risk has already been pointed out by Garnaut and
Clunies Ross (1975).
However, only much more recently there has been an increase in the analysis of formal models with
expropriation risk within the natural resource sector. An early example is the model by Gaudet, Laserre
and Long (1995). The basic version of this model assumes two periods and no commitment on behalf
of the government, and so the second-period royalty rate is (re)set after the rst periods tax rate has
been set. To my knowledge, this constitutes one of the few models assuming an explicit problem of time
inconsistency. Clearly, their model does not capture the problem of expropriation and other types of
risks associated to a not-fully-benevolent government. The model lls this gap by assuming a politicians
objective function that combines a rent-seeking behavior with a political-power-keeping behavior, the
2
For comprehensive surveys on this literature, see the classic work by Heaps and Hellwell (1985) and the more recent
by Lund (2010).
3
The classical treatment based on Hotellings (1931) model is surveyed by section 5 in Heaps and Helliwell (1985).
Within this tradition we nd Manzanos (2000) extension considering a "quality" dimension in the prot function applied
to the analysis of the Venezuelan market. For the case of Argentina, Vera (2000) presents a Hotelling model to analyze
two types of royalties: a sales-based tax and a long-run sustainable revenue tax system. The extension to stochastic
environments includes Leland (1978), Campbell and Lindner (1985), Ball and Bowers (1983), Lund (1992) and Zhang
(1997). Postali (2007) applies an explicit option pricing technique to analyze the impact of taxation on the petroleum
sector in Brazil.
4
second of which is more consistent with a benevolent uninformed government than the rst type of
behavior.
Another relatd study is Bohn and Deacon (2000). These authors develop and estimates a theoretical
model of asset expropriation. When estimated with an international panel data they nd that ownership
instability indeed reduces the speed of oil exploitation. This type of expropriation risk is also a major
subject in several chapters of the book by Hogan and Sturzenegger (2010)
4
. In the end, this expropriation
risk reects a time-inconsistency problem by a government that may want to extract more rents by
increasing tax rates over the original arrangement after the international price of the resource increases.
This time inconsistency issue is discussed in more detail when presenting the formal model presented
in this paper.
The rest of the paper is organized as follows. Section 2 presents some key facts about the oil and gas
sectors in Argentina after the abandonment of the currency board system in 2002. Section 3 presents the
model with the analysis. Section 4 presents the main results concerning the equilibrium characterization
of the model. Section 5 presents several policy implications coming from the equilibrium analysis of the
model. Finally 6 presents the conclusions with several suggestions for future research.
2 Some facts about the oil and gas sectors in Argentina after
the Convertibility
This section presents what it is understood as major empirical facts characterizing the oil and gas sec-
tors in Argentina for the period following the abandonment of the currency board in 2002. From the
relationship between the government and producers, this period has been characterized by a steady
increase in regulations concerning prices and trade restrictions. This same period also shows an impres-
sive fall (after a recovery in both production and investment in exploration (in the latter case, at least
until 2010). Thus, it is apparent that those stronger regulations seemed to have aected the incentives
to invest and to produce oil and especially natural gas. These and other facts are presented as follows.
1. Fall in production. In the period considered here, the decrease in domestic production of both
oil and natural gas (in this last case, since 2004) has been an increasing public concern. Figures
1 and 2 below document such fall for each of these two energy resources.
4
Mainly: Tomz and Wright (2010), Rigobon (2010) and Schwartz and Trolle (2010). Also in this book, Wernerfelt and
Zeckhauser (2010) discuss several actions that rms can take to avoid expropriation (that may include a slower speed of
production of the natural resource, as Bohn and Deacon (2000) predict). Also, Engel and Fisher (2010) discuss optimal
auction desigm under the threat of expropriation.
5
Figure 1. Oil production in Argentina. 2002-2012. Source: IAPG
Figure 2. Natural gas production in Argentina (2002-2012). Source: IAPG.
Indeed, the cumulative decrease in oil production between 2002 and 2012 is above 25%. The
decrease in gas production between the 2004 peak and 2012 is more than 15%. As Navajas (2006)
and other authors state, such a decrease (together with a sharp increase in domestic gas use, see
gure 5 below) explains a big portion of the energy problems faced in Argentina, especially after
2007.
2. Fall (and recovery?) in exploration investment. Those falls in production shown in fact 1
can also be linked to a decrease in exploration activity. Figures 3 and 4 below illustrates this fact.
6
Figure 3. Number of exploratory wells in Argentina (2005-2012): oil. Source: IAPG.
Figure 4. Number of exploratory wells in Argentina: natural gas (2005-2012). Source: IAPG
These gures strictly show that, between 2005 and 2010 (with some exceptions) the number of
wells dedicated to exploration of new reserves in oil and gas showed a tendency to decrease. This
seems to show a short run reversion in 2011 and 2012, possibly as an eect of the restatization of
the major oil and gas company in Argentina, Yacimientos Petrolferos Fiscales (YPF). However,
during all this period the quantity of wells are well below the technical level to increase production
signicatively. Overall, beyond these short term uctuations, the general level of this variable
suggests a strong lack of incentives to invest in the sector.
7
3. Reversion of international trade balance in gas: from surplus to decit. As a conse-
quence of the decrease in domestic production, and also as a consequence of the increase in the
domestic demand for natural gas, then it was inevitable the reversion of the trade surplus, turning
soon into a decit.
Figure 5. Natural gas domestic consumption in Argentina (2000-2012). Source: Secretary of
Energy.
Almost moving in opposite directions, gure 5 shows a steady increase (only interrupted by the
international nancial crisis in 2008-2009) in natural gas consumption. The increase between
2002 and 2012 is in the order of 48%. As stated above, these gures suggest that the immediate
consequence would be the reversion of the trade surplus in gas into a decit.
8
Figure 6. Trade balance in Argentina: natural gas (1999-2012). Source: IAPG.
The last gure is a key aspect of the domestic natural gas sector in Argentina after the 2002 crisis.
Especially, after 2006, exports decreased to almost zero in 2011 and 2012 (they are not exactly zero
just because of the pre-committed sales to Chile). Imports, on the other hand, steadily increased
since 2008 until 2012. The trade decit in physical volumes is 43% higher than the peak of the
trade surplus in 2004.
4. Government intervention on producer prices (lower than international prices). As
early as in February 2002 the Duhalde administration implemented by decree
5
export taxes on
oil and gas (see fact 6 below). Several other policy decisions also introduced tight restrictions
regarding the capability of selling oil and gas abroad. In a sense, several of these policy decisions
implied a complete departure from the policy implemented by the Menem administration in the
ninenties, which included market deregulation and privatization of YPF
6
. These interventions have
implied a departure of domestic prices for oil and gas from international prices, at least until the
2008 crisis and later.
Figure 7. WTI monthly spot prices and volume-weighted average oil domestic prices for
Argentina (2002-2012). Sources: EIA and Secretary of Energy.
5
Decree 310/02. This decree was later complemented by several resolutions of the Secretary of Energy (Resolutions
337/04 and 394/07) introducing changes in the tax rate scale.
6
For surveys on these 1990s policies see Gadano (2006 and 2010), Roccaro and Fernandez (2005) and Scheimberg
(2007). About the policy decisions after 2002 see Scheimberg (2011).
9
Figure 8. Natural gas prices: imports from Bolivia and average domestic producer prices.
Sources: Hidrocarburos Bolivia and Secretary of Energy.
Figure 7 shows that, after two years with domestic prices higher than the international WTI spot
price reference, then since 2004 the latter was systematically above the domestic one, except for
the international crisis years (which was clearly a transitory event). In the natural gas case the
dierence is even more clear, at least since 2007, the estimated import price from Bolivia
7
is at
least 2.5 (typically, more than 3) times the volume-weighted average domestic price, as shown
in gure 8. This price dierence, combined with a virtual prohibition to export gas (given by
the domestic-market priority rule introduced in 2004) seems to partially explain the decrease in
domestic production and the low level of investment exploration within the period in which YPF
was still private (mainly, until 2010, as seen above). If one accepts that the inclusion in economic
costs some measure of opportunity costs given by import prices, then such a regulation strongly
would lead to very high costs and low prots, as Navajas (2006) and Scheimberg (2007 and 2011)
state, and unlike what other papers such as Kozulj (2005) states
8
. If one accepts that protability
is key to understand incentives to invest then this may help explaining the lack of investment in
recent years. A recent paper by Barril and Navajas (2011) reinforces this view through a model-
based econometric test to study the determinants of the decline in production, using a panel-data
set at the area-level from 2004 through 2009. This exercise explains the rate of growth in natural
7
The data in gure 8 includes quarterly estimates from the site HidrocarburosBolivia.com between 2007 and 2009 and
unit values of gas imports from ocial data released by the Secretary of Energy. The lack of ocial data of gas exports
from Bolivian sources force to provide such estimates (for a discussion, see Aguilar and Valdivia (2011)).
8
Interestingly, this same author writes an article in 2012 arguing the necessity of adjusting taris upwards, neglecting
in part his diagnosis made in 2005. (See Kozulj (2012)).
10
gas production as a function of dierent variables, including the ratio of cumulative production to
resource size, exploration investment eort, and some concession-renegotiation dummies. One of
their main results is that an increase of 1% in the ratio of cumulative production to resource size
implies a 20% decrease in the rate of growth of natural gas. The extension of the concession to the
company Pan American Energy in 2007 by the government of the Chubut province also shows a
signicant increase on the rate of growth of gas for that company. Ceteris paribus this regulation
implies a 39% increase in the growth rate. These and other results induce the authors to suggest
that such a decline can be attributed mostly to "depressed economic incentives" on intra marginal
reserves and depressing also incentives to invest.
5. Natural gas taris (for nal demand) much below the import prices (implying high
levels of subsidies). Since the 2002 crisis, the government kept taris for nal use in the natural
gas markets clearly below dierent measures of opportunity costs, mainly, those based on import
prices. Cont et al (2011) estimated such dierences, which are illustrated in the next gure.
Figure 9. Natural gas: import prices from Bolivia and taris for nal use in the Cuty of Buenos
Aires and its Metropolitan Area. Source: Cont et al (2011).
In gure 9 the blue line shows the upward evolution of the import price of natural gas from Bolivia.
The red and green lines show the evolution of taris for nal users in the city of Buenos Aires and
the metropolitan area (the Greater Buenos Aires area). For the 2008-2010 taris the observation
corresponds to the upper scale of use (more than 1800 cubic meters per year). Even for such
intensive users of gas (some which correspond to the upper decile of the income distribution in
11
this area) the price paid is way below the cost of importing it. This is the origin of the increasing
value in the subsidies that the government had to pay to gas distributors as a compensation for
such price disparity
9
.
6. No major changes in taxation on the upstream of oil and gas. Regarding taxation on
this sector, upstream companies face two separate types of scal costs. The rst corresponds
to the regular taxes applied to all businesses. The main ones are corporate income taxes (with
a at tax rate of 35%), value-added taxes (a at rate of 21%) and a tax on major shareholder
assets. The rst two are the most important regarding impact on scal revenues. However,
until 2006 there has been no dierentiated treatment towards oil and gas companies regarding
their application. In that year the government released the Act 26154 with the intention of
providing investment and production incentives to the sector. Such incentives included a faster
accounting-based capital depreciation for computing the corporate income tax base and also faster
reimbursement of VAT retentions, among others. However, according to tax law experts, this Act
never was eectively applied. The evidence above (especially in gures 1-4) conrms this absence
of any relevant reaction towards this particular tax policy measure. The second source of scal cost
for companies comes from royalties. In the 1990s the Menems Administration released Act 24145
(1992). According to this, the ownership of oil, gas and minerals passed from its original National
domain to the Provinces. The Constitution Reform in 1994 conrmed this reform. This implied
that the Provinces were now the government levels receiving royalty revenues, not the National
government. However, the same legislation conrmed the royalty rate that was in the old regime
(12% over output value at domestic prices) and since then such rate was not modied. Thus, there
has been no relevant reform regarding the scal cost imposed on upstream oil and gas companies
coming from royalty rate changes. As stated in fact 4, with the crisis following the abandonment
of the Convertibility plan in January 2002, the government in that year implemented measures to
increase its revenues. One of them was the introduction of a tax on oil exports. Initially the tax
rate was a at 20% independent of the value of exports. However, in 2004 the government changes
this tax introducing a stepwise tax rate regime, up to a maximum of 45%, as a function of the
international price of a barrel. The increase in the rate was 3 percentage points per each $2 increase
in the price. In 2007 the government set a completely contingent tax rate that implied a price
cap of the barrel of $ 60.9. However, such export tax is currently (at least, since 2010) not very
relevant, given the low volume of oil exports (corresponding to oil types not used for the distilling
9
Actually, Cont et al (2011) estimate the amount of subsidies that nal users received as a consequence of such tari
regulation. For example, they estimate that, in the 2008-2010, the upper decile of users received about US$ 102.5 millions
in implicit subsidies.
12
process to get gasoline and other type of fuels). In sum: this brief summary shows that taxation
legislation changes in oil and gas has not shown to have an important impact on investment and
extraction decisions. This was mainly due to government decisions implying almost null changes
in the scal costs for those companies. Royalty rates did not change. Corporate income tax and
VAT rates remained also unchanged. Export taxes became important but also for a limited period
of years. As stated below, other regulations implied that, especially since 2010, were more relevant
to understand the production, investment and trade balance negative performances.
These ve facts characterize key variables (regulated prices and taris, production, energy consump-
tion and investment) of the oil and gas sectors in Argentina after the abandonment of the Convertibility.
For several authors who studied them in more detail some of those facts constitute a clear unsustain-
ability problem regarding the long run future of this sector. Also, for critics of the administration who
decided such policies these facts show some form of "irrationality", based on those sustainability issues.
The question is whether it is possible to at least qualitatively interpret or explain part of those facts.
The model presented in the next section is an attempt to do so.
3 A model of state intervention for Argentina in the gas (and
oil) markets
This section introduces an intertemporal theoretical model to understand the potential driving forces
behind the type of the state intervention observed in the gas sector and, less strongly, in the oil sector.
The models assumptions mainly reect institutionally relevant features of those two markets partially
collected in the last section.
3.1 The set-up
Consider an economy lasting for an innite number of periods dated as t = 0. 1. ... In each period there
exists three types of agents:
3.1.1 Domestic consumers and downstream rms.
Consumers care for leisure and two goods: energy and an internationally tradable numeraire good
denominated in dollars, called the consumption good from here on. There is a third good not consumed
by agents, an internationally tradable commodity whose units are also denominated in dollars, called
the exported commodity. Each consumer receives two endowments in each period t: labor time and a
13
random amount
t
of the exported commodity. The endowment of time for each consumer is equal to
one unit per period t. The endowment of the commodity follows a stochastic process with a compact
support =
_
.

