數(shù)據(jù)與實證研究

數(shù)

據(jù)

與

實

證

研

究高

寧

博

士國泰安信息技術有限公司常務副總裁西安交通大學教授香港浸會大學商學院Honorary

Associate香港浸會大學公司管制與金融政策研究中心Research

Fellow什么是實證研究？––

以事實、實際情況和收集到的數(shù)據(jù)為對象，通過分析、計算、實驗、研究，解釋和預測會計金融實

務，回答“實際是什么”的問題。◎實證研究要求客觀、準確、理性的描述現(xiàn)實◎實證研究以解釋現(xiàn)實為目的，認為存在就是事實◎實證研究采用客觀中立的立場◎目前，在國際上，實證研究方法廣泛的應用在經(jīng)濟、金融、會計等社會學科的研究中實證研究的發(fā)展與趨勢----實證經(jīng)濟學1953弗里德曼《實證經(jīng)濟學方法論》發(fā)展歷程----實證會計學1968

Ball，R.J.,P.Brown《An

Empirical

Evaluation

of

AccountingIncome

Numbers》《Journal

of

Accounting

Research》1986

Watts，Zimmerman《實證會計理論》趨勢由于金融市場每天都產(chǎn)生海量的數(shù)據(jù)，這些數(shù)據(jù)又是從真實的交易

過程中產(chǎn)生的,這一特性使實證研究成為現(xiàn)代金融研究的主流話語”――Ross20世紀80年代《Accounting

Review》上實證性研究的論文占半數(shù)以上，有的年份還高達81％?，F(xiàn)在實證研究已成為會計，金融研究的主流。推動實證研究發(fā)展的因素（William

Beaver)推動實證研究發(fā)展的因素（William

Beaver)財務和經(jīng)濟學的發(fā)展1證券市場在經(jīng)濟中的地位2政府對證券市場的積極監(jiān)管，不斷推出新的課題3機構投資者占股權比重的增大4計算機技術和數(shù)據(jù)庫的發(fā)展56學術刊物受重視程度的增強實

證

論

文

篇

數(shù)實證論文篇數(shù)類型1994199519961997199819992000200120022003200420052006實證研究論文13616120532538154764183011531751248235665043經(jīng)濟類實證論文8272115168188236276335432697102613991979實

證

的

要

素實證的要素結論推理檢驗假設模型數(shù)據(jù)實證的要素數(shù)據(jù)：反映客觀狀況的數(shù)字材料。模型：刻畫客觀現(xiàn)象的數(shù)學形式。假設：對所研究問題的結果或狀態(tài)的◆一種預期。檢驗：利用數(shù)據(jù)，使用統(tǒng)計學知識對假設的統(tǒng)計顯著性作出判斷。推理：基于知識和經(jīng)驗對假設檢驗結果進行推理。結論：利用假設檢驗的結果，通過合情的邏輯推理得出的結論，觀點。實證研究方法步驟確立研究課題實

證

研

究

方

法

步

驟尋找相關理論提出命題假設設計研究方案搜集事實數(shù)據(jù)分析數(shù)據(jù)檢驗命題得出研究結論金

融

實

證

研

究的

主

要

領

域投資組合選擇和資產(chǎn)定價–包括現(xiàn)代投資組合理論、資本資產(chǎn)定價理論、套利定價模型、期權定價模型、有效邊界、資本市場線、證券市場線等。資金成本和資本結構理論–包括資金成本傳統(tǒng)理論、凈利理論和營業(yè)凈利理論、權衡理論和融資偏好次序等。市場微觀結構–研究交易價格發(fā)現(xiàn)過程與交易運作機制，包括價格發(fā)現(xiàn)的模型和市場結構與設計。行為金融學–研究投資者的心理、個人特征等因素與其交易行為之間的關系，包括個人信仰(過度自信、樂觀主義、代表性、保守主義、確認偏誤、定位、記憶偏誤),個人偏好(展望理論、模糊規(guī)避)會

計

實

證

研

究的

主

要

領

域會計制度的選擇–研究企業(yè)會計制度的選擇與企業(yè)營運績效之間的關系盈余管理–研究企業(yè)管理當局借助會計政策的選擇和會計估計的變更，尋求對自己有利結果的行為及其影響會計舞弊–研究公司采取偽造、掩飾的手法編造假賬損害股東權益、影響投資者做出正確投資決策的行為財務預測–研究如何根據(jù)財務活動的歷史資料和現(xiàn)實情況對企業(yè)未來財務活動進行科學的預計和測算會計信息披露效應–研究上市公司會計信息披露與公司股票價格之間的關系財務困境–研究企業(yè)陷于財務困境的特征及影響因素主要包括財務困境企業(yè)與非財務困境企業(yè)之間財務項目的分析會計信息的價值相關性–研究會計信息價值相關性對于會計準則制證券市場監(jiān)管和投資者進行決策的作用CSMAR

