\documentclass[10pt,a4]{article}
\usepackage{fullpage}
\usepackage{setspace}
\usepackage{parskip}
\usepackage{titlesec}
\usepackage[section]{placeins}
\usepackage{xcolor}
\usepackage{breakcites}
\usepackage{lineno}
\usepackage{hyphenat}
\PassOptionsToPackage{hyphens}{url}
\usepackage[colorlinks = true,
linkcolor = blue,
urlcolor = blue,
citecolor = blue,
anchorcolor = blue]{hyperref}
\usepackage{etoolbox}
\makeatletter
\patchcmd\@combinedblfloats{\box\@outputbox}{\unvbox\@outputbox}{}{%
\errmessage{\noexpand\@combinedblfloats could not be patched}%
}%
\makeatother
\usepackage{natbib}
\renewenvironment{abstract}
{{\bfseries\noindent{\abstractname}\par\nobreak}\footnotesize}
{\bigskip}
\titlespacing{\section}{0pt}{*3}{*1}
\titlespacing{\subsection}{0pt}{*2}{*0.5}
\titlespacing{\subsubsection}{0pt}{*1.5}{0pt}
\usepackage{authblk}
\usepackage{graphicx}
\usepackage[space]{grffile}
\usepackage{latexsym}
\usepackage{textcomp}
\usepackage{longtable}
\usepackage{tabulary}
\usepackage{booktabs,array,multirow}
\usepackage{amsfonts,amsmath,amssymb}
\providecommand\citet{\cite}
\providecommand\citep{\cite}
\providecommand\citealt{\cite}
% You can conditionalize code for latexml or normal latex using this.
\newif\iflatexml\latexmlfalse
\providecommand{\tightlist}{\setlength{\itemsep}{0pt}\setlength{\parskip}{0pt}}%
\AtBeginDocument{\DeclareGraphicsExtensions{.pdf,.PDF,.eps,.EPS,.png,.PNG,.tif,.TIF,.jpg,.JPG,.jpeg,.JPEG}}
\usepackage[utf8]{inputenc}
\usepackage[ngerman,english]{babel}
\begin{document}
\title{Analysts' forecasts and stock returns: an empirical assessment I.}
\author[1]{Nelson Seixas dos Santos}%
\affil[1]{Universidade Federal do Rio Grande do Sul (UFRGS)}%
\vspace{-1em}
\date{\today}
\begingroup
\let\center\flushleft
\let\endcenter\endflushleft
\maketitle
\endgroup
\newpage
\sloppy
\begin{center}
Associate Professor\\
Departamento de Economia e Rela\selectlanguage{ngerman}ções Internacionais\\
Faculdade de Ci\selectlanguage{ngerman}ências Econômicas\\
Universidade Federal do Rio Grande do Sul\\
Av. Jo\selectlanguage{ngerman}ão Pessoa, 52, 1o. andar, sala 18-D.\\
Centro, Porto Alegre, Rio Grande do Sul, Brazil\\
CEP 90040-000\\
Tel: +55(51)3308-3332\\
e-mail:\href{mailto:nelson.seixas@ufrgs.br}{nelson.seixas@ufrgs.br}\\
URL:\href{http://professor.ufrgs.br/nelsonseixas}{http://professor.ufrgs.br/nelsonseixas}\\
\\
Research Project Submitted and to The Alfred P. Sloan School of Management to apply to a Visiting Scholar Position with Professor Rodrigo Verdi.
\end{center}\selectlanguage{english}
\begin{abstract}
This project aims to proceed an empirical investigation into the impact of analysts' forecasts on stock returns. To do so we will collect actual and target prices data from NYSE market makers and, in accordance with the interpretation that target prices comes from the estimation of Lucas asset pricing model, we will test for Granger causality between actual and target prices. In addition, we will collect US consumption and dividend from major stocks traded at NYSE in order to fit a Lucas asset pricing model taking target prices and Fama and French factors as instruments to investigate the relationship between analysts's forecasts (target prices), Fama and French factors, risk aversion, individual discount and stock returns in dynamic setting.