with 0. For simplicity, assume that the consumer can work as much as she wants
and produce the consumption good using a linear production with a stochastic productivity (equal to
the market wage) given by .
t
. Assume that the support of .
t
is a nite selection of a compact set
= [.. .] with . 0. The price of energy in dollars is :
t
. The model assumes that the consumers
preference is represented by
1
0
_
1

t=0
,
t
_
(.
t
)
1
1
s
1
1
.
+
c
c
t
|
1c
t
c
c
(1 c)
1c
__
(1)
where .
t
is the consumption of energy, |
t
is the consumption of leisure and c
t
the consumed quantity of
the consumption good. Here assume that 1 and 0 < c < 1. Consumers pay a tax on the endowment
of the exported commodity, whose rate is t
t
. Consumers cannot save nor borrow. Thus, the problem
of each consumer in each period t is to choose (.
t
. |
t
. c
t
) to maximize each term in (1) subject to the
following budget constraint:
:
t
.
t
+ c
t
= (1 t
t
)
t
+ .
t
(1 |
t
) (2)
10
.
Energy is traded in a regulated monopoly
11
. The unique downstream rm faces a production
function for .
t
given by
.
t
=
_

1
t
if .
t
_
1
t

1
t
+
n
t
if .
t

1
t
(3)
Here the variable
1
t
denotes the quantity of the input bought to the upstream rm and
n
t
is the
quantity of the input imported from abroad. Given that the input for the downstream rm is the same
10
For reference, standard calculations show that the demand for energy is given by:
z
d
t
=
!
"(1)
t

"
t
Thus, consumers welfare as a function of ; and other exogenous variables is as follows
1
(" 1)
_
!
(1)
t

t
_
"1
+ !

t
+

t
(1
t
)
!
1
t
11
Although historically the downstream market for gasoline and other oil-derived sources of energy presented some
degree of competition (see, e.g., Serebrisky, 2001), after 2003 the perception is that price is not the variable in which rms
compete, given the strong degree of the intervention by the State to regulate the price.
14
resource as that produced by the upstream rms, then its importing price is still j
t
. Thus the minimum
cost function is obviously linear:
C
1
(.
t
; j
t
) =
_
j
t
.
t
if .
t
_
1
t
j
t

1
t
+ j
t
_
.
t

1
t

if .
t

1
t
(4)
This technology allows to infer that the governments upstream-market regulations may aect the down-
stream regulated energy monopoly
12
. The latter produces energy with the technology (3) to satisfy the
demand of consumers derived from the maximization of (1) subject to (2).
At the beginning of each period consumers decide (through an implicit voting mechanism) which
politician in power or replaces her for a dierent one (although with the same preferences). The binary
variable G
t
(:
t
) 0. 1 captures this choice. Consumers derive a periodic utility
c
0 from replacing
a bureaucrat with a new one. This assumption comes directly from Yared (2010) and captures aversion
to political persistence.
3.1.2 The upstream rms
In each period there is a large number of a unique type of (upstream) rms. The basic set-up assumes
that each rm is owned by risk-neutral shareholders. The rm produces a natural, possibly exhaustible
resource, a second commodity whose main use is the production of energy (either within the country or
abroad). The technology of the upstream rm comes is taken directly from Deacon (1993). Let r
t
be
the quantity of the variable input and 1
t
the stock of reserves available at the beginning of period t.
Then the following production function represents the upstream rm technology:

t
1 (r
t
. 1
t
) (5)
with
1 (r. 1) = r
j
1
1j
(6)
where 0 < j < 1. That is, the production function is a constant-return-to-scale Cobb-Douglas. This
parametrization also corresponds to that in Deacon (1993) chosen for the quantitative exercise performed
in that paper. On the other hand, reserves in period t equals the cumulative reserve additions up to
12
One possibly important issue is related to the way integration between downstream and upstream production is
modelled. This may be be important in the case of Argentina. However, the main point that the model answers does
not seem to be strongly related to "transaction costs" that may arise in the relantionship between the upstream and the
downstream. To keep the model as simple as possible, and considering that such transaction costs problems may not be
rst order for this paper, the assumption is to consider two separate sectors for the upstream and the downstream.
15
that date less cumulative production. The term denotes the total factor productivity for each rm.
In this basic version
t
is assumed to be exogenously stochastic. More precisely, each
t
is assumed to
follow a rst-order Markov process with support being a nite subset of a compact set
_
.

, with 0.
Reserves in period t equals the cumulative reserve additions up to that date less cumulative production.
Using Deacons notation, let \
t
be the cumulative drilling and 1
t
be the cumulative output (extracted
resources). Thus
1
t
= (\
t
) 1
t
(7)
where
(\) = (1 exp (\)) (8)
with and strictly positive constants. The cumulative investment is in fact the addition of all
exploration investment decisions made up to any period t. Denote n
t
as the exploration investment
decision made at time t. Associated to this investment there is a drilling cost 1(n
t
) expressed in
dollars. This 1 function is assumed to be equal to
1(n
t
) = n
t
(9)
This means that n
t
not only expresses a physical quantity of investment eort but also a monetary
amount, given the linearity of the cost function 1. Given this investment eort, the law of motion for
\
t
and 1
t
respectively are
\
t+1
= \
t
+ n
t
(10)
1
t+1
= 1
t
+
t
r
j
t
1
1j
t
(11)
The upstream rm sells the quantity produced in potentially two markets. A rst market is that where
the production of this good is delivered to the downstream rm to produce energy. Thus, the variable

1
t
introduced in subsection 3.1.1 denotes the upstream rm production for domestic use. For these
sales, the upstream rm receives a price equal to j
t
. Below it will be assumed that the domestic price j
t
are determined by the government. Thus, the rm chooses
1
t
given the regulated price j
t
. If allowed,
the upstream rm would also sell part of the output abroad, at the international price j
t
. However, this
possibility depends on the regulation imposed by the government (see below) regarding a priority rule
for domestic sales. In each period the upstream rm decides the input quantity r
t
and the investment
amount n
t
that maximizes the expected discounted prots, taking the realization of the TFP shock
t
.
the foreign price of the resource good j
t
. the regulated local price j
t
. the domestic sale restriction
t
and
the price of the
t
input good as given. The model assumes that the vector (j
t
.
t
) follow a rst-order
16
Markov process with nite support being included in the compact [j
min
. j
max
]
_
.

. with j
min
0.
0. In principle, the latter is assumed to be independent of that governing the TFP shock
t
and that
of the endowment shock .
t
. In subsequent extensions this independence assumption may be dropped
to examine its implication concerning the equilibrium.
The upstream rm owner is assumed to be risk neutral with discount factor ,. the same one as
that of consumers. Thus, the upstream rm chooses a plan r
t
. n
t

1
t=0
that maximizes the expected
discounted value of dollar prots until depletion of the reserves. The exact expression of such prots
is presented below after introducing the government policy decisions (since these decisions aect the
explicit form of the prot function).
Finally, as a consequence of the interaction between the upstream rms decisions and the government
decisions (to be presented below), as well as the exogenous variables, there exists an endogenous random
period 1 (whose at least one of the realizations may be innite) in which resource reserves are completely
exhausted. Thus, whenever 1 is nite then in periods 1 + ,
1
)=1j
reserves are equal to 0, i.e., 1
T+)
=
0. and so the upstream rm is shut down after reserves are exhausted.
3.1.3 Government decisions and upstream rm prots
This model assumes the existence of a (national-level) government run by one of many ex-ante identical
politicians or bureaucrats, whose preferences are dened over stochastic sequences of dollars consumed,
q
t
. They are represented by the following expected utility function:
1
0
_
1

t=0
,
t
q
t
_
(12)
where the discount factor , is the same as that of the upstream rm owners and consumers.
The government sets the tax rate on potential foreign sales (which becomes eective only when
exports are positive). Also, the government decides the price of energy :
t
and the price paid to the
upstream producers of the input good for domestic sales, j
t
. The model assumes that the government
sets the restriction that upstream rmss local sales must satisfy the domestic demand from the energy
producer. There is of course an implicit assumption that the government has the power to close down
any upstream rm that intends to violate such rule. Thus, upstream rm can only export the remains
of production after servicing the local demand. If production is below the local demand, then all
production goes to the latter. On the other hand, given that the government directly sets the price
of energy, it subsidizes the energy producer when the latter incurs in losses. Let o
t
denote the dollar
amount of subsidies provided by the government to the energy producer. Finally, the government taxes
17
a fraction t
t
(0. 1) of each consumers wage income in terms of the numeraire good.
The budget constraint relevant for the national governments incumbent in period t is
q
t
= t
t

t
o
t
(13)
where o
t
satises:
o
t
= max
_
j
t
.
.(1c)
t
:
.
t
+ ( j
t
j
t
)
1
t

.
.(1c)
t
:
.1
t
; 0
_
(14)
If the incumbent is nally kicked out of power in a given period t, the former sets the policy variables so
as to obtain a maximum consumption in period t. These decisions include the conscation of positive
prots that the downstream rm may get given the price :
c&t
t
set by the exiting incumbent. In that
period the latter gets a utility loss of

