Sei sulla pagina 1di 15

Groundwater pumping by heterogeneous users

Alexander E. Saak & Jeffrey M. Peterson

Abstract Farm size is a signicant determinant of both resource users are identical. This, along with the assump-
groundwater-irrigated farm acreage and groundwater-irriga- tion of instantaneous inter-seasonal transmissivity, sim-
tion-application rates per unit land area. This paper analyzes plies the analysis because there exists a representative
the patterns of groundwater exploitation when resource users user. However, this approach does not take into account
in the area overlying a common aquifer are heterogeneous. In the spatial distribution of users, and the dependence of
the presence of user heterogeneity, the common resource individual groundwater stocks on the history of past
problem consists of inefcient dynamic and spatial allocation extractions (Brozovic et al 2003; Koundouri 2004).
of groundwater because it impacts income distribution not Recently, some authors have taken into account the spatial
only across periods but also across farmers. Under competitive variability in groundwater use, either by relaxing the
allocation, smaller farmers pump groundwater faster if farmers assumption of instantaneous lateral ows (e.g., Brozovic
have a constant marginal periodic utility of income. However, et al. 2010) or by introducing spatial heterogeneity in the
it is possible that larger farmers pump faster if the Arrow-Pratt marginal value of resource use (e.g., Gaudet et al. 2001;
coefcient of relative risk-aversion is sufciently decreasing in Xabadia et al. 2004).
income. A greater farm-size inequality may either moderate or This article addresses another source of heterogeneity,
amplify income inequality among farmers. Its effect on that of variation in the size of the land area from which
welfare depends on the curvature properties of the agricultural each user can access the resource. This is an important
output function and the farmer utility of income. Also, it is issue because irrigated agriculture, one of the major
shown that a at-rate quota policy that limits the quantity of consumers of groundwater, is comprised of farms of
groundwater extraction per unit land area may have unintend- widely varying sizes (Schaible 2004; Hoppe et al. 2010).
ed consequences for the income distribution among farmers. Knapp and Vaux (1982), Feinerman (1988), (Foster and
Rosenzweig 2008), and Sekhri (2011) are among the few
Keywords Agriculture . Conceptual models . studies addressing variation in farm size or in pumping
Groundwater management volume.
It is well known that, to the extent that groundwater is
a common property resource, private decisions lead to
inefcient allocation. This result holds unless the aquifer
Introduction is relatively large in comparison to total groundwater use,
users can cooperate, or hydraulic conductivities are so
Theoretical models of groundwater extraction typically small that the resource is effectively private (Feinerman
assume that the resource is non-exclusive or that the and Knapp 1983). However, it is not clear whether
heterogeneity in farm size alleviates or exacerbates the
so-called tragedy of the commons (Hardin 1968). To the
Received: 27 June 2011 / Accepted: 29 March 2012
Published online: 17 May 2012 extent such effects are present, there are potentially
important policy implications, because redistributive pol-
* Springer-Verlag 2012 icies will then interact with policies to correct the common
property externalities: policies targeting one of these
A. E. Saak ())
domains may have unintended impacts in the other.
Markets, Trade, and Institutions Division, To understand the presence and nature of any such
International Food Policy Research Institute, interactions, the following questions are posed in this
2033 K Street, NW, Washington, DC, 20006-1002, USA article: What are the determinants of the relationship
e-mail: asaak@cgiar.org between farm size and groundwater use intensity? How
Tel.: +1-202-8624543
Fax: +1-202-4674439 does the distribution of farm sizes in the area inuence the
efciency of groundwater allocation? What are the
J. M. Peterson distributional impacts of farmland ownership structure
Department of Agricultural Economics,
Kansas State University, and water management policies? To analyze these ques-
216 Waters Hall, Manhattan, KS 66506-4011, USA tions a two-period model is developed where land above
e-mail: jpeters@agecon.ksu.edu an aquifer, all of which can be irrigated but is of

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


836
undifferentiated quality, is gathered into farms of unequal with which the aquifer is depleted. The spatial component
size. The differences in pumping rates across farms of refers to the effect of farm-size inequality on the
different sizes in this framework are entirely due to an distribution of pumping rates and income across farmers
endogenous interaction between common property effects in each period. The direction of the overall effect depends
and farm-size inequalities. on the magnitude and direction of both these components,
For both methodological and policy reasons, it is which are determined by additional curvature conditions
helpful to distinguish between the cases where farmers on the production function.
utility-of-income functions are linear and where they are Sufcient conditions are derived that identify the cases
concave. In the rst case, marginal utility of income is where an increase in inequality leads to a reduction in
constant, which is an appropriate representation of cases social welfare. These conditions are quite restrictive,
where small farmers supplement their incomes with off- requiring specic curvature properties of the production
farm sources (e.g., off farm employment of some function, suggesting that there are many cases where
household members). Even if the underlying utility inequality is not welfare reducing. Indeed, in many cases
functions are concave, in these cases there is no inherent inequality may actually raise social welfare because it
reason that small farmers have smaller incomes than (or a dampens the tragedy of the commons problem. Moreover,
marginal utility of income that differs from) large farmers. as illustrated with a numerical example, greater farm-size
The second case presumes that income from irrigated inequality may imply less income inequality. This is
farming activities is the sole source of income, which is because of an effect similar to that identied by Foster and
more appropriate for many developing country contexts. Rosenzweig (2008): smaller farmers have a strategic
As small farms have a smaller capacity to generate advantage as they are able to poach more groundwater
income, they have a higher marginal utility of income per unit land than their larger neighbors.
that raises the stakes of the tradeoffs in allocating water When utility is concave, the analysis has another layer
across farmers and across periods. of complexity. The pure income redistribution effect of the
Linear utility is a helpful starting point because in that land ownership structure, keeping the allocation of
case farm-size inequality, in itself, does not affect average groundwater xed, must be disentangled from its effects
utility (equivalently, it has no direct effect on total utility, on the equilibrium average pumping rate and the spatial
which is taken here to be the measure of social welfare). distribution of groundwater withdrawals across farmers.
However, as shown in the following, the common Here, it is possible that small farmers actually pump less
property nature of the resource creates differing incentives in the rst period than large famers. This will occur if the
to pump water across size classes, so that an increase in utility functions are sufciently concave, so that small
inequality may either amplify or moderate the common farmers (who have lower incomes) face a greater
property externalities and social welfare may either rise or differential between marginal utilities of present and
fall. future income, and therefore, have a greater incentive to
In the linear utility case, the basic intuition is that large save groundwater for future use. With this as an additional
farms have greater spatial extent of resource access or determinant of pumping rates, the results discussed in the
ownership, so that they perceive the resource as being previous continue to apply, however.
more private. By the same token, a small farmer This paper may contribute to the continuing debate on
effectively owns a smaller share of the aquifer, and the magnitude of the welfare difference between optimal
perceives groundwater as a more common resource. control rules and competitive outcomes (Gisser 1983,
Therefore, smaller farmers tend to pump faster. In the Gisser and Sanchez 1980; Koundouri 2004). Provencher
aggregate, more water is always withdrawn in the rst and Burt (1993) identify three sources of inefciency
period compared to the efcient solution (the tragedy of associated with groundwater use in agriculture: stock,
the commons still applies), but the magnitude of over- pumping cost, and risk externalities. In the presence of
pumping depends on the inequality in land holdings. In an user heterogeneity, an access inequality externality is
alternative distribution of farm sizes with greater inequal- added to this list. The access inequality externality arises
ity, aggregate pumping in the rst period may change in when the rates of groundwater extraction differ across
either direction depending on the nature of the change in farms of varying size overlying a common aquifer. This
the distribution. Aggregate withdrawals increase if land externality can be both positive and negative, depending
area is shifted towards small farmers, but the converse on whether smaller farms appropriate, on a per unit land
holds if acreage is shifted towards large farmers. The area basis, a greater share of the common resource. Small
direction of the change is shown to depend on specic and large farmers can be thought of as, respectively, low
curvature properties of the production function relating and high income groups. And so, a common resource such
agricultural output to irrigation. as groundwater may become a natural vehicle for income
A separate but related question is how greater transfer, and can either neutralize or amplify income
inequality in farm sizes affects social welfare. The model inequality caused by the inequality in farmland holdings.
reveals that there are dynamic as well as spatial This paper also analyzes the effects of a specic but
components determining this effect. The dynamic compo- commonly implemented water-management policy, name-
nent refers to the effect of farm-size inequality on ly pumping quotas, on the distribution of income across
aggregate withdrawals in the rst period, or the speed farm-size classes. Using an example of a at-rate quota