\end{abstract}
\par\null\par\null
\section{The Research Problem and its
Importance.}
{\label{510325}}
The role of information in determining the dynamics of stock returns is
one of the cornerstone problems in financial economics
since~\hyperref[csl:1]{(Bachelier, 1900)}. ~His results on the irrelevance of past
information (including accounting one) to forecast stock returns were
more widely known as efficient market hypothesis
after~\hyperref[csl:2]{(Fama, 1965)} and~\hyperref[csl:3]{(Fama, 1970)}. Actually, that means
any information is incorporated instantaneously in stock prices so that
no information set can help forecasting stock prices. For
instance,~\hyperref[csl:4]{(Marques \& dos Santos, 2016)} questioned whether political news could
affect stock returns in Brazil and, after performing a thorough
webscrapping to get news from most important media websites, found
nothing but Presidential election itself seemed to have influenced stock
returns despite they had even a death of one of the top Presidential
candidates in the middle of campaign and Presidential impeachment
process in their sample.
In accounting literature, though, the answer to Fama's challenging
propositions came soon as~\hyperref[csl:5]{(Beaver, 1968)} and~\hyperref[csl:6]{(Ball \& Brown, 1968)}
found evidence that informational content of accounting data can to
change investors' expectations and, then, stock returns. ~Both
approaches - informational content and market efficiency - has brought
on an extensive literature as shown by~\hyperref[csl:7]{(Malkiel, 2003)}
and~\hyperref[csl:8]{(Kothari, 2001)}. ~These conflicting viewpoints nevertheless seem
to have been converging as~\hyperref[csl:9]{(Merton, 1973)}
and~\hyperref[csl:10]{(Lucas, 1978)}~established the importance of agents'
information set on describing martingale properties of stock prices. ~In
addition,~\hyperref[csl:11]{(Hansen \& Richard, 1987)}~showed the conditioning information has a
role in estimating stock returns and~\hyperref[csl:12]{(Ross, 1976)} provided the
theoretical fundamentals on which~\hyperref[csl:13]{(Fama \& French, 1993)}~ built an
empirical model which found evidence balance sheet numbers help pricing
securities. Also, ~\hyperref[csl:14]{(Ohlson, 1995)}~includes accounting variables in
a classical expected discounted dividend model to develop his valuation
model.
Finally,~\hyperref[csl:15]{(Easley et al., 1996)},~\hyperref[csl:16]{(Easley et al., 1997)},~\hyperref[csl:17]{(Easley et al., 2002)}~and~\hyperref[csl:18]{(Easley \& O'hara, 2004)}
have given an~theoretical explanation~ of the importance of accounting
numbers to correctly evaluate stock returns and also provided evidence
of it allows us to unite both approaches. Recently, dealing with
Brazilian data~\hyperref[csl:19]{(de Andrade \& dos Santos, 2017)}~applied panel data and vector auto
regressions methods to investigate impact of current ratio, earnings per
share and book value per share on Brazilian stock returns and found no
evidence of it. ~
~In particular,~\hyperref[csl:20]{(Kothari et al., 2016)}~surveyed the literature on analysts'
forecast and asset pricing and presented ~models which intended to link
analysts' forecasts and expected returns in valuation framework (dynamic
models) and in asset pricing framework (static models). ~The dynamic
models they presented are derived from ~\hyperref[csl:21]{(Ohlson, 1995)} while the
static ones descend from~\hyperref[csl:18]{(Easley \& O'hara, 2004)}
and~\protect\hyperlink{Sharpe_1964}{Sharpe (1964)}. ~
Notwithstanding~\hyperref[csl:20]{(Kothari et al., 2016)}~argue models in valuation framework
are more successful in providing estimations for expected returns
proxies than asset price ones, as it can be seen, for instance,
in~\hyperref[csl:22]{(Gebhardt et al., 2001)}~and~\hyperref[csl:23]{(Pastor et al., 2008)}, those approaches have
two main problems.~
The first one is that they are not akin to be easily statistically
evaluated, because they use simulation methods instead of statistical
inference.~ The second problem concerns the fact that the validity of
Ohlson's model - as any other discounted cash flow model - rests upon
the hypothesis individuals are risk neutral which forbids us to inquire
the role of individual risk preferences in expected returns.~
This main objective of this project is to investigate empirically the
impact of analysts' forecasts on expected and actual stock returns and a
secondary aim is to evaluate the information content of Fama and French
factors to forecast stock returns all in a generalized dynamic set up
where risk preferences and individual discount rates might have a role.