^
(1o)
o
0 in addition to the loss of future rents (consumption).
On the other hand, the expelled incumbent gets a total utility loss of

1o
13
.
The denition of subsidies in (14) merits a warning. According to empirical evidence (see, e.g., Cont
et al (2012)) the subsidies that gas companies receive are calculated as the dierence between the import
price and the domestic price of gas for nal demand. In the notation of this paper, subsidies should be
dened as

o
t
= max
_
0. (j
t
:
t
)
_
.
.(1c)
t
:
.
t

1
t
__
In this case, the upstream rms would receive the price :
t
instead of j
t
. This version of the model
considers a slightly more general assumption of a government potentially paying a price to the resource-
extracting upstream rms dierent from the nal-demand energy price, even though the marginal pro-
ductivity with respect to the resource in the downstream rm is equal to one. As it will be clear below,
this seemingly more general assumption may lead to an empirically counterfactual result that prices
received by upstream rms may not follow the same dynamics than then price of energy, which may
be read as not realistic. The task of checking how results below may change with this change in the
subsidies denition is left for future research.
Assume that
,
_
_
(1 j) j
_

_

1
j
max
+ 1
_
_
= o : 0 < o < 1 (15)
13
Note the governments lack of access to credit markets or storage technologies. Even though bureaucrats in power
are all risk neutral, access to storage may change the features of the equilibrium regulations (see Cole and Kocherlakota
(2001) for a risk-sharing problem with hidden income and hidden storage, albeit with risk averse agents). Possible future
research may explore how the addition of such technologies may change the results of this paper.
18
This condition will have an importance when considering equilibria with strictly positive subsidies.
The government regulations, together with assumptions on the exogenous stochastic variables and
parameters claried below, will imply that, in equilibrium,
1
t
will be strictly less than .
t
in every period
t. In other words, upstream will just produce in equilibrium only to satisfy the local demand. Therefore,
in such an equilibrium

1
t
=
t
r
j
t
1
1j
t
(16)
In such an equilibrium export taxes become irrelevant. All these conditions allow to write the (equilib-
rium) period- t prot for the upstream rm as
j
t

t
r
j
t
1
1j
t

t
r
t
n
t
(17)
Given the objective of the upstream rm, as long as exports are zero in the analyzed equilibrium, then
the government will aect production and investment decisions by the upstream rm only through the
regulated domestic prices j
t

1
t=0
.
3.2 The game
The interaction between producers and dierent government levels comes through a (repeated) game
whose main elements are taken from the sustainable taxation literature in macroeconomics,especially
the recent work by Yared (2010) and Acemoglu et al (2010)
14
. This timing is taken directly from Yared
(2010). The sequence of actions is as follows:
At the beginning of period t. given observing history :
t
. domestic consumers choose to re-elect the
incumbent government (that is, G
t
(:
t
) = 1) or to choose a dierent one (that is, G
t
(:
t
) = 0).
After the decision on the government, the incumbent politician chooses tax rates and regulation
variables.
Markets open and consumer and upstream rm owners make their respective decisions.
Whenever is applicable, government replacement takes place at the end of the period.
Under this timing, the government faces the decision of investing in political reputation for period
t +1 on or just decide to extract as much surplus as possible provided that at the end of the period she
14
This concept is an extension to the well-known sustainable equilibrium concept in Chari and Kehoe (1990), Chang
(1998), Phelan and Stacchetti (2001) and Fernndez-Villaverde and Tsyvinski (2004). In those papers the policy-maker
is assumed to be benevolent, whereas in this model the policy-maker-regulator is a selsh (rent-seeking) politician. This
adaptation allows to study "time-consistent taxation and regulation" within this particular framework.
19
will be removed from power. Hence, this timing implies that in every period t the bureaucrat in power
who would like to stay sets values for the policy variables that provides incentives to consumers to vote
again for her at the beginning of period t +1. At the same time, for this re-election to be rational for the
bureaucrat such policy must give her more utility value of remaining in power than of being expelled
from oce. These constraints are part of the equilibrium concept introduced below.
3.3 The equilibrium notion
Given the timing presented in the subsection above, and the absence of asymmetries of information in
the benchmark case, the basic equilibrium notion follows very closely that of Yared (2010) and Acemoglu
et. al. (2010b). In such an equilibrium the history of play at the beginning of period t includes the
realization of all stochastic variables up to period t. the history of play by the government and the
allocations chosen by rms and consumers. Thus, this notion yields the following denition.
Denition 1 A self-enforcing equilibrium is a plan of governments
_
t
A
. t

. j

. :

_
. a plan of
output, investment and input use by upstream rms

. n

. r

. a plan for consumers .

. c

. |

and
random period 1

such that:
1. at every node each plan maximizes the continuation value of the corresponding agent;
2. at the random time 1

reserves are equal to 0 from that period onwards.


As stated above, this denition borrows from that in Yared (2010) and Acemoglu et al (2010),
with the addition of the random period 1 when reserves are exhausted. It captures the fact that the
plans in the sustainable equilibrium do not present "time inconsistency" problems (given the constraints
in the model). In other words, given that in Emerging Markets countries such as Argentina policies
cannot be conceived as coming from a time consistent planner or government, but from (complex)
political processes, this model captures some dimensions of such process that are considered relevant
for the analysis of state intervention in the exhaustible-resources sector. It is a concept that allows
endogeneizing the values that dierent policy instruments take every period, explained by the political
process embedded in the game described in subsection 3.2.
4 Self-enforcing equilibrium characterization with no exports,
positive imports and positive subsidies.
This section characterizes the self-enforcing equilibrium when the upstream rm does not export in any
given period t. As stated in the introduction, this type of equilibrium in fact reects mainly the features
20
of the natural gas and a possible future oil market in Argentina. Of course, the 0-export condition
is part of a self-enforcing equilibrium when several conditions over exogenous variables hold. Those
conditions are provided also in this section.
4.1 Upstream rms optimal choice under no exports.
The equilibrium begins by the characterization of the optimal plan decided by each upstream rm. The
following proposition characterizes the solution to this problem (all proofs are in Appendix B).
Proposition 2 Suppose a self-enforcing equilibrium with 0exports. Thus the optimal policies by each
representative rm are given by the following equations: for period 1
15
r
T
= [ (1 exp (\
T
)) 1
T
]
_
j
T
j
T

T
_ 1
1
(18)
and for periods t < 1
r
t
= [ (1 exp (\
t
)) 1
t
]
_

t
[j j
t

t
]

t
_ 1
1
(19)
where

t
= ,
_
1
t
_

t+1
r
t+1
1
t+1

t + 1 < 1
_
+ 1
t
_
_
j
T
j
T
(
T
)
j
_ 1
1

t + 1 = 1
__
= ,
_
1
t
_
_

t+1
[j j
t+1

t+1
]

j
t+1
_ 1
1

t + 1 < 1
_
+ 1
t
_
_
j
T
j
T
(
T
)
j
_ 1
1

t + 1 = 1
__
(20)
The optimal level of investment n
t
in any self-enforcing equilibrium is given in the following expression.
n
t
= \
t
+
1

ln
, (1 j)
(1 ,) j
+
1

ln
t
(21)
Finally, in period 1 the following condition must hold:
_

_
1
T

_

_
1
,1
T1
_
_
j
T
j
T
q

T
_ 1
1
_
1 ,
_

_
_

_
_
1
1
1
T
_
j j
T

T
_

1
_
_ 0 (22)
15
The upperscript + is dropped for simplicity.
21
Several comments are worth making regarding the upstream rm optimal decision. Period 1 corre-
sponds to that in which reserves are exhausted. The exhaustion condition is equation (22). The latter
states that the accumulated production exceeds the accumulated investments. These are sucient con-
ditions to ensure that there are no incentives to invest in this last period. In such period, the optimal
use of the input becomes static. Indeed, the value of its marginal product (valued at the domestic price
of the resource) equalizes the input price.
For periods before the exhaustion of the resource (t < 1) the demand for the input not only depends
on the current values of the TFP shock and prices (including the price of the resource set by the
government) but also on the variable
t
.whose value depends on expectations of future values of those
same prices and TFP shocks (but not on present ones). Variable
t
essentially embeds the expected
next-period marginal rate of substitution between the two inputs of the production function. Equation
(19) is obtained from a rst-order condition that is a variant of the well-known Hotelling rule. It relates
the price-marginal-cost dierence of two consecutive periods. However, it explicitly takes into account
the fact that reserves could be exhausted in the next period (which is given by the second expression
within brackets in equation (20)). Given the stochastic nature of the model, such possibility cannot be
ruled out (and it should even have positive probability, at least for t not too small).
On the other hand, given the particular functional forms of this model (again, borrowed from Deacon
(1993)) the optimal investment decision depends on the accumulated investment spending \
t
, as well
as on the expectation about future TFP shocks as well as future output and input prices. On the
one hand this feature is consistent with the view about the dependence of the investment in natural
resources to future "perceived macroeconomic risk" and future "perceived political risk", as discussed in
e.g., Navajas et al. (2005). Also, this dependence combined with a constant-return-to-scale technology,
a lack of commitment by the incumbent and the lack of access to public debt makes the decision on
export taxes independent on the investment behavior. The domestic price set in a given period t could
have only aected the investment decision in t 1. but such investment is already taken as given by
the incumbent. Had the incumbent partial commitment for, say, one period ahead, then the decision in
period t would also include one-period ahead tax rates, aecting the investment decision in the current
period.
4.2 Self-enforcing equilibrium government policies.
The next steps is to present a characterization of the self-enforcing equilibrium. The following propo-
sition presents such characterization for the benchmark case of ` = 1 and risk-neutral owners of the
upstream rm.. The reasoning follows very closely that of Yared (2010). Thus, the main result is the
22
following:
Proposition 3 In any self-enforcing equilibrium with 0 exports and both positive imports and subsidies
the policy decisions are characterized by the following properties:
1. For any period t :
(a) the self-enforcing equilibrium price of energy :

t
follows the following law of motion:
:
t
_
:
t
_
=
_

_
: (:) =
_
.
.1\

r
_
c,
r
ouI
(,.)+_
1c
_
.
(1o)
_
j
t
if : (:
t1
)
.j
I1
_
.
_

I
r(s
I
)
1
_
+1
__
.
I
.
I1
_
(1o)
+.1
_
.
__

I1
r
I1
(s
I1
)
1
_
.+1
_
_
.
I1
.
I
_
(1o)
+.1
_
j
t
if : (:
t1
)
_
.j
I1
_
.
_

I
r(s
I
)
1
_
+1
__
.
I
.
I1
_
(1o)
+.1
.
.j
I1
_
.
_

I
r(s
I
)
1
_
+1
__
.
I
.
I1
_
(1o)
+.1
_
: (:) =
_
.
.1\

r
(c, (c)).
(1o)
_
j
t
if : (:
t1
) <
.j
I1
_
.
_

I
r(s
I
)
1
_
+1
__
.
I
.
I1
_
(1o)
+.1
(23)
where \
j

_
:.

