Abstract
Recent studies highlight the effect of uncertainty on investment decisions, i.e., the “wait and see” effect. In the existing literature, however, uncertainty is viewed as just an exogenous variable. In this paper, uncertainty is not exogenously given but endogenously determined, generating a feedback-loop between micro-behaviors and the macroeconomic environment. In particular, we assume that when uncertainty is high, expectations and behaviors of micro-agents are more heterogeneous. Under these settings, we show that on the one hand heterogeneity contributes to the stability at the macroscopic level, but on the other hand this macro-stability is suddenly lost when investors are sensitive to uncertainty. Namely, investors suddenly and simultaneously “wait and see”.
Similar content being viewed by others
Notes
Caballero (1999, p.815) argues that “[i]t turns out the changes in the degree of coordination of lumpy actions play an important role in shaping the dynamic behavior of aggregate investment.” On the other hand, Thomas (2002) presents the result that plant-level lumpy investment is irrelevant for aggregate activities in the general equilibrium model: “In contrast to previous partial equilibrium analyses, model results reveal that the aggregate effects of lumpy investment are negligible” (Thomas 2002, p.508). However, Gourio and Kashyap (2007) recalibrate the Thomas’s model and obtain different properties from the standard RBC model.
The relationship between uncertainty and disagreement and related literature are discussed in Sect. 4.
The importance of a feedback mechanism is recognized in the existing literature. For example, Caplin and Leahy (1997, pp.601–602) emphasize that “[o]ne of the most limiting aspects of these models is that they focus exclusively on the impact that microeconomic inertia has on aggregate dynamics. They ignore the feedback from aggregates onto individual behavior”. Thomas (2002) constructs a dynamic stochastic general equilibrium model in order to include a feedback mechanism. We also follow the statement of Caplin and Leahy (1997) by introducing the feedback effect between micro units and macroscopic distributional properties.
In the existing literature, the mean-field effect has been intensively investigated (see, e.g., Opper and Saad 2001), but to the best of our knowledge, the effect of the variance has been ignored. We present an alternative method to analyze collective behavior of this type in this paper.
It should be noted that while X(t) is treated as a stochastic process in our model, it does not imply that firms’ managers randomly choose their investment. Our assumption of X(t) is not in conflict with the optimal behavior. Their decisions are optimally determined based on X(t) and other variables such as market conditions, the interest rate, the technology of firms, and the personality of entrepreneurs, and if we should obtain these information in detail, investment decision making would be modeled without any randomness. However, the detailed information is not available and, therefore, X(t) is viewed as a stochastic process from a third party’s point of view.
In this regard, cancer and life insurance give us a good example of how our assumption works. Cancer is a metabolic disease and can be in principle explained by chemical reaction without any random component. However, since its mechanism is too complicated, occurrence of cancer can be dealt with as a random variable, even though it is caused by some deterministic process. It cannot be (and need not be) distinguished with the one generated by the action of chance. Indeed, whether it is generated by a deterministic process or by the action of chance does not yield any difference in the calculation of the premium of life insurance.
Here, \(C_c^{\infty }(\mathbb {R})\) denote the set of smooth functions on \(\mathbb {R}\) with compact support. We simply assume \(C_c^{\infty }(\mathbb {R}) \subset \mathcal{D}(A)\), that is, A is defined on the set of test functions that is large enough.
In general, a conservative Markov semigroup (i.e., firms do not disappear) does not imply that the killing rate e is equal to 0. However, if jump size is bounded as discussed in Appendix 1, conservativeness implies \(e=0\) by Lemmas 2.28 and 2.32 in Böttcher et al. (2013). Thus, we assume \(e=0\) in the following.
Recall that the intensity rate \(\lambda (x)\) determines the number of jumps, and \(\nu (x,{\text {d}}y)\) determines the size of independent jumps.
For the Krylov–Bogolyubov theorem, see, e.g., Corollary 3.1.2 in Da Prato and Zabczyk (1996).