~Actually, we follow the lead of~\hyperref[csl:20]{(Kothari et al., 2016)} who concluded
~``{[}\ldots{}{]} the current state of literature presents a promising
opportunity for future research.'' (p. 209), that ``although the
implications of analysts' forecasts to cash flows is clear and the
empirical evidence is vast, the links between analysts' forecasts and
expected returns are less established.'' and later go further saying
``Evidence on the link between analysts' forecasts and expected returns
is relatively scarce'' (p. 212). ~
\par\null
\section{Theoretical Background and Main Empirical
Hypotheses}
{\label{753571}}
From a theoretical viewpoint, in equilibrium agents' expectations
collapse into actual prices as clearly posed in~\hyperref[csl:10]{(Lucas, 1978)}
and~\hyperref[csl:24]{(Breeden, 1979)}. ~They say nothing though about the role of
accounting numbers in the formation of those expectations. ~Actually,
they are compatible with~\hyperref[csl:3]{(Fama, 1970)}~ semi-strong market
efficiency hypothesis where all public information is somehow already
into market prices, which doesn't leave room for any kind of forecast
based on accounting information. ~On the other side, the empirical
literature derived from~\hyperref[csl:25]{(Fama \& French, 1992)} and
~\hyperref[csl:13]{(Fama \& French, 1993)}~finds price effects of accounting indices in
expectation of returns while ~\hyperref[csl:14]{(Ohlson, 1995)}~develops a model where
accounting data matters in a, as the author says, ``neoclassical
framework'' (p. 662), which means in his terms that ``value equals the
present value of expected dividends'' (p. 662).
Recently,~\hyperref[csl:26]{(Ghosh et al., 2017)}~ incorporated~\hyperref[csl:13]{(Fama \& French, 1993)}~into
~\hyperref[csl:10]{(Lucas, 1978)} dynamic model of asset pricing in order to
factorize the stochastic discount factor in business cycle and Fama and
French factors leaving room for macroeconomic and accounting factors to
affect asset returns. Our approach to solve the problem we posed is to
substitute for Ohlson's model for Lucas setting which means the pricing
equation now is the following:
\(p_{t}= E_{t}\left\{ \beta. \frac{u'(c_{t+1})}{u'(c_{t})}.\left(p_{t+1} + d_{t+1}\right) | I_{t} \right\}\)
where~\(p_{t}\) is stock price,~\(\beta\) is
individual inter-temporal discount,~\(d_{t+1}\)~is dividend paid
at period ~t+1 and~\(I_{t}\) is the information set available
to agents at period t. The whole equation means actual prices are the
expected value of future cash flows discounted by the marginal rate of
inter-temporal substitution given the information set available.
Nevertheless, since we aim to a different target
than~\hyperref[csl:26]{(Ghosh et al., 2017)}, we will explore this equation in different way
than they did. ~We first notice that, in Lucas~setting, we can have a
twofold interpretation of analysts' forecasts like it's shown below:
\textbf{1. First Interpretation}
In the first interpretation we suppose analysts' forecasts are in
agents' information set which implies strong correlation and even
causality in~the sense of~\hyperref[csl:27]{(Granger, 1969)} between forecasts and
actual prices. ~ Formally, the interpretation means:
~\(p_{t}= E_{t}\left\{ \beta. \frac{u'(c_{t+1})}{u'(c_{t})}.\left(p_{t+1} + d_{t+1}\right) | \left\{p^{T}\right\} \cup I_{t} \right\}\)
where~\(p^{T}\) means analysts' forecast for prices, that is,
target prices and~\(I_{t}\) means the set which elements are
any other variable which helps but not target prices .
\textbf{2. Second Interpretation}
The~second interpretation rests on the fact all agents are equal in
Lucas setting and then the analysts are themselves the agents who is
solving Lucas problem. ~Consequently, in the second interpretation,
actual prices should converge in distribution to analysts' forecasts.
~Mathematically, we can express it as:
\(p_{t}= E_{t}\left\{ \beta. \frac{u'(c_{t+1})}{u'(c_{t})}.\left(p^{T}_{t+1} + d_{t+1}\right) | I_{t} \right\}\)
In the following we'll show how we'll use both interpretations above to
test for the relation of analysts' forecast ~and expected stock returns.
\section{Methods and Data}
{\label{333181}}
The empirical strategy consists in two steps. ~The first one is to
collect actual prices and target prices from analysts of Brendan E.
Cryan \& Co., Citadel Securities LLC, GTS Securities LLC, IMC Financial
Markets, and Virtu Financial Capital Markets LLC, which are market
makers in New York Stock Exchange (NYSE). ~Then I will investigate the
first interpretation of Lucas model, that is, we will consider target
prices as the result of an estimation of Lucas asset pricing model and
test for causality in the sense of~\hyperref[csl:27]{(Granger, 1969)} between those
actual and target prices.
In the second step we'll take for granted the second interpretation .