ouI
(j,.)+
1o
_
represents the marginal shadow value (for the government) of de-
creasing the consumers total expected discounted welfare when the latter is equal to the outside
option value (i.e., when consumers are indierent between reelecting the incumbent and not
doing so), and also where \
j

(:. (:)) is the same marginal shadow value (for the govern-
ment) of decreasing the consumer
0
s total welfare when the incumbent is indierent between
continuing in power and not doing so;
(b) for suciently high values of . and both
j
and
c
then the energy producer receives strictly
positive subsidies with probability 1,
(c) for suciently high values of ., the tax rate on the endowment t

t
is equal to 1.
2. For nodes :
t
with t = 1. then the self-enforcing-equilibrium domestic price j

T
paid to the upstream
rms equals jj
T
.
3. For any period t < 1. then the self-enforcing-equilibrium domestic price j

t
paid to the upstream
23
rms satises:

j
t
j j
t
= ,1
t
_

_
_

_
(1 j)
_
_

t+1

j
t+1

t+1
_
_

1
1
_

_
j j
t+1
+

j
t+1
2 (1 j) j
_

t+1

j
t+1

j
t+1
_ 1
1

t + 1 < 1

_
(24)
, (1 j) j
2
1
1
t
_
_
j
T

j
T

_ 1
1

t + 1 = 1

_
where
j
t
= jj
t

j
t
= j j

t

t
and where
t
follows (20).
4. Subsidies are strictly positive in every state :
t
such that
[ (1 exp (\
t
)) 1
t
]
_

t
[j j
t

t
]
j

j
t
_ 1
1
< .
.(1c)
t
:
t
_
:
t
_
.
where :
t
(:
t
) is given by (23).
5. The self-enforcing-equilibrium exhaustion period 1

is characterized by the minimum 1 such that


_
T

)=1
(1)
)1
_
1
, (1 j)
(1 ,) j

T)
_
_

)
|=1
_

T|
_
[j j
T|

T|
]

T|
_
j
_ 1
1
_
+
,1
T1
_
_
j
2

T
j
T
q

T
_ 1
1
_
1 ,
1
_

_
_
1
1
1
T
_
j
2
j
T

T
_

1
_
(25)
_ 0
In this equilibrium, the incumbent at the national government level never leaves oce.
24
The last proposition presents the main results regarding key aspects of the self-enforcing-equilibrium
prices set by the incumbent bureaucrat. A major aspect is that a sucient condition for positive
subsidies is a suciently high values of the vector (
j
.
c
) . This means that when the threat against the
incumbent bureaucrat of punishing him had the bureaucrat deviate from the equilibrium policy (which
includes subsidies) then the latter are part of the equilibrium. The interpretation is not too dicult
to see. High values of such parameters induce the incumbent to maintain energy prices low enough in
any state :. Of course, the motivation to keep these prices low is to induce consumers to re-elect the
incumbent next period. Such incentives are stronger the higher are both
c
and
j
.
Regarding the evolution of the self-enforcing equilibrium energy prices :
t
, a remarkable feature
is the existence of state-contingent lower and upper bounds : (:) and : (:) . which is the analog of
the bounds-on-the-income-tax result in proposition 3 (equation (40) in Yared (2010). Clearly, such
bounds come from the incentive-compatibility constraints, one for the government (stating that the
government prefers to follow the self-enforcing equilibrium strategies rather than deviating from it, with
the automatic consequences of not being re-elected), and the other for the consumers (stating that each
consumer prefers to re-elect the incumbent rather than electing a new bureaucrat). The appendix shows
the recursive versions of such constraints. Interestingly, the second expression in (23) states that, as
long as the last period energy price is within the current period bounds, then the current-period price
is increasing in the last-period price. This shows some (weak) tendency towards an increase in the price
of energy, although this is subject to the condition of past prices within the bounds. This feature may
be read as partially consistent with the evidence observed for Argentina after 2011.
The result of the tax rate on the exportable endowment equal to 1 is partially a consequence of
both the preferences assumed for consumers (the voters) and the fact that consumers do not value the
exportable commodity. Given the consumerss preferences, a marginal increase in the tax rate decreases
the consumers utility by an amount which, rst, is proportional to the endowment and, second,
whose absolute value is decreasing in the labor productivity .. On the other hand, given the linearity
of bureaucratss preferences, the period-t utility is increasingly linear. In this case, a marginal increase
in the tax rate which is also proportional to the endowment . Thus, for the bureaucrat, a marginal
increase in the tax rate on the one hand increases both his utility but on the other hand makes the
incentive constraint for the consumers tighter (since it represents a decrease in their utility). However,
with the assumption of . being high enough the second eect is milder than the rst one, implying
that the net eect of increasing t is always positive, this providing the nal result of full taxation of
the exportable commodity (given that energy prices are low enough, and so compensating for this full
taxation outcome).
25
Another feature of this equilibrium is the independence of the domestic price of the natural resource
(produced by the upstream rm) from the price of energy (produced by the downstream rm). The
reasons for this independence can be found in two key aspects of this model: on the one hand, the
linearity of the Bernoulli utility function for the bureaucrats, and, on the other hand, the linearity in
the technology of energy production. These two features then bring the benet (for the modeler) of
allowing for a relatively simple characterization of the equilibrium law of motion of these prices, but at
the cost of predicting an independence between those two prices that may be at odds with the empirical
evidence in Argentina.
5 Interpreting facts with the model results and some policy
implications.
Some of the facts presented in section 2 can be read using the results of proposition 3. Part (1.b) presents
a sucient conditions for fact 5 (energy price much lower than the import price) with probability one.
Note that these sucient conditions include a suciently high value of the endowment of the minimum
realization of the commodity shock. The rationale for this result is as follows. High enough values of both

j
and
c
induce the bureaucrat to set very low energy prices to keep incentives for re-election, implying
subsidies. For this to be a self-enforcing equilibrium the bureaucrat should get enough revenue from
other sources. That revenue undoubtedly comes from commodity taxes. Thus, one way of interpreting
fact 5 is from this part of proposition 3 is that regulations would keep energy prices low enough following
electoral incentives, being that "sustained" by tax revenues from other commodity (i.e., tax on cereals
exports such as soybean).
Part 3 of proposition 3 characterizes the domestic producer price dynamics before reserves exhaus-
tion. Unfortunately, equation (24) does not look simple enough to state under which conditions j

t
<
j
t
. For the depletion equilibrium period 1

, however, it is true that the domestic price is strictly less


than the international price given that j < 1. For earlier periods, however, the Euler equation that
determines the dynamics of the domestic producer price j

t
does not lead to such inequality in a sharp
way. Nevertheless, a sucient condition for such inequality to hold in periods before 1

is that the term


inside the rst conditional expectation is strictly negative, which, at the same time, could occur when,
on expectation,

j

t+1
is strictly greater than
j
I+1
2
and less than or equal to j j
t+1
. There are clearly less
demanding conditions for j

t
< j
t
to hold, but the last discussion at least gives a rough idea that such
inequality depends on the expectations of future domestic prices. Given that the stochastic terminal
condition for j

T
is known, it is possible that an iteration of the expectation on the right hand side
26
of equation (24) would lead to the denitive answer. However, the problem is that such expression is
highly non-linear, and thus such conditions can only be studied by means of numerical simulations.
Part 4 of proposition 3 characterizes the states under which resource imports occur in equilibrium,
which is what has occurred with natural gas in Argentina since 2008, according to fact 3. Given the
characterization of energy prices, a sucient condition for such inequality to hold is a domestic price
much lower than the input price
t
. i.e., with a low domestic-price-to-marginal-cost relationship. Part of
this is documented in fact 4, but given lack of reliable data on unit production cost the whole inequality
cannot be put under test properly. However, fact 4 suggests that part of the explanation of energy
imports in Argentina is indeed the low incentives to produce, given the low domestic producer price,
which is also what fact 1 yields.
Regarding policy implications, as stated in the section above, a major lesson from proposition (3
is that political incentives may be the cause for strictly positive subsidies. Another lesson is that the
parameters inuencing the energy price do not aect the resource domestic price as long as subsidies
are positive. This means that, from the perspective of the resource production and investment, political
considerations do not seem to aect the natural resource sector decisions, at least at the margin. On
the other hand, the equilibrium law of motion for the natural resource domestic price is the direct con-
sequence of a "rent-seeking" bureaucrat decision. Such law of motion partially determines the behavior
of the hitting time 1. Although it is not possible to get precise sharper analytical characterization re-
sults on 1 it is clear that the bureaucrat decision (assuming that his preferences depend on his own
consumption) yields directly this equilibrium hitting time.
What kind of policy (or institutional) implications do these result imply? A direct answer would
require the full characterization of the Pareto-ecient allocation, which is out of the scope of this
paper. However, a general intuition emerges directly from the model itself. The latter assumes that
policy makers are selsh agents who consume units of the numeraire goods. In making decisions, the
bureaucrat only takes into account political incentives coming from the consumers, who are in charge of
electing the bureaucrat in charge of the government in every period. Bureaucrats have no explicit interest
in the "long-run sustainability" of upstream production (extraction) of the non-renewable resource.
The intuition then is that a more benevolent government would weight the benets of a possibly low
domestic price paid to the upstream rms for the resources (and low expected future prices embedded in
the variable
t
) with the cost of such policy to the upstream rm in the long run, especially concerning
the distribution of the equilibrium stopping time 1

.
Paradoxically, the main characterization result states that, if the objective were to reduce the amount
of subsidies paid by the government, either the parameter
c
or
j
should not be "high enough". Thus,
reducing the value of these parameters would relax the governments incentives to keep the price of
27
energy very low. However, as long as equilibrium subsidies paid by the government are still positive,
then the latter policy would not aect the domestic price of the resource paid to upstream rms. The
latter only depends on exogenous variables and parameters. This also applies clearly to the exhaustion
time 1