實

證論

文

舉

例文章研究了中國上市公司盈余公告時間選擇對股票交易量和未預期收益的影響。研究發(fā)現(xiàn)，與較晚月份公告盈余的公司相比，較早月份進行年度盈余公告的公司具有較強的股票交易量反應。文章認為愿意早些公告盈余的公司往往擁有利好的信息，并且這些較早的盈余公告含有更大的信息量,帶來較大的交易量增幅和未預期收益；較晚公告盈余的公司則往往擁有利差的信息，而且更容易被市場預期,因而帶來的交易量增幅和未預期收益也較小。作

者發(fā)表刊物摘

要題

目

InformationContent

and

Timing

ofEarnings

Announcements陳工孟

高

寧 鄭子云（香港理工大學）Journal

of

Business

Finance

and

Accounting,

January

2005,

Vol

3Iss.

1-

2,

Pg.

65－95數(shù)

據(jù)

樣

本以1995年至2002年間發(fā)行A股或同時發(fā)行A，B股，在時間區(qū)間內(nèi)發(fā)表年度盈余公告的上市公司為研究樣本。樣本容量為3802。年份樣本數(shù)１月（％）２月（％）３月（％）4月（％）19952656(2.26)9(3.40)81(30.57)169(63.77)19962941(0.34)6(2.04)33(11.22)254(86.40)19973504(1.14)10(2.86)52(14.86)284(81.14)19985904(0.68)45(7.63)269(45.59)272(46.10)19993508(2.28)9(2.57)87(24.86)246(70.29)200053145(8.47)50(9.42)188(35.41)248(46.70)200166313(1.96)108(16.29)299(45.10)243(36.65)200275915(1.98)84(11.07)277(36.50)383(50.45)Total380296(2.52)321(8.45)1286(33.82)2099(55.21)CSMAR

總

體樣

本

容

量CSMAR

總

體

樣

本

容

量文

獻

回

顧和

假

設為什么選交易量而不是價格Bamber,

Barron

and

Stober

(1997)

suggest

that

trading

volume

is

relateto

the

magnitude

of

the

disagreement

among

investors

about

a

firm’searnings.Kim

and

Verrecchia

(1991a)

argue

that

price

changes

reflect

the

averagchange

in

the

aggregate

market’s

average

beliefs,

while

trading

volumis

the

sum

of

all

individual

investors’

trades,

which

also

depends

onprevailing

information

asymmetry

level

before

disclosure.

They

suggethat

although

all

investors

have

equal

access

to

public

pre-disclosureinformation,

they

acquire

private

pre-disclosure

information

withdifferent

degrees

of

precision.為什么選交易量而不是價格Atiase

and

Bamber

(1994)

and

Kross

et

al.

(1994)suggest

that

trading

volumeincreasing

function

of

the

degree

of

divergent

pre-disclosure

expectatiBamber

and

Cheon

(1995)

argue

that

the

reason

for

different

reactions

is

threactions

reflect

the

average

belief

revision,

while

trading

volume

ariindividual

investors

make

differential

belief

revisions.更

進

一

步

的

分

析Kim

and

Verrecchia

(1994)

suggest

that

there

may

be

more

information

asymmeat

the

time

of

an

announcement

than

in

a

non-announcement

period.

This

isbecauseearnings

announcements

provide

information

that

allows

certain

tradersjudgements

about

a

firm’s

performance

that

are

superior

to

the

judgemenothertraders.Lobo

and

Tung

(1997)

find

that

the

trading

volume

around

quarterly

earningsannouncements

is

related

to

the

level

of

pre-disclosure

information

asymForfirms

with

a

high

level

of

pre-disclosure

information

asymmetry,

the

travolumeis

low

prior

to

and

after

the

announcement,

but

high

during

the

announceme更

進

一

步

的

分

析Bamber(1986)

employs

the

divergence

of

earnings

forecasts

from

analysts’

forecasts

as

a

proxy

forinformation

asymmetry.

She

finds

thatthe

higher

the

information

asymmetry,

the

greater

the

abnormalreaction.In

this

study,

we

first

use

unexpected

earnings

as

a

control

variable

for

information

asymmetry.Earlier

announcements

should

generate

a

greater

surprise

in

the

market

because

it

is

more

difficult

toearlier

announcements

than

later

announcements.

