Professional Documents
Culture Documents
KMasuda SI
KMasuda SI
Kazuto Masuda
Author Note
Tel: 81-3-3279-111
E-mail: kazuto.masuda1@gmail.com
Abstract
We incorporate production into Becker and Murphy’s (2000) social interaction model,
based on Kiley (2003). We create a particular utility function for the concrete
satisfying the strict quasi-concavity. When solving our model, we use the Karush-Kuhn-
Tucker theorem, satisfying some general requirements discussed by Fuente (2000) and
specifying its domain. Our solution with this particular utility function coincides with
We provide a particular utility function for the social interaction model developed
by Becker and Murphy (2000). Then, we analyze the behavior of the equilibrium in
comparison with that of a standard business cycle model without this particular utility
function based on Kiley (2003). Furthermore, we calculate the social multiplier effect
under this general equilibrium model. My model is static at first, but we show that our
model could be extended into the dynamic model easily. Additionally, we show the first-
order conditions of our heterogeneous agent model. Finally, we show the existence of
the unique global optimum in this social interaction model with production, using the
Sundaram (1996), Fuente (2000), Jehle and Reny (2001), and Osborne (2022).
This particular utility function is based on the Cobb–Douglas utility function but
has additional utility from the behaviors of other consumers in the same community to
which the consumer in question belongs. As a result, this utility function includes the
particular case of social capital, that is, social capital with the perfect depreciation as the
In addition to Becker and Murphy’s (2000) model which analyzes the social
multiplier effect in the consumer’s partial equilibrium with many goods, other models of
1
social interaction exist, including Brock and Durlauf’s (1995), which is a dynamic
model focusing on choice, such as logit. Brock and Durlauf (1995) pointed out the
existence of multiple equilibria and got different conclusions from those reached in this
paper, although they used a different model from the one applied in this paper (i.e., that
example. Although Aiyagari (1994) did not argue for the uniqueness of the optimum in
his model, Achdou et al. (2022) discussed the uniqueness of the optimum in Aiyagari-
type models. For numerical nonlinear programing, an example is the work of Tamura
(1990) is known to guarantee the uniqueness of the optimum. To secure the unique
we extend the mathematical result shown in Dixit (1990), Sundaram (1996), Fuente
(2000), Jehle and Reny (2001), and Osborne (2022) leading to the concave
programming. Note that Becker and Murphy (2000) were not interested in the
uniqueness of the optimum in their model, at least explicitly, because their interest was
mainly in the multiple equilibria and the separating equilibrium in various social
interaction models.
2
Moreover, in this paper, we focus on consumers’ demand behavior. In this kind of
model, which assumes the existence of social capital, (a) in the utility function, the
choice modeling has already been discussed by Brock and Durlauf (1995). Furthermore,
(b) (depending on the definition of social capital) growth theorists in the late 1980s and
1990s also intensively discussed the firm production model that assumes the existence
of social capital in the production function rather than in the utility function (e.g., Barro
Function for the Social Interaction Model With Production section, we discuss a
particular utility function that satisfies the conditions for confirming the social
multiplier effect described by Becker and Murphy (2000), after introducing the general
equilibrium model with the standard Cobb–Douglas utility function and no physical
accumulation based on Kiley (2003). Then, using this particular utility function, we
construct our general equilibrium model without physical capital accumulation based on
Kiley (2003) and derive its (unique global as we discuss below) optimum together with
this production function under the Karush-Kuhn-Tucker theorem at the same time
(Karush, 1939; Kuhn & Tucker, 1951). Moreover, we show that the presence of
3
this section, usually increases (decreases) consumption under some conditions, by
further reducing (or further increasing) leisure, compared with the case without the
our particular utility function and our constraints, and confirm the existence of the
Heterogeneity section, we briefly discuss the heterogeneity of consumers and its effect
on our conclusion about the unique global optimum. we conclude that heterogeneity in
our particular setting does not alter our conclusion about the unique global optimum in
briefly discuss our alternate assumption about physical capital accumulation in the
kind of Lagrangian and easily extend our model into the dynamic setting with physical
how to obtain analytical solutions, reviewing the Kiley’s (2003) solution technique,
based on Bénassy (1995). In Appendix B, we briefly show the effects of the signs of
4
A Particular Utility Function for the Social Interaction Model With Production
Based on Kiley (2003), we consider the social planner problem under certain
resource constraints and production technologies, like the following production function
for the case of total physical capital depreciation. This model is almost identical to the
standard business cycle model often discussed in macroeconomics, only without the
physical capital accumulation. More precisely, we assume the investment at the current
textbooks (see Appendix A for the physical capital accumulation case with
Hamiltonian).
Therefore, for the time being, the following model is essentially static because it
does not consider physical capital accumulation (on this point, I benefited from a
discussion with Professor Akihiko Kaneko. Again, we can extend our model into the
𝛽
subject to 𝑌𝑡 ≥ 𝑃𝐶,𝑡 𝐶𝑡 + 𝑃𝐼,𝑡 𝐼𝑡 , 𝑌𝑡 = 𝑍𝑡 𝐼𝑡 (1 − 𝐻𝑡 )1−𝛽 , 𝐶𝑡 , 𝐼𝑡 ≥ 0, 𝐻𝑡 ∈ [0,1],
𝐷 ∈ (0, ∞), and 𝛽 ∈ [0,1] are given parameters; (𝐶𝑡 , 𝐼𝑡 ) is a convex set on ℝ2≥0 ;
5
𝐻𝑡 ∈ [0,1] is a convex set; and 𝑍𝑡 is a convex set on ℝ+ and is given together with
the utility function of the consumer (in the above case, we assume the Cobb–Douglas
𝑃𝐶,𝑡 is the real price of consumption (goods) at period 𝑡 (standardized by the price of
production [goods]); and 𝑃𝐼,𝑡 is the real price of investment (goods) at period 𝑡 (also
In this standard model, the consumer obtains positive utility from consumption
and leisure and experiences a trade-off between consumption 𝐶𝑡 and leisure 𝐻𝑡 . When
both consumption 𝐶𝑡 and leisure 𝐻𝑡 in this standard model double simultaneously, the
consumer’s utility also doubles. However, the increase of marginal utility gradually
which all individual consumers are identical, there is no need to treat each consumer
separately, such as consumer 𝑗, for all the variables in the model including production
6
there are 𝑁 consumers in our model, we do not explicitly use the subscript 𝑗 to
indicate the individual variables our model uses unlike Heterogeneity section.
production 𝑌𝑡 . When the investment 𝐼𝑡 and labor input 1 − 𝐻𝑡 double, the production
𝑌𝑡 also doubles. However, if only one of the production factors (investment or labor
input) increases, the production 𝑌𝑡 increases, but each input’s increase of marginal
In addition, if the real prices 𝑃𝐶,𝑡 , and 𝑃𝐼,𝑡 are assumed to be one (i.e., the same
economics dealing with multi-goods models, we assume that the goods baskets are
different between consumption and investment, and therefore the real price 𝑃𝐶,𝑡 is
naturally different from 𝑃𝐼,𝑡 . The deflators for consumption 𝐶𝑡 and investment 𝐼𝑡 are
Unlike the standard macroeconomic models, our social planner problem does not
explicitly deal with the firm’s profit maximization problem. However, if we make the
7
perfect competition assumption, under the firm’s profit maximization assumption, the
factors 𝐼𝑡 and 1 − 𝐻𝑡 and thus satisfies one of the constraints of the above social
planner problem with equality, 𝑌𝑡 = 𝑃𝐶,𝑡 𝐶𝑡 + 𝑃𝐼,𝑡 𝐼𝑡 as the firm’s profit maximization
production. Therefore, we follow Kiley (2003) and effectively use his formulation that
if 𝐻𝑡 ∈ (0,1),
𝛽 1 −𝛽
(1−𝛽)𝑌𝑡 𝛽𝑌𝑡
𝐶𝑡 = , 𝐼𝑡 = 𝑃 , 𝑌𝑡 = 𝛽 1−𝛽 1−𝛽
𝑍𝑡 1−𝛽
̅.
