Abstract
A recent strand of macroeconomic literature has placed sentiment fluctuations at the forefront of the academic debate about the foundations of business cycles. Waves of optimism and pessimism influence the decisions of investors and consumers, and they might therefore be interpreted as a driving force for the performance of the economy in the short term. In this context, two questions regarding the formation and evolution of psychological moods in an economic setting gain relevance: First, how can we model the process of transmission of sentiments across economic agents? Second, is this process capable of generating endogenous and persistent fluctuations? This paper answers these two questions by proposing a simple and intuitive continuous-time dynamic sentiment spreading model based on the rumor propagation literature. As agents contact with one another, endogenous fluctuations are likely to emerge, with trajectories of sentiment shares potentially exhibiting periodic cycles and chaotic behavior.
Similar content being viewed by others
Notes
At this initial stage, we make a terminological clarification that is important for the discussion that follows. To be rigorous, the concepts of sentiments, understood as a predisposition to be optimistic or pessimistic, and of animal spirits, in the strict Keynesian sense, are not necessarily synonymous. Although some authors refer to these terms interchangeably (e.g., Benhabib et al. 2015, page 549, state that “Fluctuations can be driven by waves of optimism or pessimism, or as in Keynes’ terminology, by ‘animal spirits’ that are distinct from fundamentals”), Keynes’ notion of animal spirits has a precise meaning; in the General Theory (Keynes 1936) animal spirits are referred to as a spontaneous urge to action rather than inaction. Because our sentiment spreading model deals with behavioral features that do not necessarily comply with this idea of propensity to take actions, we enclose them in the less formal notion of sentiments and refrain from using the term “animal spirits.”
Henceforth, to simplify notation and when no ambiguity arises, the time argument in time-dependent variables will be suppressed.
Keep in mind that this outcome occurs with a probability 𝜃; with a probability 1 − 𝜃, the contact implies no sentiment change.
This product, although convenient for analytical purposes, merges the role of the probability of transition with the role of the degree of connectivity, hiding the individual effects of each parameter when later assessing the model’s dynamics. One should note that all that matters for the purpose of the analysis of dynamics is that any specific value of ζ corresponds to an array of values of 𝜃 and κ, where a relation of opposite sign exists between the two, i.e., if one considers, e.g., ζ = 40, this might mean a transition probability 𝜃 = 0.5 and a connectivity level of κ = 80 or, alternatively, 𝜃 = 0.25 and κ = 160, or any other combination of parameters such that ζ remains at the specified value.
As in Angeletos and La’O (2013), we attribute a pivotal role to those in a state of exuberance. Neutral and non-exuberant individuals do not search actively for new information and are influenced solely by local interaction. Exuberant people, instead, are pro-active in the sense that besides being influenced by direct contact, they also search for economy wide information on the acceptance about the sentiment they hold (measured by the net inflow of people into the sentiment category). In this case, exuberance is synonymous with a proclivity to make an informed decision.
The specific shape of functions (3) and (4) is intended to capture the idea that the transition probability approaches zero when the inflow into the sentiment is stronger than the outflow and approaches 𝜃 > 0 when the respective net outflow is a positive value. This is done rather than considering a piecewise function that would be analytically less tractable and arguable less realistic; the continuity is assured by the specific hyperbolic tangent function. This function introduces a convenient S-shaped nonlinearity similar to the one used in some evolutionary switching models where specific relations are defined through the arctangent function (see, e.g., the supply curve (1.2) in Hommes (2013)).
The assumption of a profit maximizing firm is not, as we shall see below, essential for the analysis, since the only requirement is to consider some benchmark value for investment over which waves of optimism or pessimism might exert some effect. Nevertheless, by solving this problem, expressions for the investment, profits and output will be explicitly derived, and these will be important for the simulation exercise to be performed in the next section.
Given that the system is four-dimensional, it involves four Lyapunov characteristic exponents (LCE). The identification of chaos, based on the evidence of sensitive dependence on initial conditions, requires the existence of at least one positive LCE, as is observed in this particular case.
Note that productivity grows over time at a constant rate g − ρ τ (productivity growth is not affected by sentiment oscillations). Effective investment, profits and output, in turn, suffer the influence of sentiment changes, and hence they will grow at a rate that fluctuates around the productivity growth rate (which is also the growth rate of the optimal levels of investment, profits and output). Therefore, a straightforward way of displaying graphically the oscillations underlying each of the variables is presenting them as ratios with respect to productivity. In this way, the values of the variables will fluctuate around a constant level.
Below, we consider alternate values for this parameter.