. Thus, in such equilibrium no policy can change the dynamics of the domestic price of the
resource. Again, the only possible reform to analyze is a discrete change of the values of
j
and/or

c
. so that in equilibrium subsidies may become null. However, in that type of equilibria, the domestic
price may be indeterminate. This issue is left for future research.
6 Concluding remarks
This paper develops a theoretical intertemporal model of to give one possible explanation for several
stylized facts about production, investment and especially, government regulation of the Argentine oil
and gas sectors. The assumptions of the model partially reect some of the distinctive features of some
of the sectors described above. In terms of its results, the theoretical model, although very stylized in
several dimensions, can qualitatively replicate several features of such regulations. The model predicts
downstream prices below the import prices (as observed in the data) as long as the threat of exclusion
from power for the regulator is strong enough. The model predicts that, with low enough resource
regulated prices relative to the variable input price then the economy imports resources from abroad.
This prediction is clearly consistent with the observed natural gas imports in Argentina after 2008.
The model also predicts a forward looking behavior for the domestic-producer price of the resource.
This behavior is at least consistent with a decrease in investment coming from a perception of lower
future (expected) prices of the resource. That is, this model would explain the decline in exploration
investment seen in the data as the result of a expected future decline in domestically regulated prices.
Regarding the policy lessons, this model suggests that, to change government incentives regarding
price and quantities regulations in energy-related-non-renewable sectors, some political reforms regard-
ing electoral incentives are needed. In particular, how the electoral system favors either political turnover
or political stability is not neutral relative to government incentives to keep prices below international
ones. As discussed in section 5, changes in
j
or
c
may aect such incentives in subtle ways. It is
important to keep producing further research to get a more precise answer about which type of electoral
reforms may improve government regulations to induce a reversion in the decline of both investment
and production.
As discussed in several parts of this paper, there is a lot of room for other extensions and improve-
ments of this basic model. First, as stated in subsection 3.1.3, the type of subsidies considered on this
version of the model may be dierent from that employed by the Argentine government in the gas sec-
28
tor after 2004. Also, although the model can qualitatively predict the equilibrium behavior of dierent
policy variables, however, it does not provide precise quantitative results to compare then with the
quantitative facts provided in section 2. In particular, how the equilibrium dynamics for resource and
energy prices, as well as its impact on production and investment and the average duration of reserves,
vary with the values of parameters
j
and
c
. may help energy regulation designers to understand not
only what type of changes in political constraints to implement but also more quantitative answers
regarding such constraints. This is left for future research.
One major assumption in this model is the long-lived politician who in equilibrium never leaves power
and so sets taxes and regulations. However, as dierent authors (such as Spiller and Tommasi, 2003),
the political dynamics in Latin America is a much more complex phenomenon, and such complexity
may have dierent implications regarding the stability of dierent political actors. Spiller and Tommasi
(2003) stress how bad incentives coming from the way that the political systems work may imply that
politicians act as if they only care about the short run. Indeed, Manzano and Monaldi (2008) stress
this point applied to the oil sector. As stated below, dierent assumptions on preferences or dierent
political types can be introduced to the benchmark model to deal with more complex political dynamics.
However, such assumption change usually is not free: more complex political dynamics may threaten
the possibility of getting clean results out of the modied model.
Another aspect is the exogeneity of the type of concession contracts assumed in this model. In
reality, concession contracts are (endogenous) outcomes of auctions. The model could then be extended
to introduce an analysis of an optimal auction mechanism, in the spirit of the analysis by Engel and
Fisher (2010)
16
, that then yields dierent regulations and taxation systems depending on the type of
contract arising from such mechanism. Particularly important is the paper by Stroebel and van Benthem
(2012). The latter develops a dynamic contracting problem taken from Thomas and Worrall (1994) and
applied to non-renewable resources. The theoretical results regarding the contract design is tested
using a WoodMackenzie dataset of hydrocarbon contracts, showing that the theoretical predictions are
consistent with this evidence. Such an exercise may be embedded in the model here to get simultaneously
results concerning regulations and contract features.
Such extension endogeneizing contracts seems important to understand recent news from the oil and
gas sectors in Argentina about political declarations threatening private companies to unilaterally nish
concession contracts. Such a threat is clearly absent in the model. Therefore, it would be an interesting
extension of this model to introduce an assumption that the government can "kick the company out" of
16
However, the original Engel and Fisher paper only deals with a mechanism design problem within model of oil
production much more stylized than in this paper. Thus, the possibility of extending the analysis by Engel and Fisher
(2010) or even using part of the standard auction theory literature in this model may lead to a potential complexity that
may not allow for analytical solutions.
29
the concession. However, for this mechanism to become an interesting one the model there are certain
assumptions that need special consideration. For example, such an extension may need an assumption
on what would be the alternative if a company is expelled from the industry. The assumption on this
alternative is not obvious for the case of Argentina. This consideration and others are all left for future
research.
Other relevant issues have been left out of the analysis. One of them is the relationship between
corporate governance, regulation design and political incentives. This issue is basically motivated by
the recent decision from the Argentine government to re-nationalize YPF in 2012. This fact induces
to think that governance problems may start interacting in subtle ways with the policy and regulation
decisions. In the benchmark model the owners of the upstream rms are risk neutral and unique for
each rm. An extension would entail multiplier owners for the upstream rms, even with risk averse
preferences instead of risk neutral. But then a possible problem (already present in the traditional
GEI literature) is the lack of denition of the objective of the rm. With multiple risk averse owners
and incomplete nancial markets there is usually disagreement about the relevant discount factor. The
static GEI literature solves this problem introducing voting mechanisms inside the rm. For such a
dynamic model, an adaptation of such voting is required. A possible idea is to use an adaptation of the
self enforcing voting model by Maggi and Morelli (2006), meaning that the voting decision in each stage
must be part of a perfect equilibrium notion. This of course will clearly complicate the equilibrium
characterization, but it could be useful to analyze such an interaction.
Another venue for future research is the consideration of asymmetric information between the reg-
ulator and the rms. Taking ideas from the well-known model by Osmundsen (1998), who, assuming
also full commitment
17
, considers the optimal contract under asymmetric information, as well as the
tax systems that implement such contract. The extension of the model in this paper would consider
also an extension of Osmundsens analysis from a two-period model to the innite horizon case
18
.
Finally, a more recent issue opened in the Argentine gas and oil sector with the recent discovery
of shale gas and shale oil reserves in the Vaca Muerta area, in the province of Neuquen. This new
and unconventional source of gas seems to play a key role in the future regarding regulations in the
gas sector. It is then a natural next step to extend the model to include this type of discoveries of
new technologies. How the availability of these new technologies aect the incentives by the incumbent
17
As Osmundsen (1998) points out (an argument borrowed from Laont and Tirole, 1988) the combination of both non
commitment and asymmetric information makes the characterization of the optimal contract less clear. This is the main
reason of why Osmunden (1998) assumes perfect commitment.
18
The recent paper by Ales et al (2012) adapts Yared (2010) to an asymmetric information environment to explain
political turnover. This extension seems very useful to adapt it to a non-renewable resource endogenous regulation (and
taxation) problem like the one in this paper.
30
national-government politician to regulate the gas market in certain ways is an almost mandatory
exercise to provide precise policy recommendations for the future gas sector.
7 Appendix: proofs
7.1 Proof of Proposition 2
Suppose rst the upstream rms problem in period 1. The latter can be written as follows:
\
&
T
(:
T
. 1
T
. j
T
) = max
(a
T
)2R
+
_
j
T

T
r
j
T
1
1j
T

T
r
T
_
where the choice of n
T
is set to 0 here (later on we check that this is indeed optimal under the condition (22)).
The rst order condition with respect to r
T
yields directly the input demand (18). Thus the upstream rms
value in period 1 is
\
&
T
(:
T
. 1
T
. j
T
) =
__
j

T
_
j

T
j
T
_ 1
1
(1 j) 1
T
(26)
For periods t < 1 the upstream rms Bellman equation is as follows:
\
&
t
(:
t
. 1
t
. j
t
) (27)
= max
(aI,&I)2R
2
+
_
j
t

t
r
j
t
1
1j
t

t
r
t
n
t
+,1
t
_
\
&
t+1
_
:
t+1
. (1 exp ( (\
t
+ n
t
))) 1
t

t
r
j
t
1
1j
t
. j
t+1
_

t + 1 < 1

+,1
t
_
__
j

T
_
j

T
j
T
_ 1
1
(1 j)
_
(1 exp ( (\
t
+ n
t
))) 1
t

t
r
j
t
1
1j
t

t + 1 = 1
__
The rst-order conditions are as follows:
j
t
r
j1
t
1
1j
t
(28)
_
j
t
,1
t
_
J\
&
t+1
J1
t+1
(:
t+1
. 1
t+1
. j
t+1
)

t + 1 < 1
_
(1 j) j

1
,1
t
_
_

T
j
T

j
T
_ 1
1

t + 1 = 1
__
=
t
31
_
1
t
_
J\
&
t+1
J1
t+1
(:
t+1
. 1
t+1
. j
t+1
)

t + 1 < 1
_
+ 1
t
_
__
j

T
_
j

T
j
T
_ 1
1
(1 j)

t + 1 = 1
__
(29)
, exp ( (\
t
+ n
t
))
= 1
The envelope condition is as follows
J\
&
t
J1
t
(:
t
. 1
t
. j
t
)
= (1 j)
t
r
j
t
1
j
t
_
j
t
,1
t
_
J\
&
t+1
J1
t+1
(:
t+1
. 1
t+1
. j
t+1
)

t + 1 < 1
_
(1 j) j

1
,1
t
_
_

T
j
T

j
T
_ 1
1

t + 1 = 1
__
Replacing here the equation (28) the latter becomes
J\
&
t
J1
t
(:
t
. 1
t
. j
t
)
=
(1 j)
t
r
j
t
1
j
t

t
j
t
r
j1
t
1
1j
t
=
_
1 j
j
_

t
r
t
1
t
Thus replacing the latter in the rst order condition (28) then
j
t
r
j1
t
1
1j
t
_
j
t
,1
t
__
1 j
j
_

t+1
r
t+1
1
t+1

t + 1 < 1
_
(1 j) j

1
,1
t
_
_

T
j
T

j
T
_ 1
1

t + 1 = 1
__
=
t
Solving for r
t
yields the demand for input (19). Replacing also in the rst order condition 29 we have that
_
1
t
__
1 j
j
_

t+1
r
t+1
1
t+1

t + 1 < 1
_
+ 1
t
_
__
j

T
_
j

T
j
T
_ 1
1
(1 j)

t + 1 = 1
__
, exp ( (\
t
+ n
t
))
= 1
Solving for n
t
yields the optimal investment plan (21). Note that in period 1 it must happen that the
accumulated output is not less (in fact, equal to) reserves. This condition is actually what denes 1. Thus 1
32
is the minimum t such that

_

_
1
,1
T1
_
_
j
T
j
T
q

T
_ 1
1
_
1 ,
_

_
_ 1
T
+
_

_

_

_
1
,1
T1
_
_
j
T
j
T
q

T
_ 1
1
_
1 ,
_

_
1
T
_

1
1
T
_
j j
T

T
_

1
which is clearly equivalent to inequality (22). Under this condition there is clearly no incentive to invest so n
&
T
= 0.
Finally, note that it is possible to get

t
= ,
_
1
t
_
_

t+1
[j j
t+1

t+1
]

j
t+1
_ 1
1

t + 1 < 1
_
+ 1
t
_
_
j
T
j
T
(
T
)
j
_ 1
1

t + 1 = 1
__
as stated in the proposition. So, solving forward:

t
= ,1
t
_
_
j
T
j
T
(
T
)
j
_ 1
1

t + 1 = 1
_
+,
_
1
t
__

1
1
t+1


1
t+1
[j j
t+1
,
t+2
]
1
1
_

t + 1 < 1
__
with

t+2
= 1
t
_
_

t+2

j
t+2
[j j
t+2

t+2
]
1
1

t + 2 < 1
_

+1
t+1
_
_
j
T
j
T
(
T
)
j
_ 1
1

t + 1 = 1
_
The nal expression comes from induction.
7.2 Proof of Proposition 3
Step 1 Using the traditional primal approach let us write the consumer and bureaucrats welfare as function
of allocations and exogenous prices and shocks. From (18) and (19) we solve for j
t
as a function of r
t
and other variables:
j
T
=

T
j
T
_
r
T
1
T
_
1j
(30)
33
j
t
=
1
j
_

t
_
r
t
1
t
_
1j
+
t
_
(31)
where
t
depends on future values j
t
(which the bureaucrat takes as exogenous in t given the lack-of-
commitment assumption).
Step 2 Both the bureaucrats consumption and the consumers welfare can be written as follows:
q
t
= t
t

t
o
t
where
o
t
=
_

_
max
_
0. (j
t
:
t
) :
.
t
.
(1c).
+
_
1
j
_
qI

I
_
aI
1I
_
1j
+
t
_
j
t
_

t
r
j
t
1
1j
t
_
; t _ 1
max
_
0. (j
t
:
t
) :
.
t
.
(1c).
t
_
; t 1
and
l
c
t
=
1
( 1)
_
.
(1c)
t
:
t
_
.1
+ .
c
t
+

t
(1 t
t
)
.
1c
t
Step 3 Given the expressions obtained in step 2, we can start writing the Bellman equations characterizing
the bureaucrats optimal program. Step 3 starts from writing the Bellman equation for periods t 1.
We start by writing the value of deviating from the self-enforcing equilibrium path for both the exiting
bureaucrat and the consumer. Dropping time subindices the former is
\
j,c&t
(:) = max
t,
_
t + (: j) :
.
.
(1c).