𝑃𝐼,𝑡 (1 − 𝐻𝑡 ), 𝐻𝑡 = 1 − 𝛼 = 𝐻
𝑃𝐶,𝑡 𝐼,𝑡
(i) 𝐶𝑡 = 0, 𝐼𝑡 = 0, 𝑌𝑡 = 0, if 𝐻𝑡 = 1.
𝛽 1 −𝛽
(1−𝛽)𝑌𝑡 𝛽𝑌𝑡 1−𝛽 1−𝛽
(ii) 𝐶𝑡 = , 𝐼𝑡 = 𝑃 , 𝑌𝑡 = 𝛽 1−𝛽 𝑍𝑡 𝑃𝐼,𝑡 , if 𝐻𝑡 = 0.
𝑃𝐶,𝑡 𝐼,𝑡
investment 𝐼𝑡 , depending on the real prices 𝑃𝐶,𝑡 , and 𝑃𝐼,𝑡 , are fixed fractions of
8
production 𝑌𝑡 , whereas leisure 𝐻𝑡 is a function of the parameter of the utility function
function. Leisure 𝐻𝑡 increases only when the exponent for consumption in the utility
function decreases. Additionally, if we consider the corner solutions, these optima are
trivial. However, one corner solution with the fixed leisure at 𝐻𝑡 = 0 mostly maintains
As we will show later, the above equations of the interior solution and one corner
on 𝑃𝐶,𝑡 , and 𝑃𝐼,𝑡 , the real prices of consumption goods and investment goods. The
that of Nishimura (1990), except that this model is a general equilibrium model that
considers production.
From the above interior analytical solution, we also obtain the relation 𝑌𝑡 =
𝛽 1 −𝛽
1−𝛽 1−𝛽
𝛽 1−𝛽 𝑍𝑡 𝑃𝐼,𝑡 𝛼. Therefore, in this general equilibrium, which also considers changes in
decrease, so that utility 𝑉 will decrease. On the other hand, an increase in the real price
𝑃𝐼,𝑡 decreases production 𝑌𝑡 (due to the price effect), which lowers the level of
consumption 𝐶𝑡 , and utility 𝑉 decreases. Conversely, if the real prices 𝑃𝐶,𝑡 and 𝑃𝐼,𝑡
9
decrease, utility 𝑉 increases along the same path. This effect is the price effect in this
standard model and maintained in the interior solution and one corner solution at 𝐻𝑡 =
0.
Note that, unlike the following particular utility function case, we can measure
the welfare effects clearly because the leisure remains constant and only the
Next, we take up our social interaction model, which is based on Becker and
equilibrium in which all 𝑁 consumers are homogeneous, the particular utility function
Note that we discuss only the 𝐴, 𝐵 ∈ (0, ∞) case in the following paragraphs and
sections. This is for the convenience of our discussions. we can extend our case into the
first general case graphically in detail in Appendix B. Regarding the latter two cases,
10
these are the trivial particular cases of our 𝐴, 𝐵 ∈ (0, ∞) and do not explicitly analyze
them in this paper (for these cases, we just set 𝐴 = 0 or 𝐵 = 0 for the following
equations).
The above symmetric equilibrium assumption implies that for every consumer 𝑗
𝑗 𝑗
in a community, 𝐶𝑡 = 𝐶𝑡 , and 𝐻𝑡 = 𝐻𝑡 . Therefore, 𝐴𝐶𝑡 and 𝐵𝐻𝑡 , the linear terms in
the above particular utility function 𝑈, also represent the social capital accumulated
through consumption 𝐶𝑡 and leisure 𝐻𝑡 (i.e., the average consumption and leisure of
all consumers in society). Thus, again, this extension model is also the representative
consumer model.
In this paper, if there are 𝑁 consumers in the whole society, the per capita social
capital 𝐾𝑙,𝑡 , 𝑙 = 1,2 accumulates with the standard per capita capital formation
𝑗
∑𝑁
𝑗=1 𝑥𝑙,𝑡 𝑗 𝑗 𝑗 𝑗
equation 𝐾𝑙,𝑡 = + (1 − 𝛿𝑙 )𝐾𝑙,𝑡−1 , where 𝑥1,𝑡 = 𝐶𝑡 , 𝑥2,𝑡 = 𝐻𝑡 , 𝑡 is time, and
𝑁
𝑙 = 1,2, 𝛿𝑙 denotes the social capital depreciation rate for each good. However, in the
𝑗
∑𝑁
𝑗=1 𝑥𝑙,𝑡
standard per capita social capital accumulation equation as 𝐾𝑙,𝑡 = .
𝑁
𝑗 𝑗
that for all consumers 𝑗 in the society, 𝐶𝑡 = 𝐶𝑡 , and 𝐻𝑡 = 𝐻𝑡 again. Thus, by
substituting this relationship into the standard per capita social capital formation
11
equation, we obtain 𝐾1,𝑡 = 𝐶𝑡 , 𝐾2,𝑡 = 𝐻𝑡 , which are expressed in the above linear
Moreover, we note from this result that the linear terms 𝐴𝐶𝑡 and 𝐵𝐻𝑡 also
generate the relationship that as social capital increases, consumer utility also increases.
Although Becker and Murphy (2000) initially assume the existence of social capital
such as a religion, their model can also represent the broader social capital, such as
community and connections created by human networks (both of the above two points
are owed to Professor Kuniyoshi Saito’s suggestions). For this particular case in which
we fully amortize social capital each period, these linear terms represent the effect of
Finally, our model also represents the network externality defined by Egashira et
al. (1995), given Egashira et al.’s (1995) definition that “‘Network externality’ refers to
the case in which an individual’s utility from the demand for a good or service depends
not only on the use-value of the good or service itself but also on the extent to which
others demand the same good” (translated by author. This point is also owed to
The characteristics of this particular utility function 𝑈 are the additional linear
terms about consumption 𝐶𝑡 and leisure 𝐻𝑡 (the terms that generate the social
12
multiplier) to the utility function 𝑉 of the previous standard model. In other words,
consumers have linear terms in which their utility always increases in a fixed ratio as
the average consumption and leisure of other consumers (i.e., their own consumption
and leisure from the symmetry assumption) in society increases, as long as 𝐴 and 𝐵
are positive.
Therefore, the marginal utilities from the two goods, consumption 𝐶𝑡 and leisure
𝐻𝑡 , in our model are more significant than in the previous model. However, the second-
order differentiation of this utility function will be the same as in the previous model.
The decreasing degree of the marginal utility of the particular utility function 𝑈 is the
Using this particular utility function 𝑈, subject to the same constraints as for the
previous individual maximization problem when using the standard utility function 𝑉,
we solve this maximization problem using the Karush-Kuhn-Tucker theorem and obtain
𝛽 1 −𝛽
(1−𝛽)𝑌𝑡 𝛽𝑌𝑡 1−𝛽 1−𝛽
𝐶𝑡 = , 𝐼𝑡 = 𝑃 , 𝑌𝑡 = 𝛽 1−𝛽 𝑍𝑡 𝑃𝐼,𝑡 (1 − 𝐻𝑡 ), if 𝐻𝑡 ∈ (0,1).