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


837
policy, policy-induced gains and losses are shown to be This paper captures some of the same effects
unequally distributed across farmers. In general, the through a simple model where wells of equal depth
results suggest that the interactions between policies are already in place and each farmer faces an
addressing farmland ownership structures and groundwa- irrigation-application-rate decision. A two-period
ter management should not be ignored. An effort to reduce framework with a quasi-bathtub aquifer is particular-
inequities may worsen the common property problem, ly well suited to fully work out the equilibrium effects
while efforts to reduce the common property problem may of farm-size inequality on the welfare difference
cause greater inequities. Of course, the directions of these between the competitive and efcient allocations. By
impacts may be the opposite so that the policies are assuming an initial stock that is scarce enough to
mutually reinforcing. However, careful empirical analysis impose tradeoffs between the two periods, both the
that differentiates farmers production relationships across pumping cost externality and stock externality naturally
size classes (e.g., Sekhri 2011) is required to determine the arise in the model, which are then either amplied or
nature of the interactions moderated by the farm-size inequalities. The pumping
cost and stock externalities are the costs that one user
imposes on others through higher future pumping costs
Literature review and reduced groundwater availability, respectively.
Knapp and Vaux (1982) and Feinerman (1988) are among Following Gisser and Sanchez (1980), groundwater
the few studies that consider equity and distributional economic studies in multiperiod settings typically
effects of groundwater management schemes. Knapp and consider only the pumping cost externality; Provencher
Vaux (1982) consider groups of farmers differentiated by and Burt (1993, 1994) are notable exceptions.
their derived demand for water, and present an empirical Given the seasonality of production in irrigated
example that demonstrates that some users may suffer agriculture, a groundwater resource can be regarded as
substantial losses from quota allocation policies even a quasi-bathtub with features of a common property
though the group as a whole benets. Feinerman (1988) resource over time. The quasi-bathtub property means
extends their analysis and considers a variety of manage- that the resource at each extraction point is private
ment tools including pump taxes, quotas, subsidies, and within each period, but the aquifer becomes a
markets for water rights. Using simulations calibrated to bathtub or purely common pool across periods. This
Kern County, California (USA), Feinerman concludes that happens when the time period during which ground-
while the welfare distributional effects on user groups may water is extracted is relatively short, and does not
be substantial, the negotiations between the policy-makers allow for seepage from one point in the aquifer (such
and the users are likely to be difcult because the as a well or a pool) to another. However, the water
attractiveness of policies varies across users and is level tends to be more uniform throughout the aquifer
sensitive to the parameters. However, following Gisser in the long run. The quasi-bathtub assumption is
and Sanchez (1980), these studies ignore the stock appropriate if (1) the irrigation season is considerably
externality, and assume that under competition users shorter than the time that elapses between the two
behave myopically and base their decisions solely on the seasons, and (2) wells are spaced so that the localized
consideration of their immediate (periodic) prots. Also, cones of depression caused by pumping from neigh-
there is no investigation of the effect of the extent of user boring wells do not overlap within each irrigation
heterogeneity on the properties of competitive allocation. season.
There is a rather thin literature base in development The analysis also assume no time discounting, al-
economics that is concerned with the effect of inequality in though farmers time preferences of income are captured
land holdings on groundwater exploitation. Motivated by the in the concave utility model. These assumptions ensure
role of groundwater in sustaining the Green revolution and that the results are not an artifact of any other source of
developing agrarian economies, Foster and Rosenzweig spatial or temporal heterogeneity other than that intro-
(2008) consider the patterns of groundwater extraction in duced by size inequality. However, the main insights and
rural India. They develop a dynamic model of groundwater policy implications obtained in this framework carry on to
extraction that captures the relationships between growth in more realistic settings.
agricultural productivity, the distribution of land ownership, From here, the paper presents a simple two-period
water-table depth, and tubewell failure. Using data on model of groundwater extraction in the presence of farm-
household irrigation assets including tubewell depth as a size heterogeneity. The social planners solution is
proxy for irrigation intensity, they nd that large landowners considered. Then the paper analyzes the equilibrium
are more likely to construct tubewells, but their tubewells allocation and the effect of farm-size inequality on the
tend to be less deep than those dug by smaller landowners. pumping rates and farm income when farmers marginal
Foster and Rosenzweig conclude that this is indicative of a periodic utility of income is constant. Consideration is
free-riding effect in the sense that large farmers are less able given to equilibrium allocation when farmers marginal
to effectively poach the water from neighboring farmers by periodic utility of income is decreasing. Lastly, before the
lowering the water-table under their own lands. They also conclusions, consideration is given to a at-rate quota
nd evidence of land consolidation as a way to improve policy that illustrates political economy issues that arise in
efciency of groundwater exploitation. the presence of user heterogeneity.

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


838

Model Aquifer
The total stock of groundwater stored in the aquifer in the
For simplicity, the model focuses on the stock, cost, and beginning of period 1 is x1=Ah1, where h1 is the height of the
access inequality externalities. It considers the decisions water table in period 1, and A is the size of the area measured
of water application per acre taking the distribution of in acres (1 acre = 0.4047 ha). Let L = {1,,A} denote the set
irrigated acres across farmers as exogenous. With slight of acres. The hydraulic heads of the water table under each
modications, the model can be extended to include acre are the same in the beginning of each period, hi,t = hj,t = ht
decisions about the share of farm acreage allocated to i, j L and t = 1,2. Let ui,t denote the quantity of groundwater
irrigated crops. Farmers are identical except for the applied in period t on acre i. By the quasi-bathtub assumption,
distribution of land ownership, and irrigation technology the per acre quantity of groundwater withdrawn in each period
is constant returns to scale. All prots are derived from cannot exceed the per acre stock or ht
agricultural outputs using groundwater for irrigation on a ui;t  ht for all i 2 L and t 1; 2: 1
xed land area, and farmers hold exclusive pumping rights
P
on their land. The individual groundwater stocks are Let u1 A1 A i1 ui;1 denote the average pumping in
private during each irrigation season because there is no period 1. Since there is no recharge, the stock of
intra-seasonal well interference. However, the groundwa- groundwater in the aquifer in period 2 is x2 = x1 Au1,
ter is an inter-seasonal common property resource based and the level of the water table is
on the groundwater hydrology over a longer time interval.
The following assumptions are standard (e.g., Negri h2 h1  u1 : 2
1989):

