Abstract
Using an affine model to compute the price of equities based on a dataset of macroeconomic factors, we propose a measure of equity bubbles. We use a dynamic affine term structure framework to price equity and bonds jointly, and investigate how prices are related to a set of macrofactors extracted from a large dataset of economic time series. We analyze the discrepancies between market and model implied equity prices and use them as a measure for bubbles. A bubble is diagnosed over a given period whenever the discrepancies are not stationary and impact the underlying economy consistently with the literature’s findings, increasing over the shorter term economic activity before leading to a net loss in it. We perform the analysis over 3 major US and 3 major European equity indices over the 1990–2017 period and find bubbles only for two of the US equity indices, the S&P500 and the Dow Jones.
Similar content being viewed by others
Notes
As Gürkaynak (2008) concludes his review of literature: “For every test that “finds” a bubble, there is another paper that disputes it. The finding of a bubble, at best, suggests that the data is either consistent with a bubble or a myriad of other extensions of the standard model”
Dividend strip refers to a price of a single dividend claim k periods into the future.
One of the advantages of the above factors specification is that they are fully observable and have an approximate economic interpretation. Piazzesi (2010) reports that the model-implied dynamics for macrovariables (i.e., latent factor estimated from the model) are often disconnected from the dynamics of their historical observable counterparts. Also, latent factors require computationally burdensome estimation techniques that account for the joint distribution of yields or price to dividend ratios and the state vector which can make our task significantly more difficult (see Ang and Piazzesi 2003; Kim 2007 on those difficulties). Finally, specifications with latent factors are not always globally identifiable and this issue in itself was closely scrutinized in the academia (Collin-Dufresne et al. 2008; Christensen et al. 2011; Joslin et al. 2011; Hamilton and Wu 2012).
Technical condition are described in Harrison and Kreps (1979).
For technical details, see “Appendix D and E” in Duffie (2004). By construction, \(\xi _t\) is a strictly positive martingale under Novikov’s condition with \(\xi _0 = 1\), implying that P and Q are equivalent and Q is a well defined probability measure.
\(r_t\) is an affine function of \(X_t\), \(X_t\) is also affine under the risk-neutral probability measure due to the specification of the \(X_t\) process as a Gaussian VAR, the SDF and the MPR specifications.
\(d_t\) is indexed one period ahead of the corresponding \(r_{t-1}\) due to the fact that \(r_t\) is a locally deterministic discount factor and is known at time t
See “Appendix B”.
All data used here have been obtained from Bloomberg©as a datasource.
Here, we use BEST estimates.
This type of quadratic criterion is equivalent to assuming that market quotes are a mixture between the “true” quote and an Gaussian disturbance, yielding unbiased estimates.
We acknowledge the fact that yields and price to dividend ratios have different scalings. However, we employ interior point method to minimize the objective function (38) which uses gradient methods and is relatively robust to variables scaling.
Despite imposing highly nonlinear restriction on the parameters, in general we find that they actually improve the estimation procedure by promptly rejecting parameter regions where dividend strips diverge in the limit and thus where parameters thus do not solve the minimization problem (38).
The need for this approach is in part motivated by high sensitivity of to the initial point specification, even if following steps are equal.
Ideally, we would shift the matrix of risk premia parameters in all possible directions to generate a full set of initial values and then run the estimation. However, due to high number of parameters estimated (20), this approach is infeasible. Generating random matrices around the admissible point adheres to the same idea but reduces number of directions initiated.
Similar results have been obtained with the KPSS and Phillips and Perron tests.
An interesting reader could use this test methodology in a rolling way to try differentiating the bubble periods from the non-bubble ones. Here, given the quarterly frequency we aim at using, this rolling estimation did not lead to interesting findings.
Consistently with the previously mentioned FAVAR literature.
We refrained from doing the same with the MSCI Germany and France as their interaction with the overall European economy makes the control variable selection more complex.
E.g., see Hamilton (1994).