𝑃𝐶,𝑡 𝐼,𝑡
̅
𝐻
of this function around 𝑋 = 1−𝐻̅ is
13
1−𝛼 1−𝐺𝑡 𝑋𝛼
̅ 1−𝛼−𝛼𝑋−𝐺𝑡 𝑋 𝛼 ∙
𝐻𝑡 𝐻 𝛼 1+𝐺𝑡 𝑋𝛼−1
− 1−𝐻̅ = 𝑋𝑡 − 𝑋 ≅ , ∴ 𝐻𝑡 = 1−𝛼 1−𝐺𝑡 𝑋𝛼
,
1−𝐻𝑡 𝛼+𝛼𝐺𝑡 𝑋 𝛼−1 1+ ∙
𝛼 1+𝐺𝑡 𝑋𝛼−1
𝛽(𝛼−1) 𝛼𝛽
𝛼−1 𝛼 1−𝛽
where 𝐺𝑡 = 𝐴𝐺1,𝑡 𝑃𝐶,𝑡 𝑃𝐼,𝑡1−𝛽 − 𝐵𝐺2,𝑡 𝑃𝐶,𝑡 𝑃𝐼,𝑡 , 𝐺1,𝑡 = 𝐷−1 (1 −
𝛽(1−𝛼) 1−𝛼 −𝛼𝛽 −𝛼
Moreover, the corner solutions are obtained in a manner similar to the former
problem:
(i) 𝐶𝑡 = 0, 𝐼𝑡 = 0, 𝑌𝑡 = 0, if 𝐻𝑡 = 1.
𝛽 1 −𝛽
(1−𝛽)𝑌𝑡 𝛽𝑌𝑡 1−𝛽 1−𝛽
(ii) 𝐶𝑡 = , 𝐼𝑡 = 𝑃 , 𝑌𝑡 = 𝛽 1−𝛽 𝑍𝑡 𝑃𝐼,𝑡 , if 𝐻𝑡 = 0.
𝑃𝐶,𝑡 𝐼,𝑡
For the interior solution, this leisure 𝐻𝑡 can be interpreted in various ways
𝜕𝑋𝑡
depending on the size of the given parameters. However, we know that > 0. When
𝜕𝐻𝑡
parameter 𝐴 is more significant than parameter 𝐵, the right-hand side of the above
̅ . However, if
approximation equation is always negative, and 𝐻𝑡 is always less than 𝐻
𝐴 becomes much more significant than 𝐵, 𝐻𝑡 mostly become the corner solutions.
the size of 𝛼, 𝐻𝑡 mostly becomes the corner solutions. Similarly, depending on the
14
̅ . In such a case, the behavior
size of 𝛼, 𝐻𝑡 might be more significant or smaller than 𝐻
of 𝐻𝑡 is uncertain. Note that in this paper, we assume that 𝐷, 𝛽, and other variables
Simply put, leisure will always decrease when the preference for consumption
based on a horizontal consciousness is higher than that for leisure to some extent.
much higher than that for leisure, leisure could become the corner solutions. In this
On the other hand, if the preference for leisure from a horizontal consciousness is
higher than that for consumption to some extent, leisure will always increase. However,
if the preference for leisure from a horizontal consciousness becomes much higher than
that for consumption, leisure might become the corner solutions. In this case, the
The above discussion can be visually confirmed as follows. Figure 1 and 2 show
the MATLAB simulation results. Again, the right-hand side of the above equation for
Figure 1
15
̅ Under 𝐴 = 0.001 and 𝐵 = 0.05 With the Variation of
Changes in 𝑋𝑡 − 𝑋, 𝐻𝑡 , and 𝐻
Figure 2
16
̅ under 𝐴 = 0.001
Figures 1 and 2 show the effect of 𝛼 on 𝑋𝑡 − 𝑋, 𝐻𝑡 , and 𝐻
and 𝐵 = 0.05 and under 𝐴 = 0.0001 and 𝐵 = 7.0, respectively. The left graphs show
𝑋𝑡 − 𝑋 on the vertical axis, with the given value of 𝛼 on the horizontal axis varying
with the given value of 𝛼 on the horizontal axis varying from 0.00 to 1.00. For other
variables and parameters, it is assumed that 𝑃𝐶,𝑡 , 𝑃𝐼,𝑡 , 𝐷 = 1, 𝑍𝑡 = 150, and 𝛽 = 0.3
for convenience. In this paper, all the assumed parameters are not rigorously based on
any empirical analysis but are chosen to illustrate the characteristics of our model
behavior clearly. Note that the effects of 𝐴 and 𝐵 on the simulation results are
17
asymmetric. In these parameter cases, the economic interpretations of the change in the
In this paper, we do not show the figure in the case that 𝐴 or 𝐵, or both are too
large. In the case where 𝐵 is too large, the monotonicity of the unilateral increase or
in both positive and negative regions and fluctuate greatly, and 𝐻𝑡 mostly shows the
corner solution such as 𝐻𝑡 = 1 (in the case where 𝐵 is far too large, 𝐻𝑡 mostly takes
the corner solutions). In this way, the left graph of figure in this case also shows some
the corners and similarly takes one corner solution or the other.
Note that the behaviors of 𝑋𝑡 − 𝑋 and 𝐻𝑡 in the case where 𝐴 and 𝐵 are too
large are similar to that in the case where 𝐴 is too large given other parameters as
shown, although dependent on the relative sizes of 𝐴 and 𝐵 because the effects of
Additionally, the above results are based on the assumption that 𝐴 and 𝐵 are
18
are unchanged (though the behavior of 𝑋𝑡 − 𝑋 does not show any spikes different from
the above). Again, the above parameter settings are not natural and not supported by
empirical analyses. We take up these parameters only to illustrate our model behavior.
too large, with 𝛼 being constant, it results in being much more significant or smaller
in the case where 𝐴 and 𝐵 are too large, the effect of 𝐴 might usually dominate that
case because we cannot define this case to divide the numerator by zero.
In this section, we show the social multiplier effect, as discussed by Becker and
Murphy (2000), with this paper’s simple general equilibrium model without physical
capital accumulation. First, the interior solution for the original utility function 𝑉
19
𝛽 1 −𝛽
(1−𝛽)𝑌𝑡 𝛽𝑌
𝐶𝑡 = , 𝐼𝑡 = 𝑃 𝑡, 𝑌𝑡 = 𝛽 1−𝛽 𝑍𝑡1−𝛽 𝑃𝐼,𝑡
1−𝛽
̅.