Clearly t
c
= 1. The rst order condition with respect to : is
(1 ) :
.2
+ j:
.1
= 0
This yields the standard Amoroso-Robinsons formula:
:
c&t
(j) =
j
1
1
.
Note that this expression implies that when deviating from the equilibrium path the government extracts
34
all prots from the downstream rm. Thus the respective deviating values are:
\
j,c&t
( :) = +
_
j
1
__
j
1
_
.
.
(1c).
= +
_
1
j
_
.1
.
(1c).

.
where
: = (j. . .)
and so, since after deviation from the incumbent the consumers expect that the new incoming bureaucrat
would also deviate from the equilibrium policy, thus obtaining

c&t
(j. .) =
1
(.1)
_
.
(1o)
j
_
.1
.
_
_
.1
+ .
c
1 ,
Now we turn to the on-equilibrium values. Assuming positive subsidies in equilibrium (a condition that will
be checked below) the problem is:
\
j
(:. (:)) = max
t,,(c
0
)
s
0
2S

_
t + (: j) :
.
.
(1c).
+ ,

c
0
2S
\
j
(:
0
. (:
0
)) Q(:. :
0
)
_
(32)
subject to
(:) =
1
( 1)
_
.
(1c)
:
_
.1
+ .
c
+
(1 t)
.
1c
+ ,

c
0
2S

0
(:
0
) Q(:. :
0
) (33)
and

0
(:
0
) _
c&t
(j
0
. .
0
) +

c
1 ,
. \:
0
o (34)
\
j
(:
0
.
0
(:
0
)) _ \
j,c&t
( :
0
)

j
1 ,
. \:
0
o (35)
35
The Lagrangian associated to the problem is:
/
tT
= t + (: j) :
.
.
(1c).
+ ,

c
0
2S
\
j
(:
0
.
0
(:
0
)) Q(:. :
0
)
+`
_
1
( 1)
_
.
(1c)
:
_
.1
+ .
c
+
(1 t)
.
1c
+ ,

c
0
2S

0
(:
0
) Q(:. :
0
) (:)
_
+

c
0
2S
,Q(:. :
0
) j (:
0
)
_

0
(:
0
)
c&t
(j
0
. .
0
)
c
_
+

c
0
2S
,Q(:. :
0
) (:
0
)
_
\
j
(:
0
.
0
(:
0
)) \
j,c&t
( :
0
) +

j
1 ,
_
The rst order conditions are

_
1
`
.
1c
_
_

_
< 0 = t

= 0
= 0 = t

[0. 1]
0 = t

= 1
(36)
:
.
_
j:
1
( 1)

.
(1c).
`
_
.
(1c)
_
.1
:
.
= 0 (37)
`(:
0
) (1 + (:
0
)) j (:
0
) = ` (38)
where, by envelope-theorem arguments:
` =
J\
j
J
; `(:
0
) =
J\
j
(:
0
.
0
(:
0
))
J (:
0
)
together with the constraints.
The second rst-order condition yields
j =
_
`.
(1c)
+ ( 1)

:
Thus : (`. :) is
: (`. :) =
_

`.
(1c)
+ 1
_
j (39)
Note that : (`. :) < j if and only if
< `.
(1c)
+ 1
36
equivalent to
` .
1c
which implies that t

= 1. The rest of the argument follows that in the proof of lemma 3 and proposition
3 of Yared (2010). First we need to show that is contained in a compact set [
c
.
c
] for : o. First,
note that cannot go to since : 0. . 0. 0. t [0. 1] . Thus, for every : o. has a
lower bound, called
c
. Also, since the government would never choose : below :, dened as the value
of : that satises:
t:
.
(j :) .
(1c).
= 0
then
_
.
(1o)

_
.1
has an upper bound, and so has . Call
c
the upper bound for at : o. To show
that the interval is also closed, let the sequence
)

1
)=1
with
)
being in the set of feasible
0
: (which
was shown to be bounded), such that
)

0
. We need to show that
0
is also in the set. Clearly, if

)
is in the set, there exist optimal choices : (
)
) . t (
)
)
1
)=1
. Each element of this sequence satises
all the constraints, including the weak inequalities. Thus, the limit of : (
)
) . t (
)
)
1
)=1
satises the
incentive compatibility constraints. Also, for each ,. the vector (: (
)
) . t (
)
)) is contained in a compact
subset of R
2
. Thus, by the Dominated Convergence Theorem the limit of : (
)
) . t (
)
)
1
)=1
allows the
consumer to achieve
0
. Thus,
0
must be in the set. Also, to prove that \
j
is strictly concave in .
take two dierent values for in the set [
c
.
c
] . say,
0
and ". Associated to them there are optimal
plans (: (
0
) . t (
0
)) and (: (") . t (")) . Consider the transformation : () = [: ()]
1.
. Thus, for
both values of we have
=
1
( 1)
.
(1c)(.1)
: () + .
c
+
(1 t ())
.
1c
+ ,

c
0
2S

0
(:
0
) Q(:. :
0
)
Thus the right hand side is clearly linear in (:. t) . Thus, the convex combination i[: (
0
) . t (
0
)] +
(1 i) [: (") . t (")] clearly satises this equality for
i
= i
0
+ (1 i) ". Now, the bureaucrats
consumption can be rewritten as follows
t +
_
: () j: ()
s
s1
_
.
(1c).
This function is linear in t but strictly concave in :. Thus, i[: (
0
) . t (
0
)] + (1 i) [: (") . t (")]
allows the bureaucrat to achieve a strictly higher consumption and hence utility. This shows that \
j
is
strictly concave in .
The last argument allows to follow the argument given in the proof of proposition 3 in Yared.
37
Dene
`(:) = \
j

_
:.

c&t
(j. .) +
c
1 ,
_
;

`(:) = \
j

(:. (:)) . (:) such that \


j
(:. (:)) = \
j,c&t
( :)

j
1 ,
Suppose rst that `(:
t1
) < `(:) . The incentive constraint (34) and concavity of \
j
implies that
`(:
t
) _ `(:) . Thus, from (38) we have that j (:
t
) 0. implying that `(:
t
) = `(:) . Suppose also
that `(:
t1
)

`(:) . Then, the incentive constraint (35) and the monotonicity of \
j
implies that `(:
t
)
_

`(:) . Thus, from (38) we have that (:
t
) 0 and so `(:
t
) =

`(:) . Therefore, as in Yared (2010)
this argument shows that
`
_
:
t
_
=
_

_
`(:) if `(:
t1
) < `(:)
`(:
t1
) if `(:
t1
)
_
`(:) .

`(:)

`(:) if `(:
t1
)

`(:)
(40)
This expression allows to get the expression for the equilibrium value of :
t
(:
t
) . t 1. In particular the
energy price in period t. history :
t
follows:
:
t
_
:
t
_
=
_

_
: (:) =
_
.
.1\

r
_
c,
r
ouI
(,.)+_
1c
_
.
(1o)
_
j if : (:
t1
)
.j
I1
_
.
_

I
r(s
I
)
1
_
+1
__
.
I
.
I1
_
(1o)
+.1
_
.
__

I1
r
I1
(s
I1
)
1
_
.+1
_
_
.
I1
.
I
_
(1o)
+.1
_
j
t
if : (:
t1
)
_
.j
I1
_
.
_

I
r(s
I
)
1
_
+1
__
.
I
.
I1
_
(1o)
+.1
.
.j
I1
_
.
_

I
r(s
I
)
1
_
+1
__
.
I
.
I1
_
(1o)
+.1
_
: (:) =
_
.
.1\

r
(c, (c)).
(1o)
_
j if : (:
t1
) <
.j
I1
_
.
_

I
r(s
I
)
1
_
+1
__
.
I
.
I1
_
(1o)
+.1
which is the law of motion that appears in the proposition statement. Note that
.
.1\

r
_
c,
r
ouI
(,.)+_
1c
_
.
(1o)
<
1 if and only if
.
(1c)
< \
j

_
:.

c&t
(j. .) +
1 ,
_
This inequality is clearly ensured when both
c
and
j
are large enough, since \
j

_
:.

ouI
(j,.)+
1o
_
is
strictly decreasing in its second argument. This ensures that for any :
t
then :
t
(:
t
) < j
t
, implying
strictly positive subsidies for every t. Also, the latter implies that `(:) .
1c
. implying that t
.
t
= 1
in this equilibrium.
38
Step 4 Let any node such that t = 1. Assuming strictly positive subsidies, in this case then we can write the
governments Bellman equation in the following form:
\
j
T
(:
T
. 1
T
.
T
) = max
a
T
,t
T
,
T
,
T+1
(c
0
)
s
0
2S

_
j
T

T
r
j
T
1
1j
T


T
j
r
T
+ t
T

T
+ (:
T
j
T
) :
.
T
.
(1c).
T
+,

c
0
2S
\
j
(:
0
.
T+1
(:
0
)) Q(:. :
0
)
_
subject to the same constraints (33), (34) and (35), as in step 3 (together with non-negativity constraints
on r
T
). In an interior solution, the choices of t
T
. :
T
and
_

T+1
(:
0
)
c
0
2S
_
satisfy the same rst-order
conditions as in step 3. Thus, the expressions obtained in step 3 also hold for t = 1 for such variables.
Regarding the choice of r
T
then its FOC is
jj
T

T
r
j1
T
1
1j
T
=

T
j
= r

T
=
_
j
2
j
T

T
_ 1
1
1
T
Thus replacing the latter on the right hand side of the Bellman equation the rst two terms simply yield:
_
j
T

j
T
_ 1
1
1
T
(1 j) j
2
1
Thus the equilibrium domestic price of the commodity at time of complete exhaustion is equal to
j

T
= jj
T
as stated in the proposition. A sucient condition for strictly positive subsidies in period 1 is:
_
1

`.
(1c)
_
_
` .
(1c)
+ 1
_
.1
.
.(1c)

