Abstract
This paper is concerned with the multiplicative-noise rational expectations (MRE) problem. By resorting to a linear quadratic optimal control (LQOC) problem, the approximate solution to the MRE problem can be obtained. It is worth highlighting that this approximate solution can be highly close to the exact solution by adjusting weighted matrices in the cost function of the LQOC problem. Since the conditional expectation of the state is involved in the cost function, the LQOC problem is an optimization control with an additional measurability restriction, which is very involved. The orthogonal decomposition technique is used to deal with this measurability restriction. The solvability condition and the optimal decision are given by developing generalized Riccati-type equations. Numerical experiments are provided to illustrate the effectiveness of the achieved results.
Similar content being viewed by others
Data Availability
Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.
References
Ascari, G., Fagiolo, G., Roventini, A.: Fat-tail distributions and business-cycle models. Macroecon. Dyn. 19(2), 465–476 (2015)
Auerbach, A.J., Gorodnichenko, Y.: Fiscal multipliers in recession and expansion. Fiscal Policy after the Financial Crisis 63–98 (2012)
Barro, R.J.: Rational expectations and the role of monetary policy. J. Monet. Econ. 2(1), 1–32 (1976)
Başar, T.: Some thoughts on rational expectations models, and alternate formulations. Computers Mathematics with Applications. 591–604 (1989)
Benhabib, J., Dave, C.: Learning, large deviations and rare events. Rev. Econ. Dyn. 17(3), 367–382 (2014)
Blanchard, O.J.: Backward and forward solutions for economies with rational expectations. Am. Econ. Rev. 69(2), 114–118 (1979)
Blanchard, O.J., Kahn, C.M.: The solution of linear difference models under rational expectations. Econ. J. Econ. Soc. 1305–1311 (1980)
Carravetta, F., Sorge, M.M.: A “nearly ideal’’ solution to linear time-varying rational expectations models. Comput. Econ. 35(4), 331–353 (2010)
Carravetta, F., Sorge, M.M.: Model reference adaptive expectations in Markov-switching economies. Econ. Model. 32, 551–559 (2013)
Chib, S., Ramamurthy, S.: DSGE models with student-t errors. Economet. Rev. 33(1–4), 152–171 (2014)
Chow, G.C.: Estimation of rational expectations models. J. Econ. Dyn. Control 2, 241–255 (1980)
Christiano, J.L.: On the fit of new Keynesian models: comment. J. Bus. Econ. Stat. 60(2), 143–151 (2007)
Cogley, T., Sargent, T.J.: Evolving post-world war II US inflation dynamics. NBER Macroecon. Annu. 16, 331–373 (2001)
Cúrdia, V., Del Negro, M., Greenwald, D.L.: Rare shocks, great recessions. J. Appl. Economet. 29(7), 1031–1052 (2014)
Dave, C., Feigenbaum, J.: Precautionary learning and inflationary biases. Macroecon. Dyn. 24(5), 1124–1150 (2020)
Dave, C., Sorge, M.M.: Sunspot-driven fat tails: a note. Econ. Lett. 193, 109304 (2020)
De Santis, A., Germani, A., Scoglio, C.: Kalman filter approach to solution of rational expectations models. Comput. Math. Appl. 25(12), 3–47 (1993)
Durrett, R.: Probability: Theory and examples, vol. 49. Cambridge University Press, Cambridge (2019)
Fagiolo, G., Napoletano, M., Roventini, A.: Are output growth-rate distributions fat-tailed? Some evidence from OECD countries. J. Appl. Economet. 23(5), 639–669 (2008)
Li, L., Zhang, H., Wang, Y.: Stabilization and optimal control of discrete-time systems with multiplicative noise and multiple input delays. Syst. Control Lett. 147, 104833 (2021)
Ma, T., Xu, J., Zhang, H., Başar, T.: Optimal control approach for rational expectations models with longer forward-looking time. In: 2020 16th International Conference on Control, Automation, Robotics and Vision, pp. 1041–1048 (2020)
Phillis, Y.: Controller design of systems with multiplicative noise. IEEE Trans. Autom. Control 30(10), 1017–1019 (1985)
Sargent, T.J., Wallace, N.: The stability of models of money and growth with perfect foresight. Econ. J. Econ. Soc. 1043–1048 (1973)