See, e.g., Section 4.2 in Da Prato and Zabczyk (1996). More precisely, in order to apply the Krylov–Bogolyubov theorem and Doob’s theorem, we need to modify the stochastic behavior at large \(|X_t|\). For example, we have to assume that the depreciation rate \(\delta (x)\), which was simply assumed to be constant, becomes 0 at \(x \ll 0\) and the measure \(\nu (x, \cdot )\) has a compact support, in order that \(\int _{[-E, E]} p_t(x, y){\text {d}}y = 1, \ \forall t \ge 0, \ \forall x \in [-E, E]\) for some \(E >0\). Then, we redefine our stochastic process on \([-E, E]\) and apply the two theorems. As we will see later, the well-posedness and convergence are confirmed by the following stability analysis and numerical simulations.
This stability is related to the concept of “stochastic macro-equilibrium”, which is originally advanced by Tobin (1972) in his attempt to explain the observed Phillips curve. He argues that “a theory of stochastic macro-equilibrium: stochastic, because random inter-sectional shocks keep individual labor markets in diverse states of disequilibrium; macro-equilibrium, perpetual flux of particular markets produces fairly definite aggregate outcomes...” (Tobin 1972, p.9).
Bloom (2014) surveys both empirical and theoretical literatures that study the relationship between uncertainty and economic activity.
This equation is a natural generalization of linear Markov processes. In fact, if the generator is given by \(B f(x) = b(x) \frac{{\text {d}} f(x)}{{\text {d}}x}+ \frac{1}{2} \sigma ^2(x) \frac{{\text {d}}^2 f(x)}{{\text {d}}x^2}\), which corresponds to an Ito diffusion satisfying a stochastic differential equation of \({\text {d}}X_t = b(X_t) {\text {d}}t + \sigma (X_t) {\text {d}}W_t\) and the adjoint operator \(B^*\) is defined by \((B f, \mu _t) = (f, B^* \mu _t)\), Eq. (6) is reduced to the well-known Fokker–Planck equation.
As will be shown later, we do not need to specify \(\eta (x, {\text {d}}y)\) other than \(m_j(x)\) and \(\sigma _j^2\) under Gaussian approximation.
Of course, since we use the approximation method, \(m_{GA}\) and \(\sigma ^2_{GA}\) deviate from their counterparts \(m^*\) and \({\sigma ^*}^2\). However, as we will see in Sect. 6, Gaussian approximation well describes the evolution of \(m^*\) and \({\sigma ^*}^2\), i.e. that of the probability distribution qualitatively and quantitatively.
Note that the eigenvectors depend on the value of c.
Furthermore, because the number of firms \(N < \infty\) in a real economy, the collective behavior can be induced by the finite size effect, that is, the fluctuation due to \(N < \infty\) without any aggregate shocks.
To simplify the notation, the subscript t representing time is omitted below.
References
Aoki M (1996) New approaches to macroeconomic modeling: evolutionary stochastic dynamics, multiple equilibria, and externalities as field effects. Cambridge University Press, Cambridge
Bachmann R, Elstner S, Sims ER (2013) Uncertainty and economic activity: evidence from business survey data. Am Econ J Macroecon 5(2):217–249
Baker SR, Bloom N, Davis SJ (2013) Measuring economic policy uncertainty. Chicago Booth research paper 13-02
Bloom N (2009) The impact of uncertainty shocks. Econometrica 77(3):623–685
Bloom N (2014) Fluctuations in uncertainty. J Econ Perspect 28(2):153–175
Bloom N, Bond S, Van Reenen J (2007) Uncertainty and investment dynamics. Rev Econ Stud 74(2):391–415
Bomberger WA (1996) Disagreement as a measure of uncertainty. J Money Credit Bank 28(3):381–392
Böttcher B, Schilling R, Wang J (2013) Lévy matters III. Springer, Berlin
Caballero RJ (1999) Aggregate investment. Handb Macroecon 1:813–862
Caballero RJ, Engel EM (1991) Dynamic (S, s) economies. Econometrica:1659–1686