(1 − 𝐻𝑡 ), 𝐻𝑡 = 1 − 𝛼 = 𝐻
𝑃𝐶,𝑡 𝐼,𝑡
In this case, it is evident that the price 𝑃𝐶,𝑡 , and the price 𝑃𝐼,𝑡 only have direct
effects on the demands for consumption 𝐶𝑡 and investment 𝐼𝑡 , respectively, and are not
𝛽 1 −𝛽
1−𝛽 1−𝛽
𝐶𝑡 = (1 − 𝛽)𝑌𝑡 /𝑃𝐶,𝑡 , 𝐼𝑡 = 𝛽 𝑌𝑡 /𝑃𝐼,𝑡 , 𝑌𝑡 = 𝛽 1−𝛽 𝑍𝑡 𝑃𝐼,𝑡 (1 − 𝐻𝑡 ),
𝛽(𝛼−1) 𝛼𝛽
𝛼−1 𝛼 1−𝛽
where 𝐺𝑡 = 𝐴𝐺1,𝑡 𝑃𝐶,𝑡 𝑃𝐼,𝑡1−𝛽 − 𝐵𝐺2,𝑡 𝑃𝐶,𝑡 𝑃𝐼,𝑡 , 𝐺1,𝑡 = 𝐷−1 (1 −
𝛽(1−𝛼) 1−𝛼 −𝛼𝛽 −𝛼
1−𝛼 1−𝛽 −1 −𝛼 1−𝛽
𝛽) 𝛽 1−𝛽 𝑍𝑡 , 𝐺2,𝑡 = 𝐷 (1 − 𝛽) 𝛽 1−𝛽 𝑍𝑡 .
From the above, it is clear that under the particular utility function 𝑈 with the
social multiplier effect, the real price 𝑃𝐶,𝑡 and the real price 𝑃𝐼,𝑡 affect leisure 𝐻𝑡 , that
is, working hours 1 − 𝐻𝑡 , and through this effect, the above first two solutions about
consumption 𝐶𝑡 and investment 𝐼𝑡 will show more reduction than shown in the
previous standard model because 𝐴 and 𝐵 are assumed to be positive. The above
effect is called the social multiplier effect in the general equilibrium view (the partial
20
Looking more mathematically in depth, the real price 𝑃𝐶,𝑡 and the real price 𝑃𝐼,𝑡
1 𝛽 −𝛽
1−𝛽 1−𝛽 1−𝛽 1 𝛽 −𝛽
𝜕𝐶𝑡 𝜕[(1−𝛽)𝑍𝑡 𝛽 𝑃𝐼,𝑡 (1−𝐻𝑡 )/𝑃𝐶,𝑡 ] −1 𝜕𝐻
= = (1 − 𝛽)𝑍𝑡1−𝛽 𝛽 1−𝛽 𝑃𝐼,𝑡
1−𝛽
[− 𝜕𝑃 𝑡 𝑃𝐶,𝑡 − (1 −
𝜕𝑃𝐶,𝑡 𝜕𝑃𝐶,𝑡 𝐶,𝑡
−2
𝐻𝑡 )𝑃𝐶,𝑡 ] < 0,
𝜕𝐺 𝜕𝐺
− 𝑡 𝑋 𝛼 (1+𝐺𝑡 𝑋 𝛼−1 )−(1−𝐺𝑡 𝑋 𝛼 ) 𝑡 𝑋 𝛼−1
𝜕𝐻𝑡 1 1−𝛼 𝜕𝑃𝐶,𝑡 𝜕𝑃𝐶,𝑡
∵ 𝜕𝑃 = 2 𝛼−1 2 > 0,
𝐶,𝑡 (1+𝐺𝑇𝑡 ) 𝛼 (1+𝐺𝑡 𝑋 )
1−𝛼 1−𝐺𝑡 𝑋 𝛼
where 𝐺𝑇𝑡 = ,
𝛼 1+𝐺𝑡 𝑋 𝛼−1
𝛽(𝛼−1) 𝛼𝛽
𝜕𝐺𝑡 𝛼−2 𝛼−1 1−𝛽
= (𝛼 − 1)𝐴𝐺1,𝑡 𝑃𝐶,𝑡 𝑃𝐼,𝑡1−𝛽 − 𝛼𝐵𝐺2,𝑡 𝑃𝐶,𝑡 𝑃𝐼,𝑡 < 0, 1 − 𝐺𝑡 𝑋 𝛼 > 0.
𝜕𝑃𝐶,𝑡
1 𝛽 −1
1−𝛽 1−𝛽 1−𝛽
𝜕[𝛽𝑍𝑡 𝛽 𝑃𝐼,𝑡 (1−𝐻𝑡 )] 1 𝛽 −1 −1
𝜕𝐼𝑡 𝜕𝐻𝑡 1 −1
1−𝛽 1−𝛽 1−𝛽
= = 𝛽𝑍𝑡 𝛽 1−𝛽 [− 𝜕𝑃 𝑃𝐼,𝑡 − 1−𝛽 (1 − 𝐻𝑡 )𝑃𝐼,𝑡 ] < 0,
𝜕𝑃𝐼,𝑡 𝜕𝑃𝐼,𝑡 𝐼,𝑡
𝜕𝐺 𝜕𝐺
− 𝑡 𝑋 𝛼 (1+𝐺𝑡 𝑋 𝛼−1 )−(1−𝐺𝑡 𝑋 𝛼 ) 𝑡 𝑋 𝛼−1
𝜕𝐻𝑡 1 1−𝛼 𝜕𝑃𝐼,𝑡 𝜕𝑃𝐼,𝑡
∵ 𝜕𝑃 = (1+𝐺𝑇 )2 > 0,
𝐼,𝑡 𝑡 𝛼 (1+𝐺𝑡 𝑋 𝛼−1 )2
1−𝛼 1−𝐺𝑋 𝛼
where 𝐺𝑇𝑡 = ,
𝛼 1+𝐺𝑋 𝛼−1
𝛽𝛼−1 𝛼𝛽
𝜕𝐺𝑡 𝛽(𝛼−1) 𝛼𝛽 −1
𝛼−1 1−𝛽 𝛼 1−𝛽
= 𝐴𝐺1,𝑡 𝑃𝐶,𝑡 𝑃𝐼,𝑡 − 𝐵𝐺2,𝑡 𝑃𝐶,𝑡 𝑃𝐼,𝑡 < 0.
𝜕𝑃𝐼,𝑡 1−𝛽 1−𝛽
In short, when the real prices 𝑃𝐶,𝑡 and 𝑃𝐼,𝑡 increase, leisure 𝐻𝑡 always
price effects with those in the original utility function 𝑉 in detail, the price effects are
−1 𝜕𝐻 𝜕𝐶𝑡
amplified by the − 𝜕𝑃 𝑡 𝑃𝐶,𝑡 term on the right-hand side of in the case of a
𝐶,𝑡 𝜕𝑃𝐶,𝑡
−1
1−𝛽 𝜕𝐻
change in the real price 𝑃𝐶,𝑡 , and by the − 𝜕𝑃 𝑡 𝑃𝐼,𝑡 term on the right-hand side of
𝐼,𝑡
21
𝜕𝐼𝑡
in the case of a change in the real price 𝑃𝐼,𝑡 . Additionally, through those terms,
𝜕𝑃𝐼,𝑡
leisure 𝐻𝑡 is affected by prices 𝑃𝐶,𝑡 and 𝑃𝐼,𝑡 . These are the social multiplier effects.
In short, we can see that under a particular utility function 𝑈 with a horizontal
consciousness, in addition to the direct effect of the original utility function 𝑉 on the
amount of demand for consumption 𝐶𝑡 and investment 𝐼𝑡 , price 𝑃𝐶,𝑡 and price 𝑃𝐼,𝑡
induce an indirect effect that includes the so-called social multiplier in which prices
affect working hours 1 − 𝐻𝑡 through parameters such as 𝐴 and 𝐵 that indicate the
level of production 𝑌𝑡 , and this level of production 𝑌𝑡 will further affect consumption
𝐶𝑡 and investment 𝐼𝑡 .