References
Alessi L, Barigozzi M, Capasso M (2010) Improved penalization for determining the number of factors in approximate factor models. Stat Probab Lett 80(23):1806–1813
Ang A, Liu J (2001) A general affine earnings valuation model. Rev Acc Stud 6(4):397–425
Ang A, Piazzesi M (2003) A no-arbitrage vector autoregression of term structure dynamics and macroeconomic and latent variables. J Monet Econ 50:745–787
Ang A, Piazzesi M, Wei M (2006) What does the yield curve tell us about gdp growth? J Econom 131(1):359–403
Ang A, Ulrich M (2012) Nominal bonds, real bonds, and equity. In: Netspar discussion paper no. 12/2011-103
Bansal R, Yaron A (2004) Risks for the long run: a potential resolution of asset pricing puzzles. J Finance 59(4):1481–1509
Bauer MD, Rudebusch GD (2016) Resolving the spanning puzzle in macro-finance term structure models. Rev Finance 21(2):511–553
Beber A, Brandt MW, Luisi M (2015) Distilling the macroeconomic news flow. J Financ Econ 117(3):489–507
Bekaert G, Grenadier S (1999) Stock and bond pricing in an affine economy. NBER, Cambridge
Belo F, Collin-Dufresne P, Goldstein RS (2015) Dividend dynamics and the term structure of dividend strips. J Finance 70(3):1115–1160
Bernanke BS, Boivin J, Eliasz P (2005) Measuring the effects of monetary policy: a factor-augmented vector autoregressive (favar) approach. Q J Econ 120(1):387–422
Brunnermeier MK (2010) Bubbles, liquidity, and the macroeconomy. NBER Report Online 01(2):10–12
Brunnermeier MK, Oehmke M (2013) Bubbles, financial crises, and systemic risk. In: Handbook of the economics of finance, volume 2. Elsevier, pp 1221–1288
Chen L, Da Z, Priestley R (2012) Dividend smoothing and predictability. Manag Sci 58(10):1834–1853
Cheridito P, Filipović D, Kimmel RL (2007) Market price of risk specifications for affine models: theory and evidence. J Financ Econ 83(1):123–170
Christensen JH, Diebold FX, Rudebusch GD (2011) The affine arbitrage-free class of Nelson–Siegel term structure models. J Econom 164(1):4–20
Cochrane JH, Piazzesi M (2009) Decomposing the yield curve. In: Working paper
Collin-Dufresne P, Goldstein RS, Jones CS (2008) Identification of maximal affine term structure models. J Finance 63(2):743–795
Constantinides GM (1992) A theory of the nominal term structure of interest rates. Rev Financ Stud 5(4):531–552
Dai Q, Singleton KJ (2000) Specification analysis of affine term structure models. J Finance 55(5):1943–1978
Dai Q, Singleton KJ (2002) Expectation puzzles, time-varying risk premia, and affine models of the term structure. J Financ Econ 63(3):415–441
Duffee GR (2002) Term premia and interest rate forecasts in affine models. J Finance 57(1):405–443
Duffie D (2004) Credit risk modeling with affine processes. Scuola Normale Superiore, Pisa
Duffie D, Filipović D, Schachermayer W (2003) Affine processes and applications in finance. Ann Appl Probab 13:984–1053
Duffie D, Kan R (1996) A yield-factor model of interest rates. Math Finance 6(4):379–406
Eraker B (2008) Affine general equilibrium models. Manag Sci 54(12):2068–2080
Estrella A, Hardouvelis GA (1991) The term structure as a predictor of real economic activity. J Finance 46(2):555–576
Faust J, Rogers JH, Wright JH (2005) News and noise in G-7 GDP announcements. J Money Credit Bank 37(3):403–419
Fisher M, Gilles C (1998) Around and around: the expectations hypothesis. J Finance 53(1):365–383
Flannery MJ, Protopapadakis AA (2002) Macroeconomic factors do influence aggregate stock returns. Rev Financ Stud 15(3):751–782
Ghysels E, Horan C, Moench E (2014) Forecasting through the rear-view mirror: data revisions and bond return predictability. In: NY Fed Staff Reports
Grossman H, Diba B (1988) Explosive rational bubbles in stock prices? Am Econ Rev 78:520–30
Gürkaynak RS (2008) Econometric tests of asset price bubbles: taking stock. J Econ Surv 22(1):166–186
Hall SG, Psaradakis Z, Sola M (1999) Detecting periodically collapsing bubbles: a Markov-switching unit root test. J Appl Econom 14:143–154