References
Angeletos GM, La’O J (2013) Sentiments. Econometrica 81:739–779
Angeletos GM, Collard F, Dellas H (2015) Quantifying confidence. CEPR discussion papers n ∘ 10463
Asada T, Chiarella C, Flaschel P, Franke R (2010) Monetary macrodynamics. Routledge, London
Asada T, Chiarella C, Flaschel P, Mouakil T, Proaño C, Semmler W (2011) Stock-flow interactions, disequilibrium macroeconomics and the role of economic policy. J Econ Surv 25:569–599
Barabási AL, Albert R (1999) Emergence of scaling in random networks. Science 286:509–512
Beaudry P, Galizia D, Portier F (2015) Reviving the limit cycle view of macroeconomic fluctuations. NBER working paper n ∘ 21241
Benhabib J, Liu X, Wang P (2016) Sentiments, financial markets, and macroeconomic fluctuations. J Financ Econ 120:420–443
Benhabib J, Wang P, Wen Y (2015) Sentiments and aggregate demand fluctuations. Econometrica 83: 549–585
Brock WA, Hommes CH (1997) A rational route to randomness. Econometrica 65:1059–1095
Brock WA, Hommes CH (1998) Heterogeneous beliefs and routes to chaos in a simple asset pricing model. J Econ Dyn Control 22:1235–1274
Chahrour R, Gaballo G (2015) On the nature and stability of sentiments. Boston college working papers in economics n° 873
Chiarella C, Flaschel P (2000) The dynamics of keynesian monetary growth Macro foundations. Cambridge University Press, Cambridge
Chiarella C, Flaschel P, Franke R (2005) Foundations for a disequilibrium theory of the business cycle. Qualitative analysis and quantitative assessment. Cambridge University Press, Cambridge
De Grauwe P (2011) Animal spirits and monetary policy. Economic Theory 47:423–457
Durlauf SN (2012) Complexity, economics and public policy. Politics, philosophy and economics 11:45–75
Galtier F, Bousquet F, Antova M, Bommel P (2012) Markets as communication systems. Simulating and assessing the performance of market networks. J Evol Econ 22:161–201
Gomes O (2015a) Sentiment cycles in discrete-time homogeneous networks. Physica A 428:224–238
Gomes O (2015b) A model of animal spirits via sentiment spreading. Nonlinear Dynamics Psychol Life Sci 19:313–343
Hommes CH (2006) 23. In: Heterogeneous Agent Models in Economics and Finance, 1st edn. Tesfatsion L, Judd KL (eds), vol 2
Hommes CH (2013) Behavioral rationality and heterogeneous expectations in complex economic systems. Cambridge University Press, Cambridge
Huo L, Huang P, Guo CX (2012) Analyzing the dynamics of a rumor transmission model with incubation. Discrete Dynamics in Nature and Society, article ID 328151, 21 pages
Keynes JM (1936) The general theory of employment interest and money. MacMillan, London
Lafond F (2015) Self-organization of knowledge economies. J Econ Dyn Control 52:150–165
Lemmon M, Portniaguina E (2006) Consumer confidence and asset prices: some empirical evidence. Rev Financ Stud 19:1499–1529
Leonov GA, Kuznetsov NV, Vagaitsev VI (2011) Localization of hidden Chua’s attractors. Phys Lett A 375:2230–2233
Lucas RE (2009) Ideas and growth. Economica 76:1–19
Lucas RE, Moll B (2014) Knowledge growth and the allocation of time. J Polit Econ 122:1–51
Manski CF (2000) Economic analysis of social interactions. J Econ Perspect 14:115–136
Milani F (2014) Sentiment and the US business cycle. University of California – Irvine, department of Economics working paper n ∘ 141504
Nekovee M, Moreno Y, Bianconi G, Marsili M (2007) Theory of rumor spreading in complex social networks. Physica A 374:457–470
Sprott JC (2011) A proposed standard for the publication of new chaotic systems. Int J Bifurcation Chaos 21:2391–2394
Sprott JC, Xiong A (2015) Classifying and quantifying basins of attraction. Chaos 25:083101
Staley M (2011) Growth and the diffusion of ideas. J Math Econ 47:470–478
Wang YQ, Yang XY, Han YL, Wang XA (2013) Rumor spreading model with trust mechanism in complex social networks. Commun Theor Phys 59:510–516
Zanette DH (2002) Dynamics of rumor propagation on small-world networks. Phys Rev E 65:041908
Zhao LJ, Wang JJ, Chen YC, Wang Q, Cheng JJ, Cui HX (2012) SIHR rumor spreading model in social networks. Physica A 391:2444–2453
Acknowledgements
The detailed and insightful comments of two anonymous referees, which led to a substantial revision of the initially submitted manuscript, are highly appreciated. The usual disclaimer applies.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