Regarding the point that the prices affect leisure, this result is different from the
case of original utility function 𝑉 without the social multiplier effect. Note that,
different from Brock and Durlauf (1995), we cannot make definitive statements about
the welfare analysis of the price effects between the standard model and its extension,
that is, our social interaction model with production. This is because in our particular
utility function case, these price effects affect the utility, unlike in the standard utility
function 𝑉 case, where the leisure is always constant such that these price effects
22
information about them exists. For the same reason, we cannot infer and compare the
(concrete) magnitudes of the social multipliers themselves, different from Brock and
Durlauf (1995).
In summary, the existence of this social multiplier is rooted in the different setting
and equilibrium from those of the standard consumer model of mathematical economics
described by Nishimura (1990) and other numerous studies in the literature that when
prices change, the quantity demanded by other consumers has no effect on the quantity
demanded by respective individual consumers. In our model, like that of Becker and
Murphy (2000), when the price changes, the quantity demanded by the average member
of the community to which the consumer belongs will also affect the quantity demanded
Note that the corner solution cases do not admit the social multiplier effects
because the leisure and working hours are fixed at the corners of their domains and are
The Guarantee of the Unique Optimum in Our Social Interaction Model With
Production
Our additional assumption is that each of the arguments in our model with a
particular utility function 𝑈 has a non-empty set. Then, the domain of arguments in the
23
(at least twice) continuously differentiable function 𝑈 also becomes convex from the
constraints shown in A Particular Utility Function for the Social Interaction Model With
Production section form a kind of implicit function about the arguments of the utility
function 𝑈. We get the implicit function for our constraints after an easy
Second, the above particular utility function 𝑈 satisfies the following definition
in Sundaram (1996) and Osborne (2022) for the strict quasi-concavity: 𝑈[𝜆𝐶𝑡′ +
∀
𝐷[𝜆𝐶𝑡′ + (1 − 𝜆)𝐶𝑡′′ ]𝛼 [𝜆𝐻𝑡′ + (1 − 𝜆)𝐻𝑡′′ ]1−𝛼 > min[𝑈(𝐶𝑡′ , 𝐻𝑡′ ), 𝑈(𝐶𝑡′′ , 𝐻𝑡′′ )] for 𝜆∈
(0,1) given any 𝐶𝑡′ ∈ ℝ≥0 , any 𝐻𝑡′ ∈ [0,1] and any 𝐶𝑡′′ ∈ ℝ≥0 , any 𝐻𝑡′′ ∈ [0,1] and
𝐶𝑡′ , 𝐻𝑡′ , 𝐶𝑡′′ , and 𝐻𝑡′′ are different because any pair (𝐶𝑡 , 𝐻𝑡 ) is the convex set in real
spaces, and our utility function has positive first-order derivatives and negative second-
order derivatives. Additionally, using MAXIMA, we can show the necessary and
sufficient conditions of the alternate signs of the bordered Hessian for quasi-concavity
(e.g., Osborne, 2022), and this utility function is monotonically increasing (e.g., Fuente,
24
As Dixit (1990), Sundaram (1996), and Jehle and Reny (2001) suggested, the
called a constraint qualification for confirming the optima. The additional condition is
that all three partial derivatives of the resource constraint are not zero. Our constrained
maximization problem in the social interaction model with production satisfies this
requirement.
problem is to maximize the strictly quasi-concave objective function with the quasi-
concave constraints such that it has a unique optimum (see Theorem 1.20 in Fuente,
2000).
𝛽
constraint as 𝑅(𝑟𝑡 ) = 𝑍𝑡 𝐼𝑡 (1 − 𝐻𝑡 )1−𝛽 − 𝑃𝐶,𝑡 𝐶𝑡 − 𝑃𝐼,𝑡 𝐼𝑡 ≥ 0, where 𝑟𝑡 = (𝐶𝑡 , 𝐻𝑡 , 𝐼𝑡 ),
and using the definitions for quasi-concavity in Sundaram (1996) and Osborne (2022),
we ensure the necessary condition of the alternate signs of the bordered Hessian for
Then, we take up any 𝑠𝑡 = (𝐶𝑡′ , 𝐻𝑡′ , 𝐼𝑡′ ) and 𝑞𝑡 = (𝐶𝑡′′ , 𝐻𝑡′′ , 𝐼𝑡′′ ) vectors within
∀
their domains and 𝜃 ∈ (0,1) such that we have 𝑅[𝜃𝑠𝑡 + (1 − 𝜃)𝑞𝑡 ] = 𝑍𝑡 [𝜃𝐼𝑡′ +
25
(1 − 𝜃)𝐼𝑡′′ ]𝛽 {1 − [𝜃𝐻𝑡′ + (1 − 𝜃)𝐻𝑡′′ ]}1−𝛽 − 𝑃𝐶,𝑡 [𝜃𝐶𝑡′ + (1 − 𝜃)𝐶𝑡′′ ] − 𝑃𝐼,𝑡 [𝜃𝐼𝑡′ +
(1 − 𝜃)𝐼𝑡′′ ] ≥ min[𝑅(𝑝𝑡 ), 𝑅(𝑞𝑡 )], because 𝑠𝑡 = (𝐶𝑡′ , 𝐻𝑡′ , 𝐼𝑡′ ) and 𝑞𝑡 = (𝐶𝑡′′ , 𝐻𝑡′′ , 𝐼𝑡′′ )
vectors are convex sets in the nonnegative domains of real spaces from our previous
assumptions, and the first-order derivatives of our production function are positive and
second-order derivatives are negative. Note that we do not have strict inequality in the
above, different from our previous utility function 𝑈 case, because we might have the
investment in their domains, and set 𝛽, 𝑃𝐶,𝑡 , 𝑃𝐼,𝑡 , and 𝑍𝑡 = 𝑍̅ to be constant as usual.
Because it is tedious to think of complicated figures, they are set only for simplicity and
the following numerical example; their values themselves have no substantial meaning
for this proof. Let 𝐶̌𝑡 be a certain consumption under the condition 𝐶̌𝑡 ≤ 𝑌̌𝑡 − 𝐼̌𝑡 . To
𝛽
keep 𝑌̌𝑡 (= 𝐼̌𝑡 [1 − 𝐻
̌𝑡 ]1−𝛽 ≥ 𝐶̌𝑡 + 𝐼̌𝑡 ) unchanged, make 𝐻
̌𝑡 slightly smaller than
After that, we take this new 𝐼𝑡̿ and set 𝐶𝑡̿ ≤ 𝑌̌𝑡 − 𝐼𝑡̿ appropriately, where 𝑌̌𝑡 =
𝛽 ̌𝑡 )1−𝛽 = 𝑍̅𝐼𝑡𝛽̿ (1 − 𝐻
𝑍̅𝐼̌𝑡 (1 − 𝐻 ̿𝑡 )1−𝛽 = 𝑌̿𝑡 . Therefore, the value of 𝑅(𝐶𝑡̿ , 𝐻
̿𝑡 , 𝐼𝑡̿ ) can
26
concave (but not strict quasi-concave).
