导图社区 CFA道德部分导图(适用于一二三级)
CFA道德部分超详细思路导图,汇总了Quant、Eqty、FSA、Eco、Deri、Alt、Prtfl、CF、Ethnics的内容,建议下载收藏。
编辑于2023-04-14 11:20:55 北京市CFA道德部分超详细思路导图,汇总了Quant、Eqty、FSA、Eco、Deri、Alt、Prtfl、CF、Ethnics的内容,建议下载收藏。
公司金融导图lecture01,内容有:The Four Types of Firms、Ownership Versus Control of Corporations、The Stock Market.
矩阵语言下的计量经济学思维导图,内容详细,包括:Lecture 1: Algebra and Geometry of OLS;Lecture 2: Properties of Residuals, Regression Fit, and Partitioned Regression;Lecture 3: Unbiasedness and Efficiency of OLS Estimator; Variance Estimation。
社区模板帮助中心,点此进入>>
CFA道德部分超详细思路导图,汇总了Quant、Eqty、FSA、Eco、Deri、Alt、Prtfl、CF、Ethnics的内容,建议下载收藏。
公司金融导图lecture01,内容有:The Four Types of Firms、Ownership Versus Control of Corporations、The Stock Market.
矩阵语言下的计量经济学思维导图,内容详细,包括:Lecture 1: Algebra and Geometry of OLS;Lecture 2: Properties of Residuals, Regression Fit, and Partitioned Regression;Lecture 3: Unbiasedness and Efficiency of OLS Estimator; Variance Estimation。
Fixed Income
LM1.The Term Structure and Interest Rate Dynamics 利率的期限结构
都可能考,无突出重点
1.Benchmark curve

1.1 Spot Rates
disct factor
spot yield curve
zero-coupon时 s=YTM
coupon-paying时 YTM是s1 s2 ... sn的某种平均
1.2 Forward Rates
Forward rates model (s到f的计算转换,画图分析更明晰)
Relationship between s and f (s↗时 有f>s;s↘时 有f<s)
Yield curve shapes
commonly upward: ①inf expectation ②risk premium
inverted yield curve
flat yield curve
humped yield curve
Forward Pricing Model:s和f相互求解

(不推荐用,而是用Forward rates model求倒)
Active Bond Portfolio Mgmt
根据s'与f的大小关系判断buy sell策略:s'<f,P↑,应该buy
Riding the yield curve (rolling down the yield curve)
前提:yield curve↗
策略:buy bond M > I.H. (Inv.Horizon)
1.3 Par Curve
定义:刚刚发行(on the run)的Par bond导出的TYM构成的yield curve
[计算] Par Rate = YTM = Coupon Rate (C/Par)
【待补充截图】
自助法反解Par Rate和Sopt Rate:1yr的PR=spot, 2yr的spot用2yr的PR和1yr的Spot求解

1.4 YTM, Spot Rate and Return on Bonds
TYM是spot rate的某种加权平均(TYM接近期限靠后的spot rate)
TYM的三个假设:1持有到期,2及时全额支付C和Par,3收到C用于reinv at YTM
YTM不适合作为estimate of expected return
而应该用realized return 或者 HPR
int rate volatile
yield curve sloped
risk of default
embedded option 不会持有到期
1.5 Swap Rate Curve in Valuation
[计算] Swap本质是 收fix支float,求解Vfix的CR即为Swap Rate
【待补充截图】
[优点]
①高Liquidity,无需很多借贷方参与,OTC市场;常用语hedge int rate risk
②Swap合约定制化,提供多种maturities的SR利率,弥补了国债市场不足
③含有Default risk,相比国债,更适合作为公司的benchmark rate
④OTC mkt,具有去监管特征
2.Yield Spread

2.1 Swap Spread
目标利率是SR
风险来源:default risk,mkt liquidity
2.2 I-Spread
I: 线性插值
*注意:目标利率是任何债,benchmark是SR
风险来源:default risk,mkt liquidity
2.3 Z-Spread
Z-spread = Sc - St 公司债-国债
风险来源:default risk,mkt liquidity,option risk
是衡量公司债的最准确spread
Zero int rate volatility:不符合实际,因为σ变动时,option value也在变,这时spread对于含权债券yi的估计不准确
2.4 TED Spread
风险来源:default risk,mkt liquidity
反应banking sector的default risk(同业拆借vs央妈要钱的溢价)
2.5 Libor–OIS Spread
Libor–OIS = LIBOR(90天) - OIS(隔夜SR)
OIS是无抵押的隔夜Swap index rate
Libor–OIS = LIBOR(90天) - SOFR(隔夜SR)
SOFR 有T-bill作抵押的隔夜利率
反应money mkt(短期市场)的maturity risk,因为两个利率的期限不一致
3.Traditional theories of the term structure of interest rates

3.1 Expectations Theory
Pure (Unbiased) Expectations Theory 远期利率是未来即期利率的无偏估计
Local Expectation Theory 短期内的风险中性,长期具有不确定性
3.2 Liquidity preference theory 投资期限长,要求流动性溢价
3.3 Segmented Market Theory 长期利率与短期利率相互独立
3.4 Preferred habitat theory 溢价足够高, 会吸引其余期限投资者
4.Yield curve factor models

4.1 Yield Curve Factor Models 形状变动的三因素
level movement 上下平移
影响最大
steepness movement 短期、长期发生变化
curvature movement 短期、中期、长期发生变化
影响最小
4.2 Managing Yield Curve Risks 风险衡量
Effective duration
key rate duration KRD: 刻画yield curve非平移时 特定时点△y对P的影响 (公式)
注意Dur.本身对P有负的影响
三因素对P的分解影响

4.3 Maturity Structure of Yield Curve
lognormal分布下,μ与σ的年月转换

Volatilities: σ短>σ长
σ长受uncertainty regarding the real economy and inflation. σ短受uncertainty regarding monetary policy
强调yield volatility σ(y)
5.Developing Interest Rate Views