~To do so we'll collect consumption and dividend data for each stock
traded at NYSE for which there is a target price from one of those
market makers to investigate . ~Then, we'll estimate by generalized
method of moments (GMM) the parameters risk aversion and inter-temporal
discount of~Lucas model of second interpretation, that is, imposing
target prices are actual prices for each stock. with and without target
price data as instruments for each one of them. ~That way we will be
able to infer for each stock whether risk aversion and discount rates
might be a channel through which accounting may impact on stock returns.
Besides, we will surpass the lack of statistical confidence metrics in
the aforementioned approaches.
The third step will be estimate the basic pricing equation without
target prices and compare the results with results of first
interpretation in order to evaluate how an economy populated by
uninformed agents (Lucas model without target prices) compares with an
economy populated by informed agents in terms of risk aversion and and
inter-temporal discount in both setups.~
All data analysis will be performed with Python programming language and
its scientific stack including the RPy2 communication package necessary
to call R environment which will be used to estimate GMM specifications.
~Consequently, the paper will be fully reproducible for anyone and
efficient way. ~All code will be available in GitHub.
\section{Expected Results and Main
Hypothesis}
{\label{338226}}
The main hypothesis we have in the first interpretation of the model is
that target prices Granger cause actual prices. ~In terms of the second
interpretation, we expect we have no statistical difference between
parameters of basic pricing equations and parameters estimated in second
interpretation. ~Also, we expect to find uninformed traders are more
risk averse, their forecasts are more volatile.
\section{Conclusion}
{\label{711793}}
This project presented a research intended to evaluate the impact of
analyst's forecasts on expected stock returns in a dynamic setting where
we can investigate also how risk aversion might influence on it. ~To the
best of my knowledge this problem hasn't been dealt yet and this
contribution fill the gap hardly emphasized by~\hyperref[csl:20]{(Kothari et al., 2016)} as a
weak point in the literature. ~
An additional contribution is the comparison between economies with
informed and uninformed agents to evaluate the importance of analysts'
forecast in mitigating risk aversion and making agents more cautious
about the future.
~
\selectlanguage{english}
\FloatBarrier
\section*{References}\sloppy
\phantomsection
\label{csl:1}{Th{\'{e}}orie de la sp{\'{e}}culation}. (1900). \textit{Annales Scientifiques De l{\Textquotesingle}{\'{E}}Cole Normale Sup{\'{e}}Rieure}, \textit{17}, 21–86. \url{https://doi.org/10.24033/asens.476}
\phantomsection
\label{csl:2}{The Behavior of Stock-Market Prices}. (1965). \textit{The Journal of Business}, \textit{38}(1), 34. \url{https://doi.org/10.1086/294743}
\phantomsection
\label{csl:3}{Efficient Capital Markets: A Review of Theory and Empirical Work}. (1970). \textit{The Journal of Finance}, \textit{25}(2), 383. \url{https://doi.org/10.2307/2325486}
\phantomsection
\label{csl:4}{Do Political News Affect Financial Market Returns? Evidences from Brazil}. (2016). \textit{International Journal of Management, Accounting and Economics}, \textit{3}(10), 545–571. \url{http://www.ijmae.com/files/accepted/541_final.pdf}
\phantomsection
\label{csl:5}{The Information Content of Annual Earnings Announcements}. (1968). \textit{Journal of Accounting Research}, \textit{6}, 67. \url{https://doi.org/10.2307/2490070}
\phantomsection
\label{csl:6}{An Empirical Evaluation of Accounting Income Numbers}. (1968). \textit{Journal of Accounting Research}, \textit{6}(2), 159. \url{https://doi.org/10.2307/2490232}
\phantomsection
\label{csl:7}{The Efficient Market Hypothesis and Its Critics}. (2003). \textit{Journal of Economic Perspectives}, \textit{17}(1), 59–82. \url{https://doi.org/10.1257/089533003321164958}
\phantomsection