This is our proof but is too abstract. Therefore, for example, we can demonstrate
this proof with a numerical calculation (EXCEL was used with the goal-seeking
function). First, our case is sensitive to decimal numbers in a rigorous sense, but our
conclusions are not essentially changed if we ignore these decimal numbers. Therefore,
we round off the following figures in our calculations to three decimal places: Assume
̿𝑡 = 0.333, and 𝐼𝑡̿ = 184.068. Then, we have 𝑌̌𝑡 = 540.098 = 𝑌̿𝑡 . In this case, we can
𝐻
have 𝑅(𝐶̌𝑡 , 𝐻
̌𝑡 , 𝐼̌𝑡 ) = 𝑅(𝐶𝑡̿ , 𝐻
̿𝑡 , 𝐼𝑡̿ ) with 𝐶̌𝑡 = 400.098 ≠ 356.030 = 𝐶𝑡̿ .
Additionally, for the other inequality constraint 𝐻𝑡 ≤ 1, we can safely assume the
Note that in the next section regarding the heterogeneity, we denote our variables
(𝐶𝑡 , 𝐻𝑡 , 𝐼𝑡 ) by adding the right-upper subscript 𝑗 and should modify our discussions in
the above particular utility function and two inequalities for the strict quasi-concavity
and quasi-concavity. However, in these cases, without the other inequality constraint
𝑗
𝐻𝑡 ≤ 1, which should be 𝐻𝑡 ≤ 1, we should change (𝐶𝑡 , 𝐻𝑡 , 𝐼𝑡 ) in terms of
𝑗 𝑗 𝑗 𝑗 𝑗
(𝐶𝑡 , 𝐻𝑡 , 𝐼𝑡 ) as (E[𝐶𝑡 ], E[𝐻𝑡 ], E[𝐼𝑡 ]) like E(𝐶𝑡 ) = ∑𝑁 𝑁
𝑗=1 𝑝𝑗 𝐶𝑡 , E(𝐼𝑡 ) = ∑𝑗=1 𝑝𝑗 𝐼𝑡 , and
27
𝑗
E(𝐻𝑡 ) = ∑𝑁
𝑗=1 𝑝𝑗 𝐻𝑡 , where 𝑝𝑗 is the probability based on the stationary distribution
with ∑𝑁
𝑗=1 𝑝𝑗 = 1 that consumer (type) 𝑗 will appear in society (see the next section
for details). This transformation does not change our above arguments at all (consider
𝑗 𝑗
E[𝐶𝑡 ] = ∑𝑁 𝑘
𝑗=1 𝑝𝑗 𝐶𝑡 = 𝑝𝑗 𝐶𝑡 + ∑𝑘≠𝑗 𝑝𝑘 𝐶𝑡 , for example).
Again, Fuente (2000) explained that the optimum solution is the global maximum in the
Furthermore, the convexities of the feasible set {𝑟𝑡 : 𝑅(𝑟𝑡 ) ≥ 0} where 𝑟𝑡 = (𝐶𝑡 , 𝐻𝑡 , 𝐼𝑡 ),
(Fuente, 2000). In this concave programming view, we can secure our unique optimum
property.
always the unique global optimum under the particular utility function 𝑈 defined on its
convex domain secured by its strict quasi-concavity and the quasi-concave inequality
constraints, as Fuente (2000) showed. We safely apply the unique global optimum in our
social interaction model with production. This result is different from that of Brock and
Durlauf (1995), who suggested multiple equilibria (Note that the above discussions
28
similarly hold with only 𝐴 = 0 case or only 𝐵 = 0 case).
Heterogeneity
establish multiple equilibria. In our case, we consider the heterogeneity to examine its
(possible) effects on our unique optimum conclusion; for example, suppose that the
parameter 𝐷𝑗 is given randomly with the probability 𝑝𝑗 , which is based on the stationary
distribution with ∑𝑁
𝑗=1 𝑝𝑗 = 1, that consumer (type) 𝑗 will appear in society for each
consumer.
considering such heterogeneity mainly requires solving the consumer’s expected utility
maximization problem together with the static (in our case, innately stationary) setting
for the probability distribution and the market clearing condition in the previous section,
𝑗 𝑗
utility function 𝑈 , the additional deterministic linear terms 𝐶𝑡 = 𝐶𝑡 and 𝐻𝑡 = 𝐻𝑡
𝑗 𝑗
must be changed into the averaged ones E(𝐶𝑡 ) = ∑𝑁 𝑁
𝑗=1 𝑝𝑗 𝐶𝑡 and E(𝐻𝑡 ) = ∑𝑗=1 𝑝𝑗 𝐻𝑡 .
In these equations, again, 𝑝𝑗 is the probability based on the stationary distribution with
∑𝑁
𝑗=1 𝑝𝑗 = 1 that consumer (type) 𝑗 with 𝐷𝑗 will appear in society. Thus, these
29
alternative linear terms appear in this new expected particular utility function 𝑈 ∗ . The
linear terms in the new expected particular utility function 𝑈 ∗ become the joint averages
As a result, the new expected utility maximization problem in the social interaction
𝑗 𝑗 𝑗 𝑗 𝑗 𝑗
max
𝑗 𝑗 𝑗
E[𝑈(𝐶𝑡 , 𝐻𝑡 )] = 𝐴 ∑𝑁 𝑁 𝑁 𝑗 𝛼
𝑗=1 𝑝𝑗 𝐶𝑡 + 𝐵 ∑𝑗=1 𝑝𝑗 𝐻𝑡 + ∑𝑗=1 𝑝𝑗 𝐷 (𝐶𝑡 ) (𝐻𝑡 )
1−𝛼
,
𝐶𝑡 ,𝐻𝑡 ,𝐼𝑡
𝑗 𝑗 𝑗 𝛽
subject to 𝑌𝑡 ≥ 𝑃𝐶,𝑡 ∑𝑁 𝑁 𝑁
𝑗=1 𝑝𝑗 𝐶𝑡 + 𝑃𝐼,𝑡 ∑𝑗=1 𝑝𝑗 𝐼𝑡 , 𝑌𝑡 = 𝑍𝑡 (∑𝑗=1 𝑝𝑗 𝐼𝑡 ) (1 −
𝑗 1−𝛽 𝑗 𝑗 𝑗
∑𝑁
𝑗=1 𝑝𝑗 𝐻𝑡 ) , 𝐶𝑡 , 𝐼𝑡 ≥ 0, 𝐻𝑡 ∈ [0,1],
𝑗 𝑗
following some stationary distribution and ∑𝑁
𝑗=1 𝑝𝑗 = 1. Note that (𝐶𝑡 , 𝐼𝑡 ) is a convex
𝑗
set on ℝ2≥0 and 𝐻𝑡 ∈ [0,1] is a convex set. Additionally, we know that the expected
𝑗 𝑗
utility function E[𝑈(𝐶𝑡 , 𝐻𝑡 )], the resource constraint, the other equality and inequality
constraints, and the implicit function combined with the above first two constraints are
twice differentiable for the utility function and differentiable for the constraints. We
require this condition for numerical methods such as the interior point method with the
30
second-order Taylor expansion for the particular utility function and Newton’s method
For the above constrained expected utility maximization problem, we have the
following first-order conditions, using the multipliers 𝜙 and 𝜇 for the combined
𝑗
implicit inequality constraint and the inequality constraint 𝐻𝑡 ≤ 1 (see, e.g., Dixit,
1990, for the concrete solution method for the Karush-Kuhn-Tucker theorem):
𝑗 𝑗 𝑗
𝐴 + 𝛼𝐷 𝑗 (𝐶𝑡 )𝛼−1 (𝐻𝑡 )1−𝛼 − 𝜙𝑃𝐶,𝑡 ≤ 0, if 𝐶𝑡 ⋝ 0,
𝑗 𝑗 𝑗 𝛽
𝑝𝑗 𝐵 + 𝑝𝑗 (1 − 𝛼)𝐷 𝑗 (𝐶𝑡 )𝛼 (𝐻𝑡 )−𝛼 − 𝑝𝑗 (1 − 𝛽)𝜙𝑍𝑡 (∑𝑁
𝑗=1 𝑝𝑗 𝐼𝑡 ) (1 −
𝑗 −𝛽 𝑗
∑𝑁
𝑗=1 𝑝𝑗 𝐻𝑡 ) − 𝜇 ≤ 0, if 𝐻𝑡 ⋝ 0,
𝑗 𝛽−1 𝑗 𝑗
𝛽𝑍𝑡 (∑𝑁
𝑗=1 𝑝𝑗 𝐼𝑡 ) (1 − ∑𝑁
𝑗=1 𝑝𝑗 𝐻𝑡 )
1−𝛽
− 𝑃𝐼,𝑡 ≤ 0, if 𝐼𝑡 ⋝ 0,
𝑗 𝛽 𝑗 𝑗 𝑗
𝑍𝑡 (∑𝑁 𝑁
𝑗=1 𝑝𝑗 𝐼𝑡 ) (1 − ∑𝑗=1 𝑝𝑗 𝐻𝑡 )
1−𝛽
− 𝑃𝐶,𝑡 ∑𝑁 𝑁
𝑗=1 𝑝𝑗 𝐶𝑡 − 𝑃𝐼,𝑡 ∑𝑗=1 𝑝𝑗 𝐼𝑡 ≥ 0, if
𝜙 ⋝ 0,
𝑗
1 − 𝐻𝑡 ≥ 0, if 𝜇 ⋝ 0.