1. Fixed land ownership. The distribution of farmland


ownership does not change over time. Land ownership
2. Constant returns to scale and homogenous land There are n farmers (users of groundwater) who are located in
quality. The agricultural production function has the the area overlying the aquifer and grow irrigated crops.
property of constant returns to scale (output is Farmer k farms acres Lk L, and let Ak jLk j denote the
proportional to farm size). Land quality is identical number of irrigable acres owned by farmer k, where
Pn
k1 Ak A. In what follows, the set of acres Lk will be
across all farms. Inputs other than groundwater,
including the choice of irrigation technology, fertilizer, referred to as farm k or farmer k. For concreteness, the
crops, etc., are optimized conditional on the rate of farm indices are assumed to be ordered by farm size, A1
water extraction. Output and input prices, including A2 An. Throughout, the rst symbol in doubly
energy costs, are exogenous. subscripted variables identies the acre and the second
3. Pumping cost. The total cost of groundwater extraction identies the period, t = 1,2. Variables with one subscript
per acre increases with the pumping rate and decreases typically refer to the aggregate values in the specied period,
with the level of the water table (or the stock of unless they are farm-specic and invariant across periods. The
groundwater). letters i, j will index acres, and letters k,l will index farmers.
4. User location is irrelevant. The aquifer is conned,
non-rechargeable, homogenous, and isotropic. The
groundwater basin has parallel sides with a at bottom. Production technology
5. Quasi-bathtub. There are no intra-seasonal lateral ows The periodic per acre benet of water consumption net of
of groundwater across farms. However, inter-seasonal all costs including groundwater pumping cost is
changes in groundwater level are transmitted instanta-  
neously to all users (i.e., the groundwater has an g ui;t ; ht ; 3
innite rate of transmissivity during the time elapsed
from one irrigation season until next). Brozovic et al where g is strictly increasing and concave. While irrigation
(2003) provide a detailed discussion of the consequen- increases yield, a higher groundwater stock decreases the cost
ces of this assumption. of pumping due to a decrease in pumping lift, and increases
6. Two periods. There are only two periods (irrigation the efciency of irrigation by permitting a more exible
seasons), and farmer preferences over income are application schedule. Land quality is assumed to be homoge-
additively separable across periods. neous so that total farm income is proportional to farm size
(i.e., technology exhibits constant returns to spatial scale). For
Provencher and Burt (1994) and Saak and Peterson simplicity, the rainfall and surface water supply are the same
(2007) also consider and provide justications for a two- on all farms in both periods. For example, Eq. (3) can take the
period framework. The assumption that the aquifer is non- following form:
renewable is for expositional convenience, and a positive
rate of recharge can be easily incorporated. The ground- gu; h max z pyu; h; z  cu; h  qz;
water extractions are the net quantity of water withdrawn
if some fraction of the water percolates back to the stock. where p is the per unit price of the crop, y is yield, and c is
Next the model notation is introduced. the cost of pumping groundwater, z is the vector of other

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


839
00
inputs, and q is the price vector of other inputs. For if either farmers have linear utility, v 0, or acreage
notational convenience, let is uniformly distributed across farmers, Ak = A/n for
k = 1,n
f h gu h; h gh h; h 4 2. Characterized by smaller farmers pumping groundwa-
ter faster, usj;1  usi;1 , for j Lk, i Ll, k < l, if v00  0
denote the marginal per acre benet of water consumption
(decreasing marginal utility of income)
evaluated at the point of depletion of an individual groundwater
stock. (Here and throughout, subscripts on functions denote Equation (7) is easiest to interpret for the special case
differentiation with
0
respect to the lettered arguments.) By when the water benet depends only on water use, u. In
concavity of g, f h < 0 8h 2 0; h1 . All of the results that this case, it is efcient to equalize the
follow will also hold under weaker technical conditions, namely  marginal
 benets of

0
guu < 0, ghh < 0, and f h guu h; h ghh h; h  2guh h; h < 0, water use in the two periods: gu ui;1 gu h1  usi;1 ,
s

which are implied by concavity of g. which implies that usi;1 h1 =2 8i 2 L. This is equivalent
Let v denote the periodic utility of farm income, to the assertion that, in the absence of a pumping cost
0 00
v > 0; v  0. Each farmer maximizes the sum of utilities externality and inequality of income across farmers, the
of the whole-farm revenue in each period: efcient solution distributes the available water equally
X X   across the two periods on each farm.
p k max v g ui;t ; ht subject to 1 and 2: 5
fui;t gi2L
k
t1;2 i2Lk It is convenient to differentiate between the case when
farmers per period marginal utility of income is (1)
constant (i.e., utility is linear), and (2) decreasing (i.e.,
For simplicity, there is no discounting of future income. utility is concave). In the former case, from the social
planners point of view, a non-uniform distribution of
acreage across farmers has no effect on either the optimal
allocation of water either spatially or temporally. Howev-
Social planner er, as demonstrated in the next section, such differences
may still arise in competitive equilibrium. In the latter
Before turning to the analysis of the competitive allocation
case, as is demonstrated in part 2 of Proposition 1, the
by non-cooperating users, the efcient allocation n isorst social planner faces a trade-off between dynamic and
characterized. The social planner chooses usi;t to distributional sources of inefciencies.
maximize producer welfare conditional on the land From a policy perspective, an important insight of the
ownership distribution: analysis to follow is that, in the presence of farmer
heterogeneity, competitive allocations go beyond the
X Xn X  
W max us 
s
v g ui;t ; ht
s
6 tragedy of the commons, and affect income inequality as
i;t
t1;2 k1 i2Lk well. The welfare difference between the optimal and
subject to 1 and 2: competitive allocations may be particularly large, when,
from the societal point of view, the income distribution
matters. This happens when the equilibrium distribution of
The following result shows that the efcient allocation of pumping rates across heterogeneous farmers amplies the
groundwater compensates for income inequality caused by income inequality caused by size inequality. However, the
the inequality in farm sizes. The common resource may competitive allocation may also moderate the inherent
serve as a vehicle to decrease income inequality by inequality in income distribution caused by the inequality
redistributing income from larger farmers to smaller farmers. in land ownership, or even change its sign, whereas total
This effect is absent if either farm sizes are identical, or incomes over two periods earned by smaller farmers
farmers periodic utility functions are linear in income. Note exceed that of larger ones.
that optimal groundwater consumption in the nal period
exhausts the remaining stock on each farm, and hence, must
be identical on all acres, usi;2 usj;2 h2 8i; j 2 L, because Linear utility
the income utility and water benet functions are strictly
increasing. And so, the focus is solely on period 1 pumping. This section considers the case of linear utility functions,
All proofs that are not in the text are in the Appendix.
v 0. The competitive equilibrium is rst characterized,
00

followed by an analysis of the effect of inequality in farm


Proposition 1. (Efcient pumping) Efcient allocation of
sizes on the groundwater stock and the distribution of
groundwater is:
income.
1. Invariant across acres, usi;1 usj;1 8i; j 2 L, and is
determined by Equilibrium
Farmers are non-cooperative, and each farmer takes the
quantity of water pumped by others in each period as
gu usi;1 ; h1  f h1  usi;1 0; 7 given. In period 2, all farmers exhaust the available stocks

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


840
of groundwater on each acre, so that ui;2 h2 for 8i 2 L. that the per acre irrigation application rates be identical
By Eq. (5), in period 1 farmer ks payoff is when farmers have linear utility.
X Under linear utility, smaller farmers always deviate
p k max fui;1 gi2L gui;1 ; h1 gh2 ; h2 8 more from the socially efcient allocation. However, it is
k i2L k not clear whether the non-uniformity of the distribution of
land ownership, in and of itself, leads to a loss or gain of
subject to 1 and 2:
total farm income. As shown in the next section, the
effects of the inequality in farm sizes on the groundwater
The competitive allocation can now be characterized. stock and farm income depend on rather subtle properties
Differentiating Eq. (8), the best response by farmer k on of the agricultural production function.
acre i 2 Lk , ui;1 , satises
 
gu ui;1 ; x  ak f h2 0; if ui;1  h1 ; and 9 Inequality in farm sizes
The measure of inequality that is used to model an
ui;1 h1 ; if otherwise increase in the concentration of land ownership (a smaller
share of farmers owns a larger share of land) is introduced
where ak Ak =A is the share of the aquifer that can be next. The rest of this section analyzes the effect of
captured by farmer k. Equation (9) can be written in a inequality in farm sizes on the remaining groundwater
more compact form stock and on total income. An example is presented that
 illustrates the ndings.
ui;1 min h1 ; gu1 ak f h2 ; h1 ; 8i 2 Lk 10