Chambers

and

Penman

(1984)

argue

that

longerreportinglags

provide

the

opportunity

for

more

of

the

report’s

information

to

be

supplied

by

other

sourcethrough

search

activity

by

investors,

through

other

voluntary

disclosures

by

firms,

or

through

prthatare

supplied

in

the

earnings

releases

of

earlier

reporting

firms.Haw

et

al.

(1999)

study

the

Chinese

stock

market

and

findthat

firms

withgoodnews

publicize

their

annreports

earlier

thanthose

withbad

news,

and

loss-making

firms

are

the

lastto

release

their

annual

reThey

define

the

reporting

lag

as

the

number

of

days

from

the

fiscal

year-end

to

the

report

announcementEarlier

announcements

should

generate

a

greater

surprise

in

the

market

because

it

is

more

difficult

to

predict

earlier

announcements

than

later

announcements.

Chambersand

Penman

(1984)

argue

that

longer

reporting

lags

provide

the

opportunity

for

more

ofthe

report’s

information

to

be

supplied

by

other

sources,

either

through

search

actiby

investors,

through

other

voluntary

disclosures

by

firms,

or

through

predictions

tharesupplied

in

the

earnings

releases

of

earlier

reporting

firms.Haw

et

al.

(1999)

study

the

Chinese

stock

market

and

find

that

firms

with

good

newspublicize

theirannual

reports

earlier

than

those

with

bad

news,

and

loss-making

fi

are

the

last

to

release

theirannual

reports.

They

define

the

reporting

lag

as

thenumber

of

days

from

the

fiscal

year-end

to

thereport

announcement

date.更

進

一

步

的

分

析1.

First,

normally

due

to

potential

insitrading

and

information

leakage,

it

ispossible

that

the

market

reaction

stalong

before

the

actual

announcements.Consequently,

we

employ

[-20,

2]and

[-20,

-3]

to

capture

the

possible

pevent

reaction.2.

Second,

in

the

relatively

efficient

markannouncement

effects

shouldnot

exist

inlong

event

window.

Therefore,

we

use

fourshort

symmetrical

event

windows

to

capturannouncement

effects.They

are

[-1,

+1],+2],

[-5,

+5],

and

[-7,

+7].時間窗口的確定[-20,

2][-20,

-3][-1,

+1][-2,

+2][-5,

+5][-7,

+7]共6個250

trading

days

from

day

–280

to

day

–31.A

time

gap

between

the

end

of

the

estimation

window

and

the

begiof

the

event

window

(i.e.

from

day

–30

to

day

–21)

is

employedusing

unusualpriceor

volume

data

(due

to

information

leaka-gemodel

estimation.d比較期間（beta期間）To

focus

our

analysis

on

the

number

of

tradable

days,

we

define

the

reporting

lagthe

number

of

working

days

from

the

fiscal

year-end

to

the

annual

release

date.–

1.

a

continuous

variable,

Announcement

Timing

Index

(ATI),

to

proxy

the

reporting

lag,which

isdefined

as

ATI

=

n/N,

where

n

is

the

nth

working

day

from

January

1

on

whichthe

earnings

announcement

is

made.N

is

the

total

number

of

working

days

in

the

periodfrom

January

1

to

April

30

inthe

event

year.三個不同的時間變量(TEA)定義三個不同的時間變量(TEA)定義the

unexpected

ATI

(UATI),

a

proxy

for

the

unexpected

reporting

lag,

is

def

as

the

difference

between

the

actual

and

expected

ATI

(the

expected

ATI

of

the

current

year

should

be

the

same

as

the

ATI

of

the

previous

year),

UATI

=

ATIt

–

ATIt-1.The

final

TEA

is

a

dummy

variable,

called

MAD,

with

a

value

of

1

for

Marchand

April

announcements

and

0

otherwise.Null

Hypothesis:

Firms

with

earlier

and

laterearnings

announcements

should

receive

similarabnormal

market

reaction.簡單的假設Alternative

Hypothesis:

Firms

with

earlierearnings

announcements

should

receive

a

higherabnormalmaket

reaction.主

要

模

型■主

要

模

型tt異常交易量的決定因素多變量回歸模型CATV

(CAR)

=

0

+

1UEA

(UERW,

UEGM)+

2SIZE

+

3POWN

+

4

TEAt(UATI,

ATI,

MAD)+

5EXCH

+

iYEARi-5

+

jINDj-12

+18FORCATVPOWNUEAEXCHINDSIZETEAYEARFORCAR累積異常交易量累積異常收益率未預期盈余的絕對值人民幣計價的總資產(chǎn)的自然流通股所占百分比盈余公告時間交易所啞變量公告年的啞元變量行業(yè)啞變量外資股的啞變量Abnormal