To solve the above new consumer’s expected utility maximization problem and
satisfy these first-order conditions, we suppose that the optimal solutions are specific
𝑗 𝑗 𝑗
values of consumption 𝐶𝑡 , investment 𝐼𝑡 , and leisure 𝐻𝑡 . Because the probability
density function for the existing heterogeneous types of consumers is objective and
31
𝑗∗ 𝑗∗ 𝑗∗
stationary from our assumptions, the optimal solution pair (𝐶𝑡 , 𝐻𝑡 , 𝐼𝑡 ) in the
consumer’s expected utility maximization problem should satisfy the market clearing
𝑗 𝑗
condition 𝑌𝑡 = 𝑃𝐶,𝑡 ∑𝑁 𝑁
𝑗=1 𝑝𝑗 𝐶𝑡 + 𝑃𝐼,𝑡 ∑𝑗=1 𝑝𝑗 𝐼𝑡 (i.e., 𝜙 > 0 ) from the definition of
stationary equilibrium.
However, as we easily understand from the above first-order conditions, the new
solution satisfying our stationary equilibrium conditions can only be obtained using
numerical methods, as the representative studies, such as Aiyagari (1994) and Achdou et
al. (2021), suggest (although the latter suggests the analytical solution in the special case).
There are many numerical methods of nonlinear programming; Tamura and Muramatsu
(2002) explained some of them, for example. Furthermore, note that some numerical
methods supply the numerical gradients themselves and do not use the above analytical
For the global uniqueness of this optimum, we apply the theorems of concave
programming in A Particular Utility Function for the Social Interaction Model With
Production section to the new expected particular utility function 𝑈 ∗ and constraints. In
this way, our conclusions remain unchanged even when facing heterogeneity.
Our model based on Becker and Murphy (2000) should not permanently lose the
unique global optimum property, different from Brock and Durlauf (1995), even when
32
considering heterogeneity. Thus, it is clear that the heterogeneity does not change our
main conclusion about the existence of unique optimum in the previous section.
Conclusion
We argue for the extension of Becker and Murphy’s (2000) social interaction
model into the simple general equilibrium model with a social multiplier effect based on
Kiley (2003). We call this model the social interaction model with production. Our
particular utility function represents the existence of social capital, which is generated
representing social capitals are both positive, the increase of these social capitals will
surely bring positive utility to consumers from the goods in question (Kuniyoshi Saito
suggested the latter two points). However, a trade-off between consumption and leisure
exists.
consumption or leisure increases the consumption or the leisure more for each than that
33
makes 𝐻𝑡 take the corner solutions mostly. As Figures 1 and 2 show, the parameter 𝐴
and 𝐵 affects the behavior of 𝐻𝑡 inversely, so if 𝐴 and 𝐵 are too large, the effect of
𝐴 might usually dominate that of 𝐵. We ensure these properties even when we set 𝐴 =
0 or 𝐵 = 0.
unchanged, with negative parameters in our particular utility function regarding the
social capital case. However, depending on the size of the parameters, we cannot
generalize our cases in which the desirable stable solution behavior of the model in a
As for the price effects, from the previous discussions, we know that the existence
of horizontal consciousness affects the behavior of leisure, which the standard utility
function case does not affect, and adds this further effect to the price effects of the
standard utility function case. This is called the social multiplier effect. However, due to
the lack of empirical information about the concrete parameter setting of our model, we
do not show the welfare analysis of the price effects and measurements of their
magnitudes as the social multipliers, different from Brock and Durlauf (1995).
Therefore, it is our future task to determine the parameters in this utility function
for showing the true behaviors of the equilibrium solutions and its social multiplier
34
effect by active empirical analysis.
Because this model differs from that of Brock and Durlauf (1995), who argued for
guarantee the unique global optimum of the social interaction model with production by
our particular utility function satisfying with the strict quasi-concavity and quasi-concave
inequality constraints together with other plausible conditions, even if we consider the
heterogeneity and physical capital accumulation in our model, using existing research
results (e.g., Fuente, 2000; Jehle & Reny, 2001; Nishimura, 1990; Osborne, 2022;
Sundaram, 1996). Although Brock and Durlauf (1995) used the heterogeneous agent
model assuming two types of agents due to the binary choice, heterogeneity is not always
consumers are homogeneous and behave in the same way for convenience at first. As
we saw, this has less essential theoretical meaning, as our heterogeneity result shows
instead, in terms of our interests. However, just as Brock and Durlauf (1995) pursued
the heterogeneous agent model together with their empirical study, in recent years, the
broad field of economics has been accelerating both the theoretical and empirical
35
As Broer et al. (2016) and Kaplan et al. (2018) suggested, the recent
problems different from our unique global optimum problem, such as inequality and
poverty, which innately respects the heterogeneity of agents in society. We should also
Acknowledgments
We would like to thank Akihiko Kaneko and Kuniyoshi Saito for their valuable
comments on our first Japanese draft of this paper. Additionally, we appreciate the
helpful suggestions from Hideki Konishi for polishing this paper and the anonymous
editor at Scribendi completing the English proofreading and offering comments. The
opinions and views expressed herein belong entirely to the author and do not represent
those of the Bank of Japan and/or Central Council for Financial Services Information.
Appendix A.
For the case of the physical capital accumulation contribution instead of no physical
capital accumulation contribution in the current production, we can solve our model to
36
set the model variables as steady-state values, such as 𝐶𝑡 = 𝐶 , 𝐼𝑡 = 𝐼 , and 𝐻𝑡 = 𝐻 .