Hamilton JD (1994) Time series analysis, vol 2. Princeton University Press, Princeton
Hamilton JD, Wu JC (2012) Identification and estimation of Gaussian affine term structure models. J Econom 168(2):315–331
Harrison JM, Kreps DM (1979) Martingales and arbitrage in multiperiod securities markets. J Econ Theory 20(3):381–408
Harvey CR (1988) The real term structure and consumption growth. J Financ Econ 22(2):305–333
Hirano T, Yanagawa N (2016) Asset bubbles, endogenous growth, and financial frictions. Rev Econ Stud 84(1):406–443
Hördahl P, Tristani O, Vestin D (2006) A joint econometric model of macroeconomic and term-structure dynamics. J Econom 131(1):405–444
Jardet C, Monfort A, Pegoraro F (2013) No-arbitrage near-cointegrated VAR (p) term structure models, term premia and GDP growth. J Bank Finance 37(2):389–402
Joslin S, Priebsch M, Singleton KJ (2014) Risk premiums in dynamic term structure models with unspanned macro risks. J Finance 69(3):1197–1233
Joslin S, Singleton KJ, Zhu H (2011) A new perspective on Gaussian dynamic term structure models. Rev Financ Stud 24(3):926–970
Kim DH (2007) Challenges in macro-finance modeling. In: Working paper
Kim DH, Wright JH (2005) An arbitrage-free three-factor term structure model and the recent behavior of long-term yields and distant-horizon forward rates. In: Working paper
Leary MT, Michaely R (2011) Determinants of dividend smoothing: empirical evidence. Rev Financ Stud 24(10):3197–3249
Lemke W, Werner T (2009) The term structure of equity premia in an affine arbitrage-free model of bond and stock market dynamics. European Central Bank, Frankfurt
LeRoy SF, Porter RD (1981) The present-value relation: tests based on implied variance bounds. Econometrica 49(3):555–574
Lettau M, Wachter J (2011) The term structures of equity and interest rates. J Financ Econ 101:90–113
Martin A, Ventura J (2012) Economic growth with bubbles. Am Econ Rev 102(6):3033–3058
McCracken MW, Ng S (2016) FRED-MD: A monthly database for macroeconomic research. J Bus Econ Stat 34(4):574–589
Mönch E (2008) Forecasting the yield curve in a data-rich environment: a no-arbitrage factor-augmented VAR approach. J Econom 146(1):26–43
Monfort A, Pegoraro F (2007) Switching varma term structure models-extended version. In: Working paper
Phillips PC, Shi S, Yu J (2015) Testing for multiple bubbles: historical episodes of exuberance and collapse in the S&P 500. Int Econ Rev 56(4):1043–1078
Piazzesi M (2010) Affine term structure models. Handb Financ Econom 1:691–766
Rudebusch GD, Wu T (2008) A macro-finance model of the term structure, monetary policy and the economy. Econ J 118(530):906–926
Shiller R (1981) Do stock prices move too much to be justified by subsequent changes in dividends? Am Econ Rev 71:421–436
Simon DP (2003) The Nasdaq volatility index during and after the bubble. J Deriv 11(2):9–24
Sornette D, Woodard R, Zhou W-X (2009) The 2006–2008 oil bubble: evidence of speculation, and prediction. Physica A 388(8):1571–1576
Stock JH, Watson MW (1998) Diffusion indexes. Technical report, National Bureau of Economic Research
Swanson E (2014) A macroeconomic model of equities and real, nominal, and defaultable debt. Unpublished Manuscript, University of California, Irvine
Trapani L (2017) A randomised sequential procedure to determine the number of factors. J Am Stat Assoc 113:1341–1349
Van Binsbergen J, Brandt M, Koijen R (2012) On the timing and pricing of dividends. Am Econ Rev 102(4):1596–1618
Van Binsbergen JH, Koijen RS (2017) The term structure of returns: facts and theory. J Financ Econ 124(1):1–21
West KD (1988) Bubbles, fads and stock price volatility tests: a partial evaluation. J Finance 43(3):639–656
Wu T (2001) Macro factors and the affine term structure of interest rates. In: Working paper
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
Florian Iepo declares that he has no conflict of interest. Mikita Knihain declares that he has no conflict of interest.