where g1u ; h is the inverse of gu(u,h) obtained by Measuring inequality


treating h as a parameter. Note that per acre pumping To model the effect of increased inequality in land
rates on each farm are identical ui;1 uj;1 8i; j 2 Lk . holdings a precise measure of inequality is needed. The
Summing pumping rates (Eq. 10) over all k = 1,n and i analysis here relies on the Lorenz measure, which is widely
Lk, and substituting Eq. (2), yields
!
used to measure wealth inequality more generally. Let W
Xn      W1 ; :; Wn denote a vector of wealth (in this paper, wealth is
u1 ak min h1 ; g 1
u ak f h1  u 
1 ; h1 ; 11 measured by the area of land owned) by n individuals, where
k1 P
P W1 W2 Wn and nk1 Wk  W . The Lorenz
where u1 1=A A 
i1 ui;1 is the equilibrium average
!
! P
pumping in period 1. By concavity of g, Eq. (11) uniquely measure of W is dened as l l=n; W lk1 Wk =W ;
determines the aggregate pumping in period 1, u1 . its interpretation is the share of land held by the smallest 100
Together Eqs. (10) and (11) prove the existence and
!
(l/n)% of farmers. If W is a perfectly equal wealth
uniqueness of equilibrium. distribution (i.e., Wk W=n8k), then the Lorenz function
is linear in x = l/n with a slope of 1; for all other distributions
Proposition 2. (Competitive allocation) Suppose that it is a (weakly) convex curve that never lies above this line.
farmers utility is linear in income. Competitive equilibrium In general, increasing inequality implies more curvature of
exists, it is unique, and is given by Eqs. (10) and (11). The the Lorenz curve, so that the value of 1 at a given value of x
average pumping rate is higher than the socially efcient will be smaller.
average rate, u1  us1 . Also, smaller farmers pump faster The effect of inequality in farm size is modeled by
than larger farmers, ui;1  uj;1 , for any i Lk, j Ll, k < l. comparing the equilibrium under the given distribution of
land holdings, A1 A2 PAn, to an alternative 
Comparing the rst-order conditions that characterize distribution, B1  B2  :::  Bn nk1 Bk A . Where
the efcient and competitive allocations, Eqs. (7) and (9), ! !
distribution B is more unequal distribution A based on the
respectively, shows that the discrepancy between them
arises along both spatial and temporal dimensions. That Lorenz measure: . The proofs of
is, the competitive allocation leads to an inefciently high several of the propositions below rely on the majorization order,
aggregate pumping in period 1, which entails an inef- a general tool to compare the dissimilarity within the
cient allocation of groundwater across periods. Nonethe- components of vectors that is closely related to the Lorenz
less, it is possible that individual farmers extract measure. Marshall and Olkin (1979) provide a comprehensive
groundwater at a slower rate than the socially efcient treatment of majorization.
average rate, i.e. ui;1  us1 for some i (see section Small
and large farms: an example and Fig. 1b). Also, unless all ! !
Denition. Real vector A is majorized by B , denoted
farmers are identical, the competitive allocation results in ! m! P l P
A  B , if A  lk1 Bk for l =1,,n, and
inefcient pumping rates across farmers in period 1. Pn Pn k1 k
Recall that, by Proposition 1, part 1, efciency requires k1 Ak k1 Bk

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


841

Water stock (acre-feet)


Pumping (acre-feet) 0.95 Per acre pumping rate on
small farms
0.85

0.75

0.65 Per acre pumping rate on large


farms
0.55
0.5 Efficient per acre pumping
0 80 160 240 320 400 480 560 640 720 800 880 960 0 160 320 480 640
Farm size difference (acres) Farm size difference (acres)
a) b)
1.134 1550
Income (US$ / acre)

1450 Total profit of a large farm

Income (US$)
1.132 1350
1250
1.13
1150
1050
1.128
950
1.126 850
750 Total profit of a small farm
1.124 650
0 160 320 480 640 800 960 0 160 320 480 640 800 960
Farm size difference (acres) Farm size difference (acres)
c) d)
Fig. 1 Inequality in farm sizes, pumping rates, and income. a Per acre pumping rates. b Groundwater stock in period 2. c Average income
per acre. d Income for small and large farms (1 acre = 0.4047 ha = 4,047 m2)

Thus, the comparison of interest can be expressed as the P guuu u; h1 =guu u; h1 denote the index of concav-
! !
majorization A m B . A related notion of Schur-concave ity of the marginal output function of a farmer with
and Schur-convex functions technology g (u,h1) in period 1. If g (u,h1) were a utility
! will also be needed. A real-
of income function, then R would be interpreted as the
valued function y A is called Schur-concave if
! ! ! Arrow-Pratt coefcient of absolute risk aversion, and P
! m! would be the coefcient of absolute prudence.
A  B implies y A  y B , and y A is Schur-
! As g represents technology and not preferences in the
convex, if y A is Schur-concave. Schur-concavity might model here, these indexes are employed simply as
be more intuitively called Schur-monotonicity be- measures of the curvature of the physical relation between
cause it simply requires function y to always decrease output and water. In this non-stochastic framework, they
in response to a perturbation that induces more are indicators of the strength of the motive to smooth
dissimilarity in its arguments. The Lorenz function water extraction over time (i.e., the diminishing marginal
itself is an example of a Schur-concave function. The productivity of water). Adding uncertainty will not change
analysis to follow will appeal to the following the qualitative nature of the results. There is an empirical
important property of Schur-concave functions. Sup- literature on the relationship between farmers risk
! P ! preferences and their dynamic use of groundwater (e.g.,
n
pose that y A k1 zAk . Then y A is Schur- Antle (1983, 1987) and Koundouri et al. 2006) as well as
concave if and only if z is concave. on the effects of risk preferences on farmers reaction to
water quota policies (e.g., Groom et al. 2006).

Measuring concavity
The analysis that follows will also depend on the Inequality of farm sizes and groundwater stock
curvature properties (specically the degree of concavity) With the aforementioned denitions, the relationship
of the agricultural production function, g. Even though between inequality and the residual water stock in period
there is no uncertainty in this model, it is convenient to 2 can now be analyzed.
derive its results using well-known measures of curva-
ture from the literature on decision-making under Proposition 3. Suppose that farmers utility is linear in
uncertainty. Let R guu u; h1 =gu u; h1 denote the income. Then under more unequal distribution of farm
! !
index of concavity of agricultural output function, and sizes, A m B , the groundwater stock in period 2