Sargent, T.J., Wallace, N.: “Rational expectations’’, the optimal monetary instrument, and the optimal money supply rule. J. Polit. Econ. 83(2), 241–254 (1975)
Shiller, R.J.: Rational expectations and the dynamic structure of macroeconomic models: a critical review. J. Monet. Econ. 4(1), 1–44 (1978)
Taylor, J.B.: Conditions for unique solutions in stochastic macroeconomic models with rational expectations. Econ. J. Econ. Soc. 1377–1385 (1977)
Wall, K.D.: Generalized expectations modeling in macroeconometrics. J. Econ. Dyn. Control 2, 161–184 (1980)
Xu, J., Zhang, H., Başar, T.: Stackelberg solution for a two-agent rational expectations model. Automatica 129, 109601 (2021)
Zhang, H., Li, L., Xu, J., Fu, M.: Linear quadratic regulation and stabilization of discrete-time systems with delay and multiplicative noise. IEEE Trans. Autom. Control 60(10), 2599–2613 (2015)
Acknowledgements
This work was supported by the Foundation for Innovative Research Groups of the National Natural Science Foundation of China(61821004), Original Exploration Program of the National Natural Science Foundation of China(62250056), Major Basic Research of Natural Science Foundation of Shandong Province (ZR2021ZD14), High-level Talent Team Project of Qingdao West Coast New Area (RCTD-JC-2019-05), and Key Research and Development Program of Shandong Province (2020CXGC01208).
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Xiaolu Tan.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Proof of Lemma 2.1
It follows from the definition in Notations that we have
where \({\mathcal {F}}_{k-2}=\sigma \{\omega _0,\omega _1,\ldots ,\omega _{k-2},\nu _0,\nu _1,\ldots ,\nu _{k-2}\}\). Theorem 4.1.14 in [18] implies that for every \({\mathcal {F}}_{k-2}\)-measurable W, there holds
Taking expectation gives
Associating (75) with (77), we know that \({\hat{u}}_{k}|^{k-1}_{k-2}\) is orthogonal to any \({\mathcal {F}}_{k-2}\)-measurable vector. It is evident \({\hat{u}}_{k|k-2}\) is \({\mathcal {F}}_{k-2}\)-measurable, so \({\hat{u}}_{k|k-2}\) and \({\hat{u}}_{k}|^{k-1}_{k-2}\) are orthogonal.
Proof of Theorem 2.1
Considering the truncated cost function (13) for any \(k,0\le k\le K\), we have
Consistent with (78), we assume that the prediction process by the economic agents ends at \(k=K\). Therefore, we set \(u_k=0\) for \(k>K\). From the system (11), (78) can be written as
Now, we handle the optimization problem by dynamic programming procedure. In the case of \(k=K-1\), we get
in virtue of (11) and (80), one has
where
It can be noticed that (81) contains the \({\mathcal {F}}_{K-1}\)-measurable \(u_{K}\) and the \({\mathcal {F}}_{K-2}\)-measurable \(u_{K|K-2}\) simultaneously and they are correlated, which explicitly shows that the converted optimization problem \(\min J_{K-1}\) is an optimization problem with extra measurability constraint. To eliminate the constraint, we decompose the control input orthogonally. The new inputs are no longer correlated, which guarantees dynamic programming procedures can proceed smoothly.
According to Lemma 2.1, \(u_K\) and \(x_K\) in (81) can be orthogonally decomposed as
Inserting (82)-(83) into (81) yields
where we have used the equalities (17)-(18) with \(k=K\) and the initial value \(M^1_{K+1}=A'RA+\tau A_0'RA_0\), \(M^2_{K+1}=A'QA+\tau A_0'RA_0\), \(S_{K+1}=QA\), and \(T_{K+1}=Q\).
Making complete square over \({\hat{u}}_{K|K-2}\) and \({\hat{u}}_{K}|^{K-1}_{K-2}\) in (84) results in
where
In the above, (15)-(20), and (23)-(24) are utilized with \(k=K\).
From (86)–(87) and the system (11), the optimal \(u_K\) admits (46) with \(k=K\). The optimal cost of (85) is given as
In the case of \(k=K-2\), we have
It together with (88) means that one only needs to resolve
Because the right hand of (81) can be written as
which has the same form of (90), thus we complete the induction argument. The optimal cost of (13) is as in (25).