Ethical approval
This article does not contain any studies with human participants or animals performed by any of the authors.
Additional information
Communicated by M. Squillante.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix
Derivation of bonds recursive ODE equations
In this section, we derive a system of ODE equations that arises when pricing bonds in the ATS framework. Assuming that the price of a bond maturing in n years at time t, \(p_t^n\) is exponential affine in the state vector
we show that the price of the bond maturity in \(n+1\) years is also exponential affine as follows
Matching the coefficients on the RHS and LHS in Eq. (42) leads to the system of ODE equations
where we used a general specification of the short rate \(r_t = \delta _0 + \delta _1' X_t\). The boundary conditions are given by
Finally, short rate specification used throughout this work corresponds to the case when
Derivation of equity recursive ODE equations
In this section, we derive a system of ODE equations, arising when pricing price to dividend ratio (or normalized dividend strips) under risk-neutral probability measure Q.
The price of a single dividend strip that pays at time \(t + \tau \) is given by Eq. (32), which we reiterate here for convenience
The derivation starts with the guess that price of dividend strip is exponential affine in the state vector \(X_t\)
where \(X_t\) follows Gaussian VAR as given by Eq. (5). First, we notice that we can rewrite using the guess given by Eq. (50), the price of dividend strip recursively as follows
where we used the law of iterated expectations in the third equality and Eq. (50) in the last equality.
We assume that earnings logarithmic growth rates are affine in the state vector \(X_t\) (which implies that earnings themselves are exponential affine in the state vector \(X_t\))
where \(\gamma _0\), \(\gamma _1\) are constant parameters. Next, we assume that logarithmic growth rates of the payout ratio are also affine in the state vector \(X_t\)
This implies that dividend (log) growth rates \(d_t\) satisfy
where \(\omega _0 = \alpha _0 +\left( 1+\alpha _1 \right) \gamma _0\) and \(\omega _1 = \left( 1+\alpha _1 \right) \gamma _1\) are obtained from earnings and payout ratio growths loadings on the state vector factors. \(d_t\) is thus also affine in \(X_t\).
Having dividend growth rate process in place, we can further rewrite Eq. (51) as follows (directly under Q, using the same pricing kernel as for bonds pricing)
where we used the guess from Eq. (50) in the last equality and the fact that state vector follows conditional Gaussian VAR \(X_{t+1} = \mu ^Q + \Phi ^Q X_t + \Sigma \varepsilon _{t+1}\) under Q (under essentially affine risk premia specification used throughout this work).