Proof of Proposition 1
Applying the equilibrium condition\(\overset {\text {{\large .}}}{y}=\overset {\text {{\large .}}}{v}=\overset {\text {{\large .}}}{z}=\overset {\text {{\large .}}}{w}=0\)to system (2) gives
Thesolutions of the above system are the four equilibrium points and the equilibrium line in the proposition. □
Proof of Proposition 2
The Jacobian matrix for system (2) has the general form
For each of the equilibria,
Fromthe Jacobian matrices, the respective eigenvalues are
The signs of the eigenvalues show that the equilibrium points e 2, e 3, e 4and e 5are all unstable since at least one of the eigenvalues has a non-negative sign. Point e 1is locallystable because the respective eigenvalues are two pairs of complex conjugates with negative realparts. □
Proof of Proposition 3
Equilibrium condition\(\overset {\text {{\large .}}}{y}=\overset {\text {{\large .}}}{v}=\overset {\text {{\large .}}}{z} =\overset {\text {{\large .}}}{w}=0\)is now applied to system (5) with the result,
Given that tanh(0) = 0,the solution of the above system gives the set of equilibrium points claimed in the proposition. □
Proof of Proposition 4
The Jacobian matrix of system (5) is a modified version of the one presented in the proofof Proposition 2 given by
with
For \(e_{1}^{\prime }\), observethat \(\chi (y,y)=\chi (v,v)=\frac {\zeta }{5}\), χ(y, v) = χ(y, w) = χ(v, y) = χ(v, z) = 0, and\(\chi (y,z)=\chi (v,w)=-\frac {\zeta }{5}\). Giventhe values of the derivatives of the hyperbolic tangent functions evaluated in the equilibrium, it isstraightforward to compute the respective Jacobian matrix,
The eigenvalues of \(\widehat {J}\)for \(e_{1}^{\prime }\)are
Relative to the above eigenvalues, four different cases are identifiable (excludingthe border cases that imply the existence of bifurcation points): (i) if\(\zeta >\frac {25}{2}\left (1+\sqrt {20}\right ) \)then the four eigenvalues havepositive real values; (ii) if \(\frac {75}{2}<\zeta <\frac {25 }{2}\left (1+\sqrt {20}\right ) \)then two of the eigenvalues are positive real roots, while the other two area pair of complex conjugate eigenvalues with a positive real part; (iii) if\(\frac {25}{2}<\zeta <\frac {75}{2}\), thenthe eigenvalues are two pairs of complex conjugates with positive real parts; (iv) if\(\zeta <\frac {25}{2}\)thenthe eigenvalues are two pairs of complex conjugates with negative real parts. Only in the lastcase will stability hold, and thus the condition for stability is the one claimed in theproposition.
Next, we analyze the stability of the other equilibrium points and confirm that they are alsounstable. Observe that:
• For\(e_{2}^{\prime }\)and\(e_{3}^{\prime }:\chi (y,y)=\chi (v,v)= \frac {\zeta }{3}\), χ(y, v) = χ(y, w) = χ(v, y) = χ(v, z) = 0,\(\chi (y,z)=\chi (v,w)=-\frac {\zeta }{3};\)
• For e 4 : χ(y, y) = χ(v, v) = ζ, χ(y, v) = χ(y, w) = χ(v, y) = χ(v, z) = 0, χ(y, z) = χ(v, w) = −ζ;
• For e 5 : χ(y, y) = χ(y, z) = z ∗ ζ, χ(y, v) = χ(y, w) = χ(v, y) = χ(v, z) = 0, χ(v, v) = χ(v, w) = (1 − z ∗)ζ.
With these derivatives, matrix \(\widehat {J}\) in each case will be
As in the original model, the eigenvalues of the pair of equilibrium points\( e_{2}^{\prime }\)and\(e_{3}^{\prime }\)areidentical and are given in this case by
At least one ofthe eigenvalues of \(\widehat {J}\)for \(e_{2}^{\prime }\)and\( e_{3}^{\prime }\)has a positive sign,regardless of the value of ζ,and therefore the instability of the respective equilibria is confirmed. For e 4, the Jacobianmatrices J and \(\widehat {J}\)are identical. Thus the corresponding eigenvalues are also the same, namely λ 1,2 = −ζ; λ 3,4 = ζ,and local stability is absent in this case as well. Finally, note that the eigenvalues for e 5are\(\lambda _{1}=-\frac {1}{2} z^{\ast }\zeta ;\lambda _{2}=-\frac {1}{2}(1-z^{\ast })\zeta ;\lambda _{3}=0;\lambda _{4}=\zeta \); forthe equilibrium line, one of the eigenvalues is positive and one other is zero, which again impliesinstability. □
Rights and permissions
About this article
Cite this article
Gomes, O., Sprott, J.C. Sentiment-driven limit cycles and chaos. J Evol Econ 27, 729–760 (2017). https://doi.org/10.1007/s00191-017-0497-5
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00191-017-0497-5