\label{csl:8}{Capital markets research in accounting}. (2001). \textit{Journal of Accounting and Economics}, \textit{31}(1-3), 105–231. \url{https://doi.org/10.1016/s0165-4101(01)00030-1}
\phantomsection
\label{csl:9}{An Intertemporal Capital Asset Pricing Model}. (1973). \textit{Econometrica}, \textit{41}(5), 867. \url{https://doi.org/10.2307/1913811}
\phantomsection
\label{csl:10}{Asset Prices in an Exchange Economy}. (1978). \textit{Econometrica}, \textit{46}(6), 1429. \url{https://doi.org/10.2307/1913837}
\phantomsection
\label{csl:11}{The Role of Conditioning Information in Deducing Testable Restrictions Implied by Dynamic Asset Pricing Models}. (1987). \textit{Econometrica}, \textit{55}(3), 587. \url{https://doi.org/10.2307/1913601}
\phantomsection
\label{csl:12}{The arbitrage theory of capital asset pricing}. (1976). \textit{Journal of Economic Theory}, \textit{13}(3), 341–360. \url{https://doi.org/10.1016/0022-0531(76)90046-6}
\phantomsection
\label{csl:13}{Common risk factors in the returns on stocks and bonds}. (1993). \textit{Journal of Financial Economics}, \textit{33}(1), 3–56. \url{https://doi.org/10.1016/0304-405x(93)90023-5}
\phantomsection
\label{csl:14}{Earnings Book Values, and Dividends in Equity Valuation}. (1995). \textit{Contemporary Accounting Research}, \textit{11}(2), 661–687. \url{https://doi.org/10.1111/j.1911-3846.1995.tb00461.x}
\phantomsection
\label{csl:15}{Liquidity Information, and Infrequently Traded Stocks}. (1996). \textit{The Journal of Finance}, \textit{51}(4), 1405–1436. \url{https://doi.org/10.1111/j.1540-6261.1996.tb04074.x}
\phantomsection
\label{csl:16}{The information content of the trading process}. (1997). \textit{Journal of Empirical Finance}, \textit{4}(2-3), 159–186. \url{https://doi.org/10.1016/s0927-5398(97)00005-4}
\phantomsection
\label{csl:17}{Is Information Risk a Determinant of Asset Returns?}. (2002). \textit{The Journal of Finance}, \textit{57}(5), 2185–2221. \url{https://doi.org/10.1111/1540-6261.00493}
\phantomsection
\label{csl:18}{Information and the Cost of Capital}. (2004). \textit{The Journal of Finance}, \textit{59}(4), 1553–1583. \url{https://doi.org/10.1111/j.1540-6261.2004.00672.x}
\phantomsection
\label{csl:19}{Accounting Information and Stock Returns: Evidences from Brazil}. (2017). \textit{International Journal of Management, Accounting and Economics}, \textit{4}(5), 443–485. \url{http://www.ijmae.com/files/accepted/645_final.pdf}
\phantomsection
\label{csl:20}{Analysts' Forecasts and Asset Pricing: A Survey}. (2016). \textit{Annual Review of Financial Economics}, \textit{8}(1), 197–219. \url{https://doi.org/10.1146/annurev-financial-121415-032930}
\phantomsection
\label{csl:21}{Earnings Book Values, and Dividends in Equity Valuation}. (1995). \textit{Contemporary Accounting Research}, \textit{11}(2), 661–687. \url{https://doi.org/10.1111/j.1911-3846.1995.tb00461.x}
\phantomsection
\label{csl:22}{Toward an Implied Cost of Capital}. (2001). \textit{Journal of Accounting Research}, \textit{39}(1), 135–176. \url{https://doi.org/10.1111/1475-679x.00007}
\phantomsection
\label{csl:23}{Estimating the Intertemporal Risk–Return Tradeoff Using the Implied Cost of Capital}. (2008). \textit{The Journal of Finance}, \textit{63}(6), 2859–2897. \url{https://doi.org/10.1111/j.1540-6261.2008.01415.x}
\phantomsection
\label{csl:24}{An intertemporal asset pricing model with stochastic consumption and investment opportunities}. (1979). \textit{Journal of Financial Economics}, \textit{7}(3), 265–296. \url{https://doi.org/10.1016/0304-405x(79)90016-3}
\phantomsection
\label{csl:25}{The Cross-Section of Expected Stock Returns}. (1992). \textit{The Journal of Finance}, \textit{47}(2), 427–465. \url{https://doi.org/10.1111/j.1540-6261.1992.tb04398.x}
\phantomsection
\label{csl:26}{What Is the Consumption-{CAPM} Missing? An Information-Theoretic Framework for the Analysis of Asset Pricing Models}. (2017). \textit{Review of Financial Studies}, \textit{30}(2), 442–504. \url{https://doi.org/10.1093/rfs/hhw075}
\phantomsection
\label{csl:27}{Investigating Causal Relations by Econometric Models and Cross-spectral Methods}. (1969). \textit{Econometrica}, \textit{37}(3), 424. \url{https://doi.org/10.2307/1912791}
\end{document}