Similar to bonds pricing, in order for the last equality to hold as an identity, coefficients \(a_n\) and \(b_n\) have to satisfy the following system of equations (obtained via coefficients matching in Eq. (55))
To obtain boundary conditions for \(a_n\) and \(b_n\), we note that for 1-period ahead dividend strip, we have (again, using Eq. (50))
as previously, last equality must hold as an identity. Matching the coefficients in Eq. (58), we obtain the following boundary conditions
The short rate specification used throughout this work corresponds to the case when
Parameter closed-form solutions and limiting behavior
In this section, we analyze the system of recursive ODE Eqs. (56), (57) derived in “Appendix B”. Equation (57) implies that \(b_n\) follows a simple recursive equation with initial boundary condition given by Eq. (60) and is independent of \(a_n\). Using the boundary condition, we can rewrite Eq. (57) as follows (we temporarily drop the superscript \(Q'\) from \(\Phi ^{Q'}\) to ease the notations)
where by convention we set \(\Phi ^0 = I\). Since the terms \(\delta _1\) and \(\omega _1\) are constant, last equation in Eq. (63) represents a finite sum of a geometric series, which admits a closed-form solution and is given by
In Eq. (64), we require that matrix \(I - \Phi \) is invertible. This condition is equivalent to absolute value of eigenvectors of matrix \(\Phi \) being different from one. This condition alone does not guarantee the convergence of \(\Phi ^n\) (and hence \(b_n\)). However, as long as absolute values of eigenvalues of matrix \(\Phi \) are also less than one, we have
which implies \(I - \Phi \) is invertible and that in the limit \(n \rightarrow \infty \) parameter \(b_n\) satisfies
and is also known as the fixed point in Eq. (57).Footnote 24 Importantly, restriction on eigenvalues is not the only restriction that can be imposed on matrix \(\Phi \) that leads to stable results for \(b_n\). These restriction present sufficient, but not necessary condition for \(b_n\) convergence.
Next, from Eq. (56), we have
Equation (67) provides an analytical expression for \(a_n\) in terms of model parameters, since the solution for \(b_n\) was derived in Eq. (64) and boundary condition \(a_1\) is given in Eq. (59).
As evident from Eq. (67), in order to analyze the limiting behavior of \(a_n\), we first have to analyze the limiting behavior of \(\sum _{k=1}^{n-1}b_{k}\). From Eq. (64), we have
where, as previously, we assume that matrix \(I-\Phi \) is invertible. Note that under the restriction that absolute values of eigenvalues of \(\Phi \) are less than one, using Eq. (65) in the limit we have
where two last terms in Eq. (69) is constant and two first terms increase linearly with n. This implies that generally, expression \(\sum _{k=1}^{n-1}b_k\) does not have a limit as \(n \rightarrow \infty \). The relation (69) is, however, useful to determine restrictions on parameters that determine limiting the dynamics of \(a_n\).
Finally, we analyze the limiting behavior of the term \(\sum _{k=1}^{n-1}b_k' \Sigma \Sigma ' b_k\) that appears in Eq. (67). First, we note that
where, following Eq. (64), we have
and therefore
which implies that
Since we are first and foremost interested in the limiting behavior of pricing parameters, in what follows we will only focus on terms that growth linearly with maturity n (and we will denote these terms with a subscript n for clarity). In fact, in the limit, terms of (74) either growth linearly with n or converge to constant values (under the assumption 65). We, therefore, focus on former terms, which characterize the limiting behavior of \(a_n\) and are necessary to derive the constraint that guarantees convergence of prices of dividend strips to zero.
which implies that
This allows us to analyze the limiting properties of \(a_n\) as \({n \rightarrow \infty }\)
We now recall that deterministic maturity-varying parameters \(a_n\) and \(b_n\) are used to price-dividend strips via Eq. (50). In order to ensure that prices of dividend strips converge to zero as time to maturity n approaches infinity, we impose the following restriction on the model parameters
which implies that in the limit
and since \(b_n\) and \(X_t\) are finite, as seen from Eq. (66), this implies for any time t the price of a (normalized) dividend strip satisfies
The price of equity (normalized by current dividend) is an infinite sum of individual dividend strips and is given by Eq. (37)
Finally, we note that condition (78) is also sufficient for equity-dividend ratio (81) to converge.
Tables and figures
Rights and permissions
About this article
Cite this article
Ielpo, F., Kniahin, M. Fundamental bubbles in equity markets. Soft Comput 24, 13769–13796 (2020). https://doi.org/10.1007/s00500-019-04514-1
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00500-019-04514-1