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


842
! ! !
1. Increases, h2 A  h2 B , if 2R P; where h2 h1  u1 is given by Eq. (11), and W c A
~  h B
2. Decreases, h2 A ~ , if symbolizes the dependence of total farm income (agricul-
2
tural output) on the distribution of land ownership among
~ , i.e. the smallest farm
a) B1 =A gu h1 ; h1 =f h2 A farmers.
under the new land ownership distribution is not The farm-size inequality affects both the groundwater
too small and stock in period 2 (dynamic allocation) and the distribution
b) 2R P of groundwater application rates across farms in period 1
(spatial allocation). Keeping everything else equal, a
The inequality in land ownership creates a trade-off in more stable inter-seasonal pattern of groundwater use
terms of its effect on the pumping decisions in period 1. A increases total farm income. The distributional effect of
heavier left tail of the acreage distribution implies that farm-size inequality on farm income is more difcult
there are more farmers who own a smaller share of the because a higher variability in farm sizes may or may not
aquifer and tend to pump faster than the average farmer. lead to a higher variability in the per acre pumping rates
However, a heavier right tail implies the opposite. (see Proposition 3).
Therefore, ascertaining the effect of any increase in
acreage inequality on the competitive allocation requires Proposition 4. Suppose that farmers utility is linear in
structure on the farm-size sensitivity of the difference in income. Then under more unequal distribution of farm
pumping rates between small and large farmers, ui;1  uj;1 , ! !
sizes, A m B , total farm income
where i Lk, j Ll, Ak < Al. The farm-size sensitivity of
the difference in pumping rates across farms is ! !
00 0
1. Decreases, W c A  W c B , if
ak u ak =u ak , where uak g1
u ak f h2 ; h1 < h1 . If a) 3R ! P and !
0
the pumping rate differential, u , is increasing (decreas- b) h2 A  h2 B
ing), the sensitivity is negative (positive). ! !
Condition 1 states that, when the aquifer is full, the 2. Increases, W c A  W c B , if
agricultural output, g (u,h1), is in a sense more concave a) The smallest farm under the new land ownership
!
than the marginal output, gu u; h1 . Then the perceived distribution is not too small, B1 =A  gu h1 ; h1 =f h2 A
benet from a more stable inter-seasonal groundwater use b) 3R P
! !
pattern increases with size at an accelerating rate, and a c) h2 A  h2 B
greater inequality stimulates, on average, a slower pump-
ing rate. Note that condition 2 R () P is equivalent to Conditions in part 1 guarantee that the unequal distribution
log-concavity (log-convexity) of the rst derivative of the of farm acreage aggravates both the distributional 1(a) and
demand for water with respect to output when the aquifer dynamic 1(b) inefciencies that are associated with the
is full, g1 1 competitive allocation. Condition 1(a) requires that the net
y y; h1 , where g y; h1 fu : y gu; h1 g is
the inverse of agricultural output function obtained by benet of irrigation when the aquifer is full, g (u,h1), is in a
treating the stock of groundwater, h1, as a parameter. sense more concave than the marginal benet, g (u,h1). Then a
greater inequality in farm sizes stimulates a greater variability
To guarantee that the average pumping rate increases,
in (acreage-weighted) pumping rates and lowers total output.
the additional condition a) in part 2 is needed because the
Observe that 1(a) is less stringent than part 1 in Proposition 3.
aquifer is a quasi-bathtub (see Eq. 1). This condition puts
This is because the net benet of irrigation, g (u,h1), is concave
a limit on the increase in the size of large farms. It implies
in u, which adds additional curvature, and thus, on average, a
that, under the new distribution of land ownership, the smaller (or positive) farm-size sensitivity of the spatial
number of farmers who grow irrigated crops is the same, pumping rate differential sufces to cause a total output loss.
B1 > 0, and that, under the initial distribution of land Part 2 has a similar interpretation. Condition 2(a) is the
ownership, no farmer depleted his/her stock of ground- same as in Proposition 3. But now sufcient condition 2
!
water in period 1, ui;1 A < h1 for all i L1, where 1 is (b) is more stringent compared with 2(b) in Proposition 3.
the index of the smallest farmer. This is because a negative and sufciently large (in
absolute value) farm-size sensitivity of the spatial pump-
ing rate differential is required in order to assuredly raise
total output. Note that condition 3R () P is equivalent
Farm-size inequality and farm income to concavity (convexity) of the rst derivative of the
The effect of farm-size inequality on total farm income is inverse output function (i.e., demand for water as a
now considered. In the case of linear utility, Eq. (6) becomes function of output) when the aquifer is full, g1 y y; h1 .
! Xn Xn
Combining Propositions 3 (part 2) and 4 (part 1) yields
Wc A p
k1 k
A
k1 k
          Corollary. Suppose that farmers utility is linear in income.
g min h1 ; gu1 ak f h2 ; h1 ; h1 g h2 ; h2 ; Then under more unequal distribution of farm sizes,
! m! ! !
12 A  B , total farm income decreases, W c A  W c B ,
if 2R P 3R.

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


843
s 1:8
Sufcient conditions under which more unequal distri- per acre is W =A 10  0:2 . At = 0 (i.e., s = l = 1,000),
bution of farm sizes has an unambiguously positive effect small and large farms are the same, and the distribution
on total farm income cannot be obtained in this way. To of land ownership is uniform across farmers. The effects
guarantee a lesser inequality in pumping rates, the of an increase in farm-size inequality on the equilibrium
0
pumping rate spatial differential, u ak , must be suf- groundwater stocks, pumping rates, and incomes are
ciently decreasing (in absolute value) with farm size. In analyzed next.
contrast, to guarantee a more stable average pumping rate, As shown in Fig. 1a, when the difference in farm sizes
the pumping rate spatial differential must be increasing or is relatively small, D  280, the difference in the pumping
slightly decreasing (in absolute value) with farm size. rates increases until the small farmers deplete their wells
Furthermore, as clear from the proof of Proposition 4 in period 1, ui;1 h1 1 for i 2 Lk and k 1; :::; 50. This
(see Eq. 21), the sign of k/Ak is ambiguous. Therefore, limits the ability of small farmers to steal groundwater
it is possible that smaller farmers earn more total income from their neighbors, and therefore, establishes an upper
than larger farmers, k l for k < l. Of course, larger bound on the difference in the pumping rates. Curiously,
farmers always have higher total revenues in period 2. the large farmers pump less than the efcient quantity,
However, smaller farmers have more intensive-margin ui;1  0:5h1 0:5 for i 2 Lk and k 51; :::; 100, when
operations and higher per acre revenues in period 1. The
D 2 220, 400]. In this range, the gain in the dynamic
differential in total revenues between small and large
efciency for the large farmers outweighs the loss
farmers in period 1 can be positive, and even exceed the
associated with letting the small farmers steal their
magnitude of the negative differential in total revenues in
groundwater. However, as the size of each large farm,
period 2. Intuitively, smaller farmers will earn higher
and hence the total share of the aquifer farmed by large
prots from being in a better strategic position to take
farms, increases, large farmers are able to more effectively
advantage of the common property resource; they are able
push the aggregate groundwater use towards the
to steal more groundwater per unit of land than their larger
efcient allocation. Even though the incentive to pump
neighbors. The following example illustrates.
groundwater efciently for each individual large farmer
declines, the aggregate groundwater usage in period 1
decreases. This is because the distribution of total acreage
Small and large farms: an example is skewed more (less) heavily towards large (small)
00
Let g(u, h) = (u + z)y, (0,1), z 0.5h1, and v 0. By farmers, who pump slowly (who deplete their wells in
Proposition 1, the efcient allocation of groundwater period 1).
across acres and seasons is invariant to the distribution Figure 1b illustrates the non-monotone relationship
of land ownership, and is given by usi;1 0:5h1 for i L. between the stock of groundwater in period 2 and
The maximal regional farm income is Ws = 2A(0.5h + z)y. farm-size inequality. As explained earlier, when the
For simplicity, all farms fall in one of the two categories: gap between small and large farms is small,
small and large. The size of small farms is s acres, Ak=s for k = D 2 0; 280, the large farmers are relatively ineffective
1,,m, and the size of large farms is l acres, Ak = l for k = in raising the dynamic efciency. This is because, even
m + 1,n, where s l. The number of small farms is m, though they decrease their pumping rates in order to
and the number of large farms is n m, where ms + (n m) compensate for the higher pumping rates by small
l = A. By Eqs. farmers, their weight in aggregate pumping is relative-
h (10)
 
and (11),
 equilibrium  pumping
 in period 1
 1=g1 i
is ui;1 min h1 ; As 1=g1 h1  Ez1E
1E
z 1  As ly light. And so, the negative effect of the aggressive
pumping by small farms dominates, and the ground-
h1 smum;1 =Azl=A1=1g 1
for i Lk, k = 1,,m, ui;1 l=A1=1g lnm=A
for i water stock in period 2 falls. As the share of total
Lk and k = m + 1,,n where E ms=A g=g1
acreage owned by small farmers declines, but their
pumping rates remain constant (ui;1 h1 1 for i 2 Lk
n  ml=Ag=g1 . and k 1; :::; 50), the large farmers need to give up
For concreteness, consider a special case of an increase less of period 1 pumping to push the region towards
in farm-size inequality, whereas small farms get uniformly more dynamically efcient allocation. From the per-
smaller and large farms get uniformly larger. Note spective of a large farmer, the groundwater resource is
! 0 0 ! 00 00 0 00
more private, which reinforces the diminished inuence
that A s ; m; ls m A s ; m; ls for s > s , where of aggressive pumping by small farmers. As a result,
ls A  ms=n  m. Clearly, a uniform shift of the average stock in period 2 increases, and the region
acreage from small farms to large farms, keeping the number moves towards a more dynamically (and spatially)
of farms in each size category xed, constitutes an increase efcient allocation.
in farm-size inequality. Inequality can then be measured Figure 1c shows the non-monotone effect of the
simply as the gap between the acreage on small and large inequality in farm sizes on total income. Proposition 4
farms, = l s 0, keeping the number of each type of shows that, in general, an increase in size inequality
farms, m, xed. affects the total farm income in two distinct ways. First, it
In Fig. 1, parameters are: = 0.8, z = 0.3, n = 100, m = affects the groundwater stock in period 2. Second, it
50, h1 = 1, and A = 100,000. Then the maximal farm income affects the variability of the pumping rates among farmers