Trading

Volume

around

EarningsAnnouncement

by

bi-monthly

sampleJanuaryand

February

(#

Obs

=

417)March

and

April

(#

Obs

=

3385)DayATVt-valueATVt-value-70.00151.640.00071.63-60.00242.39*0.00102.12*-50.00242.25*0.00091.86-40.00443.40**0.00071.61-30.00453.59**0.00112.38*-20.00554.41**0.00102.05*-10.00926.25**0.00193.78**00.01347.87**0.007112.02**+10.01297.63**0.007112.24**+

20.00915.62**0.00366.95**+

30.00554.05**0.00183.61**+

40.00322.64**0.00081.67+

50.00301.860.00061.31+

60.00181.440.00091.89+

70.00201.580.00102.02*IntervalCATVz-valueCATVz-value[-20,2]0.0841a13.32

**0.0380a15.67**[-20,-3]0.0340b7.57**0.0173b8.98**[-7,7]0.0808c14.62**0.0302c14.76**[-5,5]0.0731d14.94**0.0266d14.92**[-2,

2]0.0501e14.21**0.0207e16.57**[-1,

1]0.0355f12.56**0.0161f16.19**Abnormal

Trading

Volume

around

EarningsAnnouncement

by

bi-monthly

sample◎

Most

of

the

ATVs

for

all

monthly

samplesaresignificant,

whindicates

that

the

announcements

do

provide

information

tomarket.◎

The

magnitudesofthe

ATVs

and

CATVs

for

the

January

andFebruary

sample

are

much

greater than

those

for

the

March

anApril

sample.Lowest

40%

ofATI

SampleHighest

40%of

ATI

SampleDifferenceMean

CATVCATV30.02530.01410.0112cdCATV50.03370.01820.0155cdCATV110.04780.02290.0249cdCATV150.05450.02580.0287abCATV180.02650.0298-0.0033CATV230.06020.04790.0123Panel

A

:

Between

the

Lowest

40%

of

the

ATI

Sampleand

Highest

40%

of

the

ATI

SamplePositive

UATSampleNegative

UATSampleDifferenceMean

CATVCATV30.01060.0290-0.0184cCATV50.01320.0413-0.0281cCATV110.01600.0631-0.0471cCATV150.01660.0755-0.0589cCATV180.01100.0407-0.0297aCATV230.02420.0820-0.0578cPanel

B:

Between

the

Positive

UATI

Sampleand

Negative

UATI

SampleThe

lowest

40%

of

ATI

samples

demonstrates

a

significantly

greatvolumereaction

than

those

of

the

highest

40%

of

ATI

samples.The

negative

UATI

samples

demonstrate

a

significantly

greatervolume

reaction

than

those

of

the

positive

UATI

samples.earlier

announcements

provide

more

information

content

to

themarket

than

later

announcements

do.CATV3CATV5CATV11CATV15Intercept0

.15900

.24800

.47600

.7070(4

.12

)**(4

.28

)**(4

.40

)**(5

.16

)**UERW0

.00050

.00100

.00180

.0023(2

.42

)*(3

.19

)**(3

.25

)*(3

.21

)**SIZE-0

.0068-0

.0110-0

.0228-0

.0341(-3

.31

)**(-3

.57

)**(-3

.96

)**(-4

.67

)**POWN-0

.0052-0

.0105-0

.0085-0

.0362(-0

.43

)(-0

.57

)(-0

.25

)(-0

.83

)UATI-0

.0282-0

.0384-0

.0568-0

.0596(-3

.47

)**(-3

.14

)**(-2

.48

)*(-2

.06

)*EXCH0

.00820

.01590

.03360

.0392(2

.37

)*(3

.05

)**(3

.45

)**(3

.19

)**YEAR2-0

.0410-0

.0623-0

.1090-0

.1420(-5

.69

)**(-5

.74

)**(-5

.36

)**(-5

.53

)**YEAR30

.01170

.01070

.00780

.0057(1

.72

)(1

.04

)(0

.41

)(0

.24

)YEAR4-0

.0596-0

.0941-0

.1800-0

.2580–

Results

of

Regression

Model

for

CATVYEAR

5-0.0397-0.0604-0.1050-0.1550(-3.52)**(-3.56)**(-3.31)**(-3.88)**YEAR

6-0.0599-0.0893-0.1560-0.2180(-5.59)**(-5.54)**(-5.16)**(-5.71)**YEAR

7-0.0592-0.0877-0.1590-0.2230(-5.73)**(-5.63)**(-5.46)**(-6.07)**IND

10.05490.07530.15400.2190(2.84)**(2.59)**(2.83)**(3.19)**IND

20.0000-0.00190.0010-0.0011(-0.01)(-0.20)(0.06)(-0.04)IN