Then, the solutions for the current physical capital (investment) model in the text hold for
(1995) we use in this paper, even in our model with the utility functions 𝑉 and 𝑈, we
should guess that 𝐶𝑡 and 𝐼𝑡 share 𝑌𝑡 with the fixed ratio and 𝐻𝑡 is constant over
time and then verify this guess. However, this guess can hold only under the specific
utility function setting, and even the utility functions 𝑉 and 𝑈 cannot verify this guess
Initially, a particular utility function 𝑈 satisfies the strict quasi-concavity, and the
inequality (resource) constraints are quasi-concave. This fact suggests that its unique
global optimum is only the solution with the steady-state technique (Sundaram, 1996;
Fuente, 2000). To be more precise, for our model with a particular utility function 𝑈
under the case of the physical capital accumulation, as the usual macroeconomic theory
However, this technique yields similar solutions and results to those in this paper.
Therefore, we conclude that the question of whether we should consider the physical
37
With this steady-state technique, our framework with the Karush-Kuhn-Tucker
theorem can easily be extended into the one with Hamiltonian, that is, Lagrangian of the
full intertemporal maximization problem with the physical capital accumulation formula
(without necessarily the perfect depreciation) with adequate modifications of our original
(actually, static) problem, such as adding the discount factor and using the sum of
discounted utilities as the maximized objective instead (see, e.g., Dixit, 1990, for the
Regarding this physical capital accumulation case, for example, in the steady-state
technique, we additionally have 𝐾𝐼,𝑡 = 𝐾𝐼,𝑡−1 = 𝐾𝐼 for the physical capital stock
variable under the usual physical capital accumulation formula, such as 𝐾𝐼,𝑡 = 𝐼𝑡 + (1 −
𝛿𝐼 )𝐾𝐼,𝑡−1, where 𝛿𝐼 ∈ [0,1] is the depreciation rate of the physical capital stock, which
becomes the fourth constraint. Because we assume that 𝐼𝑡 and 𝐾𝐼,𝑡−1 (and thus, [1 −
𝛿𝐼 ]𝐾𝐼,𝑡−1 ) have convex domains, 𝐾𝐼,𝑡 (which is the combination, the sum of 𝐼𝑡 and a
proportional 𝐾𝐼,𝑡−1) has a convex domain. Then, 𝐼𝑡 in the combined resource constraint
should be substituted by this physical capital accumulation formula, but this new resource
in this case).
38
Finally, note that the above steady-state technique should also allow for the case of
social capitals with 𝛿𝑙 ∈ (0,1], 𝑙 = 1,2, instead of the particular case with 𝛿𝑙 = 1, 𝑙 =
1,2, in the text. However, this case is mostly the same as the original case in the text, and
we do not show it because it is trivial. Also, adding this social capital accumulation
formula (see A Particular Utility Function for the Social Interaction Model With
Production section) to our particular utility function does not change its strict quasi-
concavity. Thus, the above steady-state technique does not change our conclusions
Appendix B.
In the text, we only treat the case in which the parameters in our particular utility
function regarding social capitals are both positive. In this appendix, first, we present
Figure A1, illustrating the case when 𝐴 is negative but 𝐵 remains positive. Then,
Figure A2 shows the case when 𝐵 is negative but 𝐴 remains positive. Finally, the
combined complicated case is shown in Figure A3, when 𝐴 and 𝐵 are both negative.
Note that for these figures, the assumption for other parameters is the same as in Figures
39
To summarize our conclusions in this appendix, the more significant 𝐻𝑡
becomes, (a) the smaller 𝐴 gets and (b) the more significant 𝐵 gets. (c) As a result, we
have extreme values as zeros or ones (as corner solutions) for 𝐻𝑡 and can hardly
interpret its behaviors when 𝐴 or 𝐵 becomes too small in negative values, similar to
the case with a positive, too significant 𝐴 or 𝐵, although we do not show the
corresponding figures.
contrast to Figures 1 and 2. This is because we show the cases of the moderate
𝑋𝑡 − 𝑋 are not different from those of the positive 𝐴 and 𝐵 case without showing any
In this case, 𝐴 = -0.001 for the left and -0.0005 for the right in Figure A1. As
̅ , the smaller 𝐴
shown in the figure, the more significant 𝐻𝑡 becomes compared with 𝐻
gets. Furthermore, extreme values such as zeros or ones are taken as corner solutions for
𝐻𝑡 , and we can hardly interpret its behaviors when 𝐴 gets too small.
Figure A1
40
𝐴 = -0.001 for the Left and -0.0005 for the Right, Given 𝐵 = 0.05
In this case, 𝐵 = -0.50 for the left and -0.25 for the right in Figure A2. As shown
in the figure, the more significant 𝐻𝑡 becomes, the more significant 𝐵 gets.
Furthermore, extreme values such as zeros or ones are taken as corner solutions for 𝐻𝑡 ,
and we can hardly interpret its behaviors when 𝐵 gets too small.
Figure A2
𝐵 = -0.50 for the Left and -0.25 for the Right, Given 𝐴 = 0.0001
41
𝑨 and 𝑩 Are Negative
In this case, as shown in Figure A3, the more significant 𝐻𝑡 becomes, the more
Figure A3
𝐴 = -0.015 and 𝐵 = -1.000 for the Top Left, 𝐴 = -0.010 and 𝐵 = -1.000 for the Top
Right, 𝐴 = -0.015 and 𝐵 = -0.0001 for the Bottom Right, and 𝐴 = -0.010 and 𝐵 = -
This case shows that 𝐴 = -0.015 and 𝐵 = -1.000 for the top left, 𝐴 = -0.010 and 𝐵 =
42
-1.000 for the top right, 𝐴 = -0.015 and 𝐵 = -0.0001 for the bottom left, and 𝐴 = -0.010
and 𝐵 = -0.0001 for the bottom right. As shown in the figure, the more significant 𝐻𝑡
becomes, the more significant 𝐵 gets or the smaller 𝐴 gets, or both. These are the intuitive
results that 𝐻𝑡 becomes more significant when the horizontal consciousness for the leisure
take negative values, as in the top left, bottom left and right of Figure A3.
Finally, although we do not show the corresponding figures, extreme values such as
zeros or ones are taken as corner solutions for 𝐻𝑡 , and we can hardly interpret its behaviors
40
References
Achdou, Y., Han, J., Lasry, J.-M., Lions, P.-L., & Moll, P. (2022). Income and wealth
Becker, G. S., & Murphy, K. J. (2000). Social economics market behavior in a social
Bénassy, J-P. (2011). Money and wage contracts in an optimizing model of the business
Brock, W. A., & Durlauf, S. N. (1995). Discrete choice with social interaction I: Theory
Broer, T., Hansen, N.-J. H., Krusell, P., & Öberg, E. (2016). The new Keynesian
40
Paper No. 22418). National Bureau of Economic Research.
University Press.
Press.
https://www.nobelprize.org/prizes/economic-sciences/2000/heckman/lecture/
Jehle, G. A., & Reny, P. J. (2001). Advanced microeconomic theory (2nd ed.). Addison
Wesley.
Kaplan, G., Moll, B., & Violante, G. L. (2018). Monetary policy according to HANK.
41
American Economic Review, 108(3), 697–743.
Chicago.
Kiley, M. T. (2003). An analytical approach to the welfare cost of business cycles and
3(1), 1–23.
Ljungqvist, L., & Sargent, T. (2000). Recursive macroeconomic theory. MIT Press.
Mas-Colell, A., Whinston, M. D., & Green, J. D. (1995). Microeconomic theory. Oxford
University Press.
https://mjo.osborne.economics.utoronto.ca/index.php/tutorial/index/1/CVN/t#p:
42
CcvConds
Press.
method]. Kyoritushupppan.
43