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


844
in period 1. When the gap is small, D 2 0; 280, both Proposition 5. Suppose that farmers utility is strictly
the stock and pumping rate variability effects work concave in income. Then the average pumping rate is
in the same direction. When the gap is sufciently higher than the socially efcient average rate, u1  us1 ,
large, any further increase in farm-size inequality raises and for all i 2 Lk , j 2 Ll , k < l
the total farm income. Note that the dip in the total
income in Fig. 1c has a rather pointed peak. This is 1. Smaller farms pump faster, ui;1 uj;1 , if r0 0
because for D  280 there is an additional income gain 2 . S m a l l e r f a r m s p u m p s l o w e r, ui;1  uj;1 , i f
associated with the gain in the spatial efciency due to 1 rv1 0
1 av ahA  rahA 8a 2 ak ; al  a n d
the decline in the heterogeneity of pumping rates. The h 2 0; 0:5h1
period 1 pumping on large farms increases, while
pumping on small farms remains constant (as they Farm size has two effects on the farmers pumping
deplete their wells in period 1). decision. On the one hand, larger farmers view their
As shown in Fig. 1d, total per farm incomes are stock of groundwater as a relatively more private
also non-monotone in the extent of farm-size inequal- resource. This provides them with a greater incentive
ity. Surprisingly, the total small farm income to push the regional use towards a dynamically more
increases when the acreage on small farms decreases efcient allocation. On the other hand, larger farmers
in the range D 2 0; 280. The converse holds for large may have a smaller (negative) difference in marginal
farms. This is because small farms are in a better utilities of income in periods 1 and 2. This diminishes
position to steal groundwater from their neighbors their incentive to push the region towards a dynamically
operating on large farms. However, the cap on the more efcient allocation compared with smaller farm-
pumping in period 1, ui;1  1, eventually annuls this ers. The private resource effect dominates if the
effect. Consequently, a further increase in farm-size coefcient of relative risk-aversion is increasing in
inequality affects farm incomes in the expected income. The income scale effect dominates if the
direction because, keeping everything else equal, a coefcient of relative risk-aversion is sufciently
smaller (larger) acreage entails a smaller (larger) large and decreasing in income (in the sense of
whole-farm income. condition in part 2).
While not reported here due to space constraints, the
counterparts of Propositions 34 carry over to the case of
Concave utility concave utility as well. Competitive allocations may either
exacerbate or alleviate income inequality associated with
So far, the analysis has considered the effect of farm-size the distribution of land holdings among farmers. If the
heterogeneity on welfare in the case of farmers with linear coefcient of relative risk-aversion is increasing in
utility functions (constant marginal utility of income). As income, small farmers pump more groundwater per acre
shown next, relaxing this assumption may lead to rather than large farmers. This lessens the income inequality
different conclusions. Even the result that smaller farmers caused by an unequal distribution of acreage. The converse
pump faster under the competitive allocation may no is true if larger farmers pump more aggressively (on a per
longer hold. This section considers the case of farmers acre basis), which is possible if the coefcient of relative
with (strictly) concave per period utility functions, v00 < 0. risk-aversion is sufciently large and decreasing.
To highlight the role of concavity of utility, prot per unit Note that, in the absence of the effect of farm-size
of land area (e.g., yield) is now assumed to be a linear inequality on the disaggregated pumping rates, from
function of the amount of water applied per acre, and that the societal point of view, the heterogeneity in land
pumping costs do not depend on the hydraulic head, holdings is immaterial if farmers are risk-neutral (i.e.,
gu; h u. they value marginal income in both periods indepen-
Following the same steps as before, it can be shown dently of the number of acres they farm). When
that the equilibrium best response of farmer k on acre farmers are risk-averse, the heterogeneity in the
i 2 Lk ; ui;1 , is pumping rates can be welfare-increasing, given that
the per acre irrigation rates increase on smaller farms
ui;1 minh1 ; 1=Ak v1 0 
1 ak v Ak h1  u1 ; 8i 2 Lk 13 and decrease on larger ones, so that in period 1
income is redistributed from rich to poor farmers (see
where v1 0
1 : is the inverse of v , and the average pumping
Proposition 1). However, because of the decreasing
in period 1, u1 , solves marginal per acre benets of water, total income
always decreases under a greater variability of the
Xn pumping rates. This may create a tension between the
u1 1=A minAk h1 ;v1 0 
1 ak v Ak h1  u1  14
k1 effects of farm-size inequality on income distribution
and total income (output). The next section takes a
Let ru uv00 u=v0 u denote the Arrow-Pratt coef-
policy perspective and investigates the workings of a
cient of relative risk-aversion of a farmer with the periodic very simple groundwater use policy in the presence of
utility of income v. farmer heterogeneity.

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


845
Policy analysis: an example of at-rate quota On the other hand, from Eq. (19) it follows that the income of
policy large farmers, who are not bound by the quota, increases
because the quota policy slows down the average pumping
The analysis now considers some political economy rate in period 1.
aspects of implementing a simple policy that allocates Let mq supfk : ak  gu q; h1 =f h2 q; 1  k  ng.
per period per farm pumping quotas. Suppose that the Note that m(q) is a non-increasing function. Then farmers
policy takes the form k = 1,, m(q) are bound by the quota in period 1. Also,
X X farmers k 1; :::; mq h1 deplete their wells in period 1,
ui;1  Ak q and ui;2 15 where q h1 symbolizes the absence of the quota policy.
i2Lk i2Lk
X
 Ak q maxAk q  i2Lk
ui;1 ; 0 for k 1; :; n; Proposition 6. Suppose that the quota is applicable,
u1;1 q h1 > q0 . Then under the groundwater quota
where q 2 0; h1  is the per acre quota (measured in acre- policy q q0 < h1
feet), and the quota allocated to each farm is proportional
to its size. The quota limits the quantity of groundwater 1. The groundwater stock in period 2 increases, h2 q
extracted in each period, but allows farmers to carry over h1 < h2 q q0 8q0 < h1 . Suppose that the period 2
unused portions of their quota into the next period. There quota is not binding, q0 h1 =2. Then
is no market for water rights, and the unused quotas 2. Large farmers gain, pk q h1  pk q q0 for
cannot be bought or sold.
k mq0 1; :::; n;
For concreteness, the case of risk-neutral farmers and a
3. Small farmers lose, p k q h1  pk q q0 for
strictly concave agricultural output function (analyzed in
section Linear utility) is considered. The following result k 1; :::; mh1 , if
establishes that, while this policy always slows the rate of
the aquifer depletion, the effect on farmer incomes is a) guuu  0, guuu 0, 2guh h; h ghh h; h  0
likely heterogeneous. The setting is assumed to be such Pz1 Pn
that the equilibrium pumping rates decrease with time b) az k1 ak = kz1 a2k for all z mh1 ; :::; mq0
ui;1  ui;2 8i 2 L, so that u1  0:5h1  u2 . For example,
Farmers in the medium size range, mh1  k  mq0 ,
this is always true if all farmers are sufciently small relative may lose or gain from a quota. The intuition for this result
to the aquifer, an inf u20;h1 fgu h1  u; h1 =f h1  ug. is very clear: Small farmers, who pump faster than the
Then, under the quota policy of Eq. (15), farmers do not average farmer, stand to lose the most from a quota policy.
transfer the unused portion of their quotas from period 1 to Large farmers, who are not restricted by the policy, strictly
period 2: q  ui;1  ui;2 , if q  h1 =2, and ui;1 ui;2 q gain from the quota because of the more stable inter-seasonal
8i 2 L if q < h1 =2. Hence, for q  h1 =2 equilibrium is allocation of groundwater induced by this policy.
given by This illustrates that policies that do not account for user
heterogeneity, are likely to affect not only the inter-
seasonal but also the spatial distribution of incomes
ui;1 q minq; gu1 ak f h1  u1 q; h1  16
among farmers. The ensuing political economy issues
8i 2 Lk ; k 1; :; n; and the relative weight of small and large farmers in the
policy-making process pose additional constraints on the
Xn design of efcient groundwater management policies.
u1 q k1
ak minq;gu1 ak f h1  u1 q; h1  17