.
(1 j) (j
max
)
.
_
j
2
j
max

_

1

1
1

where

` = max
c

`(:) . ` = min
c
`(:) . Again, when both
c
and
j
are large enough, and when . is
large, then such inequality holds.
39
Note also that the Bellman equation can be also
\
j
T
(:
T
. 1
T
.
T
)
=
_
j
T

j
T
_ 1
1
1
T
j
2
1
(1 j) +
+ max
t
T
,
T
,
T+1
(c
0
)
s
0
2S

_
t
T

T
+ (:
T
j
T
) :
.
T
.
(1c).
T
+ ,

c
0
2S
\
j
(:
0
.
T+1
(:
0
)) Q(:. :
0
)
_
Thus, let
\
j
(:
T
.
T
) = max
t
T
,
T
,
T+1
(c
0
)
s
0
2S

_
t
T

T
+ (:
T
j
T
) :
.
T
.
(1c).
T
+ ,

c
0
2S
\
j
(:
0
.
T+1
(:
0
)) Q(:. :
0
)
_
(41)
Then
\
j
T
(:
T
. 1
T
.
T
) =
_
j
T

j
T
_ 1
1
1
T
j
2
1
(1 j) + \
j
(:
T
.
T
)
Note that the equality (41) is the same Bellman equation as in periods t 1.
Step 5 Suppose now any node :
t
such that t < 1.Thus, the Bellmans equation can be written in this form:
\
j
t
(:
t
. 1
t
. 1
t
.
t
)
= max
aI,tI,I,
I+1
(c
0
)
s
0
2S

__
j
t

1
j
_

t
_
r
t
1
t
_
1j
+
t
__

t
r
j
t
1
1j
t
+ t
t

t
+ (:
t
j
t
) :
.
t
.
(1c).
t
+,

c
0
2S
c
T
\
j
t+1
_
:
0
.
t+1
(:
0
) . (1 exp ( (\
t
+ n
t
))) 1
t

t
r
j
t
1
1j
t
_
Q(:
t
. :
0
)
+, (1 j) j
2
1

c
0
2S
T
_
_
j
T

j
T
_ 1
1
_
(1 exp ( (\
t
+ n
t
))) 1
t

t
r
j
t
1
1j
t
_
_
Q(:
t
. :
0
)
+,

c
0
2S
T
\
j
(:
0
T
.
T
(:
0
)) Q(:
t
. :
0
)
_
where the maximization is subject to the same constraints (33), (34) and (35). Also, recall that the value
of the Lagrange multiplier `
T
(:
0
) in state :
0
may depend on
T
(:
0
). In fact, only

` and ` depends on
. since in the rest of the range for `
T
the latter is equal to `
t
. which is independent of
T
. Thus, the
40
rst-order conditions with respect to r
t
yields:
, (1 j) j
2
1

c
0
2S
T
_
j
T

j
T
_ 1
1
Q(:
t
. :
0
) + ,

c
0
2S
c
T
J\
j
t+1
J1
t+1
Q(:
t
. :
0
)
=
r
t
j
_
jj
t

t
r
t


t
j
t
r
j
t
1
1j
t
_
while the FOC with respect to
t+1
(:
0
) (for :
0
o
c
T
) is
`
t+1
(:
0
)
_
1 +
t+1
(:
0
)
_
j
t+1
(:
0
) = `
t
(:
t
)
and the FOC with respect to
t+1
(:
0
) (for :
0
o
T
) is
`
T
(:
0
) (1 +
T
(:
0
)) +,Q(:
t
. :
0
) j
T
(:
0
) = `
t
(:
t
)
By the Envelope condition:
J\
j
t
J1
t
= (1 j)
t
r
j
t
1
1j
t
_
_
_
2
j
_

t
(r
t
)
1j
1
j2
t
_

_
j
t

1
j
(
t
)
_
1
t
_
_
_
+
_
(1 j)
t
r
j
t
1
j
t
1
_
_
_
, (1 j) j
2
1

c
0
2S
T
_
j
T

j
T
_ 1
1
Q(:
t
. :
0
) + ,

c
0
2S
c
T
J\
j
t+1
J1
t+1
Q(:
t
. :
0
)
_
_
Using the rst-order condition with respect to r
t
:
J\
j
t
J1
t
= (1 j)
t
r
j
t
1
1j
t
_
_
_
2
j
_

t
(r
t
)
1j
1
j2
t
_

_
j
t

1
j
(
t
)
_
1
t
_
_
_
+
r
t
j
_
jj
t

t
r
t


t
j
t
r
j
t
1
1j
t
_
=
2
j
(1 j)
t
r
t
1
t
+
_
1 (1 j)
t
r
j
t
1
j
t
__
j
t


t
j
_


t
r
1j
t
j
2

t
1
1j
t
41
Thus the rst-order condition with respect to r
t
is
j
t


t
j


t
r
1j
t
j
2

t
1
1j
t
= , (1 j) j
2
1

c
T
2S
T
_
j
T

j
T
_ 1
1
Q(:
t
. :
0
)
+,

c
I+1
2S
c
T
_
2
j
(1 j)
t+1
r
t+1
1
t+1
+
_
1 (1 j)
t+1
r
j
t+1
1
j
t+1
__
j
t+1


t+1
j
_


t+1
r
1j
t+1
j
2

t+1
1
1j
t+1
_
Q(:
t
. :
0
)
Recalling that
j
t
=
1
j
_

t
_
r
t
1
t
_
1j
+
t
_
Then
r
t
1
t
=
_

t
_ 1
1
[j j
t

t
]
1
1
Thus, the rst-order condition with respect to r
t
can be rewritten as
j
t


t
j

1
j
_
j
t


t
j
_
= , (1 j) j
2
1

c
T
2S
T
_
j
T

j
T
_ 1
1
Q(:
t
. :
0
)
+,

c
I+1
2S
c
T
2
j
(1 j)
_

t+1
[j j
t+1

t+1
]

j
t+1
_ 1
1
Q(:
t
. :
0
)
+,

c
I+1
2S
c
T
_

_
_

_
1 (1 j)
_
_

t+1
[j j
t+1

t+1
]

t+1
_
_

1
_

_
_
j
t+1


t+1
j
_

1
j
_
j
t+1


t+1
j
__
Q(:
t
. :
0
)
where
t
follows (20). Dening
j
t
= jj
t

j
t
= j j
t

t
42
then the latter expression is
j
t


j
t
j
= , (1 j) j
1+
1
1
t
_
_
j
T

j
T
_ 1
1

t + 1 = 1
_
+,

c
I+1
2S
c
T
1
t
_

_
2 (1 j)
_

t+1

j
t+1

j
t+1
_ 1
1
+
_

_
1 (1 j)
_
_

t+1

j
t+1

t+1
_
_

1
_

_
j
t+1


j
t+1
j

t + 1 < 1
_

_
The expression just obtained corresponds to (24) in the statement of the proposition.
Step 6 To get the expression (25) is based on the following progression in determining accumulated natural
resource output:
1
1
= 1
0

1
1
0
_
[j j
0

0
]

0
_

1
1
2
=
_

_
1
, (1 j)
(1 ,) j

1
_
1
0

1
1
0
_
[j j
0

0
]

0
_

1
_

1
1
1
_
[j j
1

1
]

1
_

1
=
_
1
, (1 j)
(1 ,) j

1
_

1
1
1
_
[j j
1

1
]

1
_

1
1
0

1
1
0

1
1
1
_
[j j
0

0
]

0
_

1
_
[j j
1

1
]

1
_

1
1
3
=
_

_
1
, (1 j)
(1 ,) j

2
_

_
1
, (1 j)
(1 ,) j

1
_

1
1
1
_
[j j
1

1
]

1
_

1
1
0

1
1
0

1
1
1
_
[j j
0

0
]

0
_

1
_
_
[j j
1

1
]

1
_

1

1
1
2
_
[j j
2

2
]

2
_

1
=
_
1
, (1 j)
(1 ,) j

2
_

1
1
2
_
[j j
2

2
]

2
_

1

_
1
, (1 j)
(1 ,) j

1
_

1
1
1

1
1
2
_
[j j
1

1
]

1
_

1
_
[j j
2

2
]

2
_

1
+1
0

1
1
0

1
1
1

1
1
2
_
[j j
0

0
]

0
_

1
_
[j j
1

1
]

1
_

1
_
[j j
2

2
]

2
_

1
43
In general
1
t
=
_
1
, (1 j)
(1 ,) j

t1
_

1
1
t1
_
[j j
t1

t1
]

t1
_

1

_
1
, (1 j)
(1 ,) j

t2
_

1
1
t2

1
1
t1
_
[j j
t2

t2
]

t2
_

1
_
[j j
t1

t1
]

t1
_

1
+
_
1
, (1 j)
(1 ,) j

t3
_

1
1
t3

1
1
t2

1
1
t1
_
[j j
t3

t3
]

t3
_

1
_
[j j
t2

t2
]

t2
_

1
_
[j j
t1

t1
]

t1
_

1
... + (1)
t1
1
0

1
1
0
...
1
1
t1
_
[j j
0

0
]

0
_

1
...
_
[j j
t1

t1
]

t1
_

1
So
1
t
=
t

)=1
(1)
)1

_
1
, (1 j)
(1 ,) j

t)
_
_

)
|=1
_

t|
_
[j j
t|

t|
]

t|
_
j
_ 1
1
_
Thus, replacing the latter in the expression (22) yields the desired result.
Step 7 The nal step corresponds to check the conditions for which subsidies are strictly positive for t < 1.
Subsidies in these nodes are strictly positive if and only if
_
`
t
.
(1c)
t
+ 1
_
.1

.
.
.(1c)
t
j
1.
t
_
1 `
t
.
(1c)
t
_
(j
t
j
t
) 1
t
_
_

t

j
t

t
_
_

1
where j
t
follows (24) and
t
follows (20). Given that 1
t
_ . and also
j
t
= jj
t

j
t
= j j
t

t
44

j
t
= j j
t
+ ,

c
I+1
2S
c
T
1
t
_

_
_

_
(1 j)
_
_

t+1

j
t+1

t+1
_
_

1
1
_

_
j j
t+1
+

j
t+1
2 (1 j) j
_

t+1

j
t+1

j
t+1
_ 1
1

t + 1 < 1
_
_
, (1 j) j
2
1
1
t
_
_
j
T

j
T
_ 1
1

t + 1 = 1
_
_ j j
t
+

c
I+1
2S
c
T
,1
t
_

_
(1 j) j
_
_

t+1

j
t+1

t+1
_
_

1
j
t+1
+

j
t+1

t + 1 < 1
_

_
_ j j
t
+

c
I+1
2S
c
T
,1
t
_

_
max
_
_

j
t+1
_

1
.

j
t+1
_
_

_
(1 j) j
_
_

t+1

t+1
_
_

1
j
t+1
+ 1
_

t + 1 < 1
_

_
Note that
max
_
_

j
t+1
_

1
.

j
t+1
_
_

j
1
1
t+1
conditional on

j
t+1
_ 1. Thus, when imposing a transversality condition on

j
t
and after some algebra
then

j
t
_ j j
t
+

c
I+1
2S
c
T
,1
t
_

j
1
1
t+1
_

_
(1 j) j
_
_

t+1

t+1
_
_

1
j
t+1
+ 1
_

j
t+1
_ 1. t + 1 < 1
_

_
45
_ j j
t
+

c
I+1
2S
c
T
,1
t
_

_
_
_
_
j j
t+1
+

c
I+2
2S
c
T
,1
t+1
_

j
1
1
t+2
_

_
(1 j) j
_
_

t+2

t+2
_
_

1
j
t+2
+ 1
_

j
t+2
_ 1. t + 2 < 1
_

_
_
_
_
1
1
_

_
(1 j) j
_
_

t+1

t+1
_
_

1
j
t+1
+ 1
_

j
t+1
_ 1. t + 1 < 1
_

_
_ j j
t
+

c
I+1
2S
c
T
,1
t
_

_
(1 j) j
_
_

t+1

t+1
_
_

1
j
t+1
+ 1
_

_
(j j
t+1
)
1
1
+
1

)=1
,
1+)
1
t
_

_
(1 j) j
_
_

t+1

t+1
_
_

1
j
t+1
+ 1
_

_
_

)
|=1
_

_
(1 j) j
_
_

t+1+|

t+1+|
_
_

1
j
t+1+|
+ 1
_

_
_

_
(j j
t+)
)
1
1
Since the support of j
t
,
t
and
t
. and since
,
_
_
(1 j) j
_

_

1
j
max
+ 1
_
_
= o : 0 < o < 1
then the right hand side is is bounded by jj
max
plus a constant multiplied by
1
1c
. This implies that

j
t
has a xed upper bound. Call this upper bound. Then as long as
_
1

`.
(1c)
_
_
` .
(1c)
+ 1
_
.1
.
.(1c)

.
j
max

_

1
then subsidies are strictly positive for every node with t < 1. Again, a sucient condition for this is to
make both
j
and
c
large enough.