Caballero RJ, Engel EM (1999) Explaining investment dynamics in US manufacturing: a generalized (S, s) approach. Econometrica 67(4):783–826
Caplin A, Leahy J (1997) Aggregation and optimization with state-dependent pricing. Econometrica:601–625
Carruth A, Dickerson A, Henley A (2000) What do we know about investment under uncertainty? J Econ Surv 14(2):119–154
Clements MP (2008) Consensus and uncertainty: using forecast probabilities of output declines. Int J Forecast 24(1):76–86
Cooper R, Haltiwanger J, Power L (1999) Machine replacement and the business cycle: lumps and bumps. Am Econ Rev:921–946
Courrège P (1965–1966). Sur la forme intégro-différentielle des opérateurs de \(C^\infty _k\) dans \(C\) satisfaisant au principe du maximum. Séminaire Brelot-Choquet-Deny. Théorie Potentiel 10(1):1–38
Da Prato G, Zabczyk J (1996) Ergodicity for infinite dimensional systems. London mathematical society lecture note series. Cambridge University Press, Cambridge
Dixit AK, Pindyck RS (1994) Investment under uncertainty. Princeton University Press, Princeton
Doms M, Dunne T (1998) Capital adjustment patterns in manufacturing plants. Rev Econ Dyn 1(2):409–429
Fuss C, Vermeulen P (2008) Firms’ investment decisions in response to demand and price uncertainty. Appl Econ 40(18):2337–2351
Giordani P, Söderlind P (2003) Inflation forecast uncertainty. Eur Econ Rev 47(6):1037–1059
Gourio F, Kashyap AK (2007) Investment spikes: new facts and a general equilibrium exploration. J Monet Econ 54:1–22
Hommes CH (2006) Heterogeneous agent models in economics and finance. In: Tesfatsion L, Judd K (eds) Handbook of computational economics, vol 2. Elsevier, Oxford, pp 1109–1186
Jurado K, Ludvigson SC, Ng S (2015) Measuring uncertainty. Am Econ Rev 105(3):1177–1216
Kirman A (2010) Complex economics: individual and collective rationality, vol 10. Routledge, Abingdon
Kolokoltsov V (2011) The L\({\rm \acute{e}}\)vy–Khintchine type operators with variable Lipschitz continuous coefficients generate linear or nonlinear Markov processes and semigroups. Probab Theory Relat Fields 151(1–2):95–123
Leahy JV, Whited TM (1996) The effect of uncertainty on investment: some stylized facts. J Money Credit Bank 28(1):64–83
Nilsen ØA, Schiantarelli F (2003) Zeros and lumps in investment: empirical evidence on irreversibilities and nonconvexities. Rev Econ Stat 85(4):1021–1037
Opper M, Saad D (2001) Advanced mean field methods: theory and practice. MIT Press, Cambridge
Thomas JK (2002) Is lumpy investment relevant for the business cycle? J Polit Econ 110(3):508–534
Tobin J (1972) Inflation and unemployment. Am Econ Rev 62(1):1–18
Zarnowitz V, Lambros LA (1987) Consensus and uncertainty in economic prediction. J Polit Econ 95(3):591–621
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
On behalf of all authors, the corresponding author states that there is no conflict of interest.
Additional information
This research was conducted as a part of the Project “Sustainable Growth and Macroeconomic Policy” undertaken at Research Institute of Economy, Trade and Industry (RIETI). This paper has greatly benefited from comments by Hiroshi Yoshikawa. We would like to thank Toichiro Asada, Jun-Hyung Ko, Akio Matsumoto as well as many colleagues, especially Masayuki Morikawa and Willem Thorbecke, for very useful comments. We are grateful to participants at the 2015 International Conference on Nonlinear Economic Dynamics (NED) at Chuo University as well as seminars at Aoyama Gakuin University, Chuo University, Kyoto University, Sophia University, and RIETI for helpful and constructive comments. This paper was financially supported by JSPS KAKENHI Grant Numbers 13J10353 (Kimura) and 14J03350 (Murakami).