The income of farmer k under the quota policy is Conclusions and policy implications

p k q Ak fgq; h1 gq; h1  qg; if q < h1 =2; and 18 This article has analyzed the economic inefciencies that
arise when farmers controlling operations of varying sizes
withdraw irrigation water from a common aquifer. Farm-
p k q Ak fgminq; gu1 ak f h1  u1 q; h1 ; h1 size inequality was shown to affect the degree of
inefciency because small farmers are more strongly
g h1  u1 q; h1  u1 qg; if q h1 =2:
inuenced by common property externalities than large
19 farmers, who have an incentive to internalize inter-well
costs within their operations. This insight alone has the
From Eq. (18) it follows that all farmers lose (gain) policy implication that the gains from groundwater
from a more restrictive quota, if the initial quota is management are likely to be greater in regions populated
sufciently small and the marginal benet of a higher by small farms such as in developing nations.
stock is small (large) relative to the marginal The overall effect of an increase in inequality on social
benet of water consumption: @pk q=@q Ak fgu q; h1 welfare was shown to be ambiguous and dependent on the
gu q; h1  q  gh q; h1  qg  0 for all k = 1,,n. agricultural production function as well as on the

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


846
differences in marginal utility between large and small Proof of Proposition 2
farmers. To the extent that these relationships vary across Suppose that us1 > u1 . Then, by Eq. (11)
regions, it is one explanation for wide gaps in the
prosperity of groundwater-dependent agricultural regions. 
Pn 1 
Sufcient conditions were established to identify the u 1 k1 ak minh1 ; gu ak f h1  u1 ; h1 
P
cases where increased inequality reduces aggregate welfare, nk1 ak minh1 ; gu1 ak f h1  us1 ; h1 
and these conditions which appear to be quite restrictive. P
This nding suggests that in many regions, there is a nk1 ak minh1 ; gu1 f h1  us1 ; h1 
meaningful, if not recognized, policy tradeoff between gu1 f h1  us1 ; h1 us1 :
common property distortions and inequality. Wealth dispar-
ities within the farm population are a concern in both high The inequalities follow by concavity of g. The equality follows
and low income countries, particularly as it relates to the by Eq. (7); thus, a contradiction was obtained. Also,
incomes of small farmers (Hoppe et al. 2010). However, in u minh1 ; g1 ak f h2 ; h1  minh1 ; g1 al f h2 ; h1  u
i;1 u u j;1
the case of access to a common aquifer, a reduction in for any i 2 Lk ; j 2 Ll ; k < l
inequality may have the unintended effect of accelerating the
depletion of the resource. Moreover, the analysis reveals that
the common aquifer can, in effect, become a conduit to Proof of Proposition 3
transfer income from large to small farmers. Part 1. Suppose that h2 A~ > h B ~ . Then, by Eq. (11),
Finally, water-management policies designed to correct 2

common property externalities were demonstrated to have  ! Xn !


potentially signicant and undesirable distributional u1 A a minh1 ;gu1 ak f h1  u1 A ; h1 
k1 k
impacts. In particular, it was shown that a quota policy
may well reduce the speed of aquifer depletion as Xn !
b minh1 ; gu1 ; bk f h1  u1 A ; h1 
k1 k
intended, but the welfare gains from groundwater conser-
vation will not be evenly distributed; in general, irrigators in Xn !
certain size classes will incur welfare losses. bk minh1 ; g 1 bk f h1  u B ; h1 
k1 u 1

!
u1 B
Appendix
The rst inequality follows because the sum of compositions
Proof of Proposition 1 of two concave functions (here minak h1 ; ak g1
u ak f :; h1 ),
First, note that in period 2, the planner optimally exhausts is Schur-concave in a1,,an. To show this, it must be
the remaining stock on each farm because g and v are demonstrated that ag1u af is concave in a. Differentiating
strictly increasing. This implies that Eq. (1) binds for twice yields
t 2i:e:; usi;2 h2 8i 2 L, so that Eq. (6) can be written
Xn X @ 2 agu1 af  f
W s max fusi;1 g v gusi;t ; ht vAk gh2 ; h2 : 2Ru  Pu  0;
k1 i2L k @a2 Ruguu u; h1

P where the inequality follows by condition 1 stated in


Because i2Lk gusi;t ; ht is symmetric and concave in Proposition 3. The second inequality follows by concavity
usi;1 , and W5 is symmetric in v(.), optimality requires that of g. And so, a contradiction was obtained.
 ! !
s s
ui;1 uj;1 for any i 2 Lk and j 2 Ll if Ak Al . Addition- Part 2. Suppose that h2 A < h2 B . Then, by Eq. (11),
ally, corner solutions are ruled out because v and g are u ! Xn !
A a g 1 ak f h1  u1 A ; h1
increasing and concave in each argument. The rst-order 1 k1 k u
conditions for a maximum are Xn !
 k1
bk gu1 bk f h1  u1 A ; h1
f h1  us1 h i
v0 Ak gusi;1 ; h1 gu usi;1 ; 1  20 Xn
1  !
Xn
A k1
b k min h 1 ; g u b k f h 1  u 1 A ; h 1
A v0 Al gh1  us1 ; h1  us1 0; Xn h i
l1 l 1  !
 k1
bk min h 1 ; gu b k f h 1  u1 B ; h1
if usi;1  h1 , and usi;1 h1 , otherwise, for all i 2 Lk and
!
k = 1,,n. Part a follows by observingPn that Eq. (20) reduces to u1 B :
00
Eq. (7) when v 0 because l1 Al A. Part 2 follows by
observing that only the rst term in Eq. (20) depends on farm
size Ak, and, by concavity of utility function, v, it decreases The equalities follow because, by condition 2(a) in the
with Ak. Then by concavity of yield function, g, this implies statement ! of Proposition 3 and concavity of g,
s  !
that ui;1 is a non-increasing function of farm acreage. gu ak f h2 A ; h1  h1 a n d g1
1 
u bk f h2 A ; h1  h1

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


847
! !
for all k = 1,,n, since A m B implies a1  b1 . The rst Part 2. By condition 2(a) in the proposition statement,
inequality follows because, by condition 2(b) in the ! ! !
Pn 1 ui;1 A < h1 for all i 2 L because A m B implies
statement of Proposition 3, a g u ak f h1   !
k1 k
 ! that a1  b1 so that g1 u ak f h2 A ; h1  h1 and gu
1
u1 A ; h1 is Schur-convex (see part a). The second
!
equality follows by assumption. And so, a contradiction bk f h2 A ; h1  h1 for all k = 1,, n. Then, by Eq. (12),
was obtained.
! Xn !
W c A A fggu1 ak f h2 A ; h1 ; h1
k1 k