46
References
[1] Abreu, Dilip, David Pierce and Ennio Stacchetti (1990). "Toward a Theory of Discounted Repeated
Games with Imperfect Monitoring." Econometrica 58: 1041-1063.
[2] Acemoglu, Daron, Mikhail Golosov and Aleh Tsyvinski (2010a). "Dynamic Mirrlees Taxation under
Political Economy Constraints." Review of Economic Studies 77: 841881.
[3] Acemoglu, Daron, Mikhail Golosov and Aleh Tsyvinski (2010b). "Political Economy of Ramsey
Taxation". Journal of Public Economics 95: 467475.
[4] Aghion, Phillippe and Luca Quesada (2010). "Petroleum Contracts: What Does Contract Theory
Tells Us?" In Hogan, William and Federico Sturzenegger (eds.) The Natural Resources Trap: Private
Investment without Public Commitment. Cambridge MA. MIT Press: 47-60.
[5] Aguilar, Ruben and Daney Valdivia (2011). "Bolivian natural gas export prices: Modeling and
forecast pooling." MPRA Working Paper No. 35485.
[6] Ales, Laurence, Pricila Maziero and Pierre Yared (2012). "Political Distorsions and Endogenous
Turnover." Working paper.
[7] Barril, Diego and Fernando Navajas (2011). "What drove down natural gas production in Ar-
gentina?" Working Paper.
[8] Bohn, Henning and Robert Deacon (2000). "Ownership Risk, Investment and the Use of Natural
Resources." American Economic Review 90: 526549.
[9] Campbell, H.F. and R. Lindner (1985). "A Model of Mineral Exploration and Resource Taxation."
Economic Journal 95: 146160.
[10] Campodnico, Humberto (2008). "Renta petrolera y minera en pases seleccionados de Amrica
Latina." Working Paper, ECLAC.
[11] Caratori, Luciano (2010). "Evolucin de las reservas de hidrocarburos en Argentina en la dcada
de 2000." Instituto Argentino de la Energa General Mosconi.
[12] Cole, Harold and Narayana Kocherlakota (2001). "Ecient Allocations with Hidden Income and
Hidden Storage." Review of Economic Studies 68: 523542.
47
[13] Cont Walter, Pedro Hancevic and Fernando Navajas (2011). Energy populism and household
welfare. MPRA Working Paper No. 35725.
[14] Deacon, Robert (1993). "Taxation, Depletion and Welfare: A Simulation Study of the U.S. Petro-
leum Resource." Journal of Environmental Economics and Management 24, 159-187.
[15] Duncan, Roderick (2006). "Price or Politics? An Investigation of the Cause of Expropriation." The
Australian Journal of Agricultural and Resource Economics 50: 85101.
[16] Engel, Eduardo and Ronald Fisher (2010). "Optimal Resource Extraction Contracts under Threat
of Expropriation." In Hogan, William and Federico Sturzenegger (eds.) The Natural Resources
Trap: Private Investment without Public Commitment. Cambridge MA. MIT Press: 161-196.
[17] Favaro, Orietta (1992). "Territorio y petrleo: la intervencin estatal en la industria petrolera. El
caso Neuqun (1884-1955)." Revista de Historia 3: 159-180.
[18] Gadano, Nicols (2000). "Determinantes de la inversin en el sector petrleo y gas de la Argentina."
In Heymann, Daniel and Bernardo Kosaco (eds). La Argentina de los Noventa. Buenos Aires,
EUDEBA/ECLAC.
[19] Gadano, Nicols (2006). Historia del Petrleo en la Argentina. Buenos Aires, EDHASA.
[20] Gadano, Nicols (2010). "Urgency and Betrayal: Three Atempts to Foster Private Investmjent
in Argentinas Oil Industry". In Hogan, William and Federico Sturzenegger (eds.) The Natural
Resources Trap: Private Investment without Public Commitment. Cambridge MA. MIT Press:
369-404.
[21] Garnaut, Ross and Anthony Clunies Ross (1975). "Uncertainty, Risk Aversion and the Taxing of
Natural Resource Projects." Economic Journal 85: 272-287.
[22] Gaudet, Grard, Pierre Laserre and Ngo V. Long (1995). "Optimal Resource Royalties with Un-
known and Temporally Independent Extraction Cost Structures." International Economic Review
36: 715-749.
[23] Hartley, Peter and Kenneth Medlock III (2008). "A Model of Operation and Development of a
National Oil Company." Energy Economics 30: 2459-2485.
[24] Heaps, Terry and John Helliwell (1985). "The Taxation of Natural Resources." In Alan Auerbach
and Martin Feldstein (eds.) Handbook of Public Economics, Volume I. Amsterdam, North Holland,
421-472.
48
[25] Hobbs, Philip, Suzan Oelofse and Jeanette Rascher (2008). "Management of Environmental Im-
pacts from Coal Mining in the Upper Olifants River Catchment as a Function of Age and Scale."
International Journal of Water Resources Development 24: 417-431.
[26] Hogan, William and Federico Sturzenegger (2010). The Natural Resources Trap: Private Investment
without Public Commitment. Cambridge MA. MIT Press.
[27] Kozulj, Roberto (2002). "Balance de la privatizacin de la industria petrolera en Argentina y
su impacto sobre las inversiones y la competencia en los mercados minoristas de combustibles."
ECLAC Working Paper.
[28] Kozulj, Roberto (2005). "Crisis de la industria del gas natural en Argentina." ECLAC Working
Paper.
[29] Kozulj, Roberto (2005). "Estrategias para el sector energtico y scalidad: un desafo crtico."
Voces en el Fenix 13: 84-91
[30] Lund, Diderik (1992). "Petroleum Taxation under Uncertainty: Contingent Claim Analysis with
an Application to Norway." Energy Economics 14: 23-31.
[31] Lund, Diderik (2010). "Rent Taxation for Non Renewable Resources." Annual Review of Resource
Economics 1: 287-307.
[32] Managi, Shunsuke, James J. Opaluch, Di Jin and Thomas A. Grigalunas (2006). "Stochastic Fron-
tier Analysis of Total Factor Productivity in the Oshore Oil and Gas Industry." Ecological Eco-
nomics 60: 204-15.
[33] Manzano, Osmel (2000). Tax Eects upon Oil Field Development in Venezuela. Working paper
2000-006. Massachusetts Institute of Technology, Center for Energy and Environmental Policy
Research.
[34] Manzano, Osmel and Francisco Monaldi (2008). "The Political Economy of Oil Production in
Latin America." Economia: Journal of the Latin American and Caribbean Economic Association
9: 59-98.
[35] Navajas, Fernando, Santiago Urbiztondo, Walter Cont and Ramiro Moya (2005). Inversin y e-
ciencia contractual: Qu hace distintos a los recursos naturales? Teora y evidencia para la
Argentina. Buenos Aires, Fundacin YPF.
49
[36] Navajas, Fernando (2006). "Energo Crunch Argentino 2002-20XX". Working Paper FIEL.
[37] Osmundsen, Petter (1998). "Dynamic Taxation of Non-Renewable Resources under Asummetric
Information about Reserves." Canadian Journal of Economics 31: 933-951.
[38] Peralta, Mara G., Andrea P. Abella y Juan F. Albarenque (2005). Tratamiento impositivo y
jurdico de las regalas en la Argentina". Petrotecnia (Review of the Instituto Argentino del Petrleo
y Gas). August: 62-97.
[39] Petrecolla, Diego and Mara Fernanda Martnez (2008). "Condiciones de competencia en el mercado
de gas natural de la Repblica Argentina: 1990-2008". Economa 33: 177-199.
[40] Piano, Horacio (2003). "Explotacin de los recursos no renovables en un sistema federal de gob-
ierno." Working Paper, Universidad Nacional de La Plata.
[41] Postali, Fernando (2007). "The Brazilian Fiscal System and the decision to invest in petroleum
reserves: A Real Options Approach." University of Sao Paulo Working Paper.
[42] Prado, Oscar (2005). "Situacin y perspectivas de la minera metlica argentina." ECLAC working
paper.
[43] Rigobn, Roberto (2010). "Dealing with Expropriations: General Guidelines for Oil Production
Contract." In Hogan, William and Federico Stuzenegger (eds.) The Natural Resources Trap: Private
Investment without Public Commitment. Cambridge MA. MIT Press: 227-254.
[44] Roccaro, Isabel and Edgardo Fernndez (2006). Aspectos tributarios del sector de hidrocarburos:
El caso argentino". Working Paper, Universidad Nacional de Cuyo.
[45] Scheimberg, Sebastin (2007). "Experiencia reciente y desafos para la generacin de renta petrolera
aguas arriba en la Argentina." Santiago de Chile. ECLAC Working Paper.
[46] Scheimberg, Sebastin (2011). "Desempeo del sector petrolero en la ltima dcada. Los efectos
distributivos del presente marco regulatorio. Amenazas y Oportunidades a futuro." Anales de la
Asociacin Argentina de Economa Poltica.
[47] Serebrisky, Toms (2001). "How Competitive is the Argentine Gasoline Market? A Dierentiated
Product Model." Working paper.
50
[48] Spiller, Pablo and Mariano Tommasi (2003). "The Institutional Determinants of Public Policy: A
Transaction Approach with Application to Argentina". Journal of Law, Economics, and Organiza-
tion 19: 281-306.
[49] Stroebel, Johannes and Arthur van Benthem (2012). "Resource Extraction Contracts Under the
Threat of Expropriation: Theory and Evidence." Review of Economics and Statistics: forthcoming.
[50] Thomas, Jonathan and Tim Worrall (1994). "Foreign Direct Investment and the Risk of Expropri-
ation." Review of Economic Studies 61: 81-108.
[51] Vera, Juan Carlos (2000). "Regalas ptimas en una cuenca petrolera." Anales de la Asociacin
Argentina de Economa Poltica.
[52] Wernerfelt, Nils and Richard Zeckhauser (2010). "Denying the Temptation to GRAB." In Hogan,
William and Federico Stuzenegger (eds.) The Natural Resources Trap: Private Investment without
Public Commitment. Cambridge MA. MIT Press: 197-226.
[53] Yared, Pierre (2010). "Politicians, Taxes and Debt". The Review of Economic Studies 77: 806-840.
51