Proof of Lemma 3.1
It follows from the definition in Notations that for \(i=1,\ldots ,d\) and \(j=1,\ldots ,d-1\), we have
From Theorem 4.1.13 in [18], (93) can be rewritten as
Theorem 4.1.14 in [18] implies that for every \({\mathcal {F}}_{k-1-j}\)-measurable W, \(j=1,2,\ldots ,d-1\), there holds
Taking expectation yields
Associating (94) with (96), we know that \({\hat{u}}_{k}|^{k-j}_{k-1-j}\) is orthogonal to any \({\mathcal {F}}_{k-1-j}\)-measurable vector. It is evident \({\hat{u}}_{k|k-i}\) are \({\mathcal {F}}_{k-1-j}\)-measurable as \( j\le i\), so \({\hat{u}}_{k|k-i}\) and \({\hat{u}}_{k}|^{k-j}_{k-1-j}\) are orthogonal for \(1\le j<i\), i.e.,
Similarly, \({\hat{u}}_{k}|^{k-j}_{k-1-j}\) is orthogonal to any \({\mathcal {F}}_{k-1-j}\)-measurable vector. From \({\mathcal {F}}_{k-i}\subseteq {\mathcal {F}}_{k-j-1}\subset {\mathcal {F}}_{k-j}\), \({\hat{u}}_{k}|^{k-i}_{k-1-i}\) with \(j<i\le d\) is \({\mathcal {F}}_{k-j-1}\)-measurable. Accordingly, there holds
Now (97) together with (98) shows that \({\hat{u}}_{k|k-1-d}\), \({\hat{u}}_{k}|^{k-1}_{k-2},\ldots , {\hat{u}}_{k}|^{k-d}_{k-d-1}\) are orthogonal to each other.
Proof of Theorem 3.1
Considering the truncated cost function (59) for any \(k,1\le k\le K\), we have
Consistent with (99), we assume that the prediction process by the economic agents ends at \(k=K\). Therefore, we set \(u_k=0\) for \(k>K\). From the system (11), (99) can be written as
where \(X_j=QA^{d-K-1+j}\), \(Y_j=A'Y_{j-1}A+\tau A_0'Y_{j-1}A_0\) with the initial value \(Y_{K-d+1}=R\), \(Z_j=(A^{d-K-1+j})'(Q-R)A^{d-K-1-j}\), and \(H_j=\sum ^{j+d-1}_{k=K+1}\eta (A^{d-k-1+j})'RA^{d-k-1+j}\). Now, we handle the optimization problem by dynamic programming procedure.
In the case of \(k=K-d+1\), we get
by virtue of (11) and (101), one has
According to (61), \({\hat{u}}_{K|j-1}\), \({\hat{x}}_{K|j-1}\) and \(u_j,j=K-d+1,\ldots ,K\) in (102) can be orthogonally decomposed as
Inserting (103)-(105) into (102) yields
where we have used the equalities (64)-(65) with \(k=K\) and the initial value \(M^{n}_{K+1}=\sum ^K_{j=K-d+2}Y_j+\sum ^K_{j=K-n+2}Z_j\), \(T^{n,j}_{K+1}=Q\) and \(S^{n,j}_{K+1}=X_j\).
By completion of squares over \({\hat{u}}_{K|K-d}\) and \({\hat{u}}_{K}|^{K-n}_{K-n-1},i=1,2,\ldots ,d-1\) in (106) results in
where
In the above, (62)-(67), and (70)-(71) are utilized with \(k=K\). From (108)-(109), the optimal \(u_K\) admits (46) with \(k=K\). The optimal cost of (107) is given as
In the case of \(k=K-d\), we have
This, together with (110), means that one only needs to resolve
where \(S^{d,K-d}_K=T^{d,K-d}_K=Q\). Because the right hand of (101) can be written as
which has the same form of (112), thus we complete the induction argument.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Song, L., Wang, H., Zhang, H. et al. Rational Expectations Models with Multiplicative Noise. J Optim Theory Appl 199, 233–257 (2023). https://doi.org/10.1007/s10957-023-02275-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10957-023-02275-4