! !
gh2 A ; h2 A g
Proof of Proposition 4
To show parts 1 and 2, we need two facts. Xn !
Fact 1. (1) p k ak Aak gminh1 ; uak  is concave  k1
Bk fggu1 bk f h2 A ; h1 ; h1
in ak when 3RP. ! !
(2) pk ak juak <h1 is convex in ak when 3R P, where gh2 A ; h2 A g
 ! Xn !
uak g1
u ak f h2 A ; h1 Bk fgminh1 ;gu1 bk f h2 A ; h1 ; h1 
k1
Proof of Fact 1. To verify, differentiate twice
with respect to ak=a: ! !
gh2 A ; h2 A g
Xn !
 Bk fg minh1 ; gu1 bk f h2 B ; h1 ; h1
@p k a @agua; h1  af 2 k1
A Agu; h1 ; and
@a ua<h1 @a guu u; h1 ! ! !
gh2 B ; h2 B g W c B :
21
The rst inequality follows because function
!
@ p k a
2
@ 2 agua; h1  WA is Schur-convex by Fact 1(2). The equality follows
A
@a 2
ua<h1 @a2 by condition 2(b) in the statement of Proposition 4. The
af 2 second inequality follows by condition 2(c) in the
A 3Ru  Pu  0:
Ruguu u; h1 proposition statement and Fact 2.
22
depending on whether 3R  P. This proves Fact 1(2). Proof of Proposition 5
To show Fact 1(1), note that ak gminh1 ; uak  Suppose that us1 u1 . Then, by Eqs. (20) and (14) in the
minak gh1 ; h1 ; ak guak ; h1  by monotonicity of g. text,
Hence, ak gminh1 ; uak  is concave in ai when 3R P
P Pn
as a composition of concave functions. us1 1=A Pnk1 minAk h1 ; v1 0
1 Pl1 al v Al h1  u1 
s
Fact 2. @W c =@h2 > 0 < 1=A Pk1 minAk h1 ; v1 l1 al v Al h1  u1 
n 1 n 0
Proof of Fact 2. @W c =@h2 inherits the sign of  1=A nk1 minAk h1 ; v1 0  
1 ak v Ak h1  u1  u1
@fgg1 0
u af h2 ; h1 ; h1 gh2 ; h2 g=@h2 af h2 =guu u; h1
f h2 > 0, where the inequality follows by concavity of g.
The inequalities follow by concavity of v. And so, a
Keeping everything else equal, as the extent of dynamic
contradiction was obtained.
inefciency of the competitive allocation increases, Part 1. Let i 2 Lk . First, consider ui;1 Ak < h1. By
welfare falls. Eq. (13), differentiation yields
Part 1. By Eq. (12),
! P ! @ui;1 =@Ak v0 Ak h2 =Ak v00 Ak ui;1 1 RAk ui;1
W c A nk1 Ak fg minh1 ; gu1 a k f h2 A ; h1 ; h1
! !  RAk h2  0
gh A ; h2 A g
Pn 2 !
k1 Bk fgminh1 ; gu1 bk f h2 A ; h1 ; h1 
! ! The inequality follows because, by Eq. (13), ui;1  h1  u1,
gh A ; h2 A g and so 1 RAk ui;1  RAk h1  u1  1 > 0. If ui;1
Pn 2 !
k1 Bk fg minh1 ; gu1 bk f h2 B ; h1 ; h1 h1 then uj;1  h1 for j 2 Ll ; k < l
! ! ! Part 2. Proof is analogous.
gh2 B ; h2 B g W c B :
!
The rst inequality follows because function WA is
Schur-concave as the sum of concave functions by
condition 1(a) in the statement of Proposition 4 and Fact Proof of Proposition 6
1(1). The second inequality follows by condition 1(b) in Part 1. Note that this is trivially true when the quota
the proposition statement and Fact 2. is binding in period 2, q0 < h1 =2, because then ui;1 q,

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


848
and ui;2 h2 h1  q 8i 2 L. So consider the case since, by 3(1),
when q0 h1 =2 and suppose that u1 q h1 <
u1 q q0 . Then, by Eq. (17), f 0 h1  u1 guu h1  u1 ; h1  u1 2guh h1  u1 ; h1  u1

ghh h1  u1 ; h1  u1  guu ui;1 ; h1


Xn
u1 q h1 ak minh1 ; gu1 ak f h1  u1 q h1 ; h1 
Xk1
n
The third inequality in Eq. (23) follows by 3(2). Hence,
ak minq0 ; gu1 ak f h1  u1 q h1 ; h1  p k q q0  p k q h1 for k 1; :::; mh1 because
Xk1

n
ak minq0 ; gu1 ak f h1  u1 q q0 ; h1  @p k q=@q  0 for all q 2 q0 ; h1 
k1
u1 q q0 ;

where the last inequality follows by concavity of g. And


so, a contradiction was obtained.
Part 2. By Eq. (19), farmer ks income for k mq0
1; :::; n is References
Antle J (1983) Testing the stochastic structure of production: a
p k q q0 Ak fggu1 ak f h1  u1 q0 ; h1 ; h1 exible moment-based approach. J Bus Econ Stat 1:192201
Antle J (1987) Econometric estimation of producers risk attitudes.
Am J Agric Econ 69:509522
gh1  u1 q0 ; h1  u1 q0 g Brozovic N, Sunding D, Zilberman D (2003) Optimal manage-
ment of groundwater over space and time. In: Berga D,
Ak fggu1 ak f h1  u1 h1 ; h1 ; h1 Goetz R (eds) Frontiers in water resource economics. Kluwer, New
York
Brozovic N, Sunding D, Zilberman D (2010) On the spatial nature
gh1  u1 h1 ; h1  u1 h1 g of the groundwater pumping externality. Resour Energ Econ 32
(2010):154164
p k q h1 ; Feinerman E (1988) Groundwater management: efciency and
equity considerations. Agric Econ 2:118
Feinerman E, Knapp K (1983) Benets from groundwater manage-
where the inequality follows by part a, and monotonicity ment: magnitude, sensitivity, and distribution. Am J Agric Econ
and concavity of g. 65:703710
Part 3. By Eq. (19), farmer ks income is p k q0 Foster A, Rosenzweig M (2008) Inequality and the sustainability of
Ak fgq0 ; h1 gh1  u1 ; h1  u1 g f o r k 1; :::; agricultural productivity growth: groundwater and the green
revolution in rural India, presented at the India Policy
mh1 . Differentiation yields Conference at Stanford University, California, June 13,
2005
Gaudet G, Moreaux M, Salant S (2001) Intertemporal depletion of
@p k q0 resource sites by spatially distributed users. Am Econ Rev
Ak fgu q; h1  f h1  u1 91:11491159
@q Gisser M (1983) Groundwater: focusing on the real issue. J Polit
Econ 91:10011027
@u1 Gisser M, Sanchez D (1980) Competition versus optimal control in
 g Ak f h1  u1 famq0 groundwater pumping. Water Resour Res 16:638642
@q Groom B, Koundouri P, Nauges C, Thomas A (2006) Stochastic
@u1 technology and risk attitudes in agricultural production. Appl
 g Ak f h1  u1 famq0 Econ 40:315326
@q Hardin G (1968) The tragedy of the commons. Science 162:1243
Pmq0 1248
Hoppe R, MacDonald J, Korb P (2010) Small farms in the United
P l1 al
 g 0: 23 States: persistence under pressure. Economic Research Service,
1 nlmq0 1 a2l EIB-63, United States Department of Agriculture, Washington,
DC
The rst inequality follows because mh1  mq0 , which Knapp K, Vaux H (1982) Barriers to effective groundwater
management: the California case. Ground Water 20:6166
follows by concavity of g. The second P 0
inequality
Pn
follows Koundouri P (2004) Current issues in the economics of
because, by Eq. (17), u1 q q mq l1 al lmq 0 1 al g1
u groundwater resource management. J Econ Surv 18(5):703

al f h1  u1 q; h1 , and implicit differentiation yields 740
Koundouri P, Nauges C, Tsouvelekas V (2006) Technology
adoption under production risk: theory and application to
Pmq0 irrigation technology. Am J Agric Econ 88:657670
@u1 q l1 al Marshall A, Olkin I (1979) Inequalities: theory of majorization and
Pn its applications. Academic, New York
@q 1 lmq0 1 a2l f 0 h1  u1 q=guu ui;1 Al ; h1 Negri D (1989) The common property aquifer as a differential
Pmq0 game. Water Resour Res 29:915
a Provencher B, Burt O (1993) The externalities associated with the
 Pnl1 l 2 ;
1 lmq0 1 al common property exploitation of groundwater. J Environ Econ
Manage 24:139158

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2


849
Provencher B, Burt O (1994) A private property rights regime for Sekhri S (2011) Public provision and protection of natural
the commons: the case of groundwater. Am J Agric Econ (76 resources: groundwater rirrigation in rural India. Am Econ
November): 875888 JAppl Econ (3 October): 2955
Saak A, Peterson J (2007) Groundwater use under incomplete Xabadia A, Goetz R, Zilberman D (2004) Optimal dynamic pricing
information. J Environ Econ Manage 54:214228 of water in the presence of waterlogging and spatial heteroge-
Schaible G (2004) Irrigation, water conservation, and farm size in neity of land. Water Resour Res 40:W07S02. doi:10.1029/
the western United States. Amber Waves (8 June): 8 2003WR002215

Hydrogeology Journal (2012) 20: 835849 DOI 10.1007/s10040-012-0854-2

Potrebbero piacerti anche