Appendices
Appendix 1: Feller processes
Here, we discuss Assumption 1 from an economic standpoint. First, we consider the strong continuity. It can be shown that the strong continuity is equivalent to pointwise convergence (e.g., Lemma 1.4 in Böttcher et al. 2013):
Because \(T_t u(x)\) is the expectation value after time t with an initial point x, (16) simply means that the process \(X_t\) stay at its initial value if t is small. Note that it does not exclude the possibility that the process \(X_t\) jumps at around 0. The property (16) means that this probability converges to 0 as \(t \rightarrow 0\). In this sense, this property is closely related to the stochastic continuity. Since there is no reason to assume that \(X_t\) jumps exactly at 0 with positive probability, the strong continuity is a plausible assumption in an economic sense.
Second, to discuss the Feller property, we introduce an additional assumption:
Assumption 4
For all \(t > 0\) and any increasing sequence of bounded sets \(B_n \in \mathcal{B}(\mathbb {R})\) with \(\cup _{n \ge 1} B_n = \mathbb {R}\), the transition function \(p_t(x, B_n)\) satisfies the following property;
Here, \(\mathcal {B}(\mathbb {R})\) is Borel sets. For example, suppose that there exist increasing adjustment costs \(\psi (\Delta X_t)\) such that \(\psi '(\Delta X_t)>0, \psi ''(\Delta X_t)>0\), or financial or physical constraints, and, therefore, capital adjustment to some target level cannot be done at one time when \(X_t\) is extremely small. In such a situation, Assumption 4 can be justified. Then, the following lemma implies the Feller property.
Lemma 5
Assumption 4 implies the Feller property.
Proof
Let \(u \in C_{\infty }(\mathbb {R})\). For an arbitrary \(\epsilon\), there exists N such that \(|u| < \epsilon\) on \(\mathbb {R}\setminus B_n\) for all \(n \ge N\). Therefore,
where \(||\cdot ||_{\infty }\) denotes the sup norm on \(C_{\infty }(\mathbb {R})\).
Since \(\epsilon\) is arbitrary, Assumption 4 implies that \(T_t u \in C_{\infty }(\mathbb {R})\). \(\square\)
Appendix 2: Stability analysis
In this section, we investigate the properties of the following system of differential equations of the first moment (mean) m and the second central moment \(\sigma ^2\ge 0\) obtained in Sect. 5:Footnote 22
An equilibrium point of the system of (17) and (18), \((m^*,{\sigma ^2}^*) \in \mathbb {R}\times \mathbb {R}_+,\) is defined as a solution of the following system of simultaneous equations:
One can easily find that \({\text {d}}m/{\text {d}}t={\text {d}}\sigma ^2/{\text {d}}t=0\) if and only if \((m,\sigma ^2)=(m^*,{\sigma ^2}^*).\) Concerning the existence and uniqueness of an equilibrium \((m^*,{\sigma ^2}^*),\) the following proposition holds.
Proposition 2
There uniquely exists an equilibrium point \((m^*,{\sigma ^2}^*) \in \mathbb {R}\times \mathbb {R}_+\).
Proof
Equation (20) can be solved for \(\sigma ^2\ge 0\) as
Let \(z=a-m.\) Then, substituting (21) in (19), we have
where g(z) and h(z) are defined as follows:
Since the right hand side of (22) is positive, for some z to satisfy (22), we must have
or
Both g and h are positive and strictly increasing in z within the range of (23) and that we have \(g(\underline{z})h(\underline{z})=0\) and \(g(\infty )h(\infty )=\infty .\) Hence, there uniquely exists a \(z^*\) that meets (22). Letting \(m^*=a-z^*\) and \({\sigma ^2}^*=[-1+\sqrt{1+4({z^*}^2+\sigma _j^2)}]/2,\) we can find that \((m^*,{\sigma ^2}^*) \in \mathbb {R}\times \mathbb {R}_+\) is a unique equilibrium point. \(\square\)
Corollary 6
Assume that the following condition is satisfied:
Then, we have \(m^*<a.\)
Proof
About this article
Cite this article
Arata, Y., Kimura, Y. & Murakami, H. Aggregate implications of lumpy investment under heterogeneity and uncertainty: a model of collective behavior. Evolut Inst Econ Rev 14, 311–333 (2017). https://doi.org/10.1007/s40844-017-0074-5
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40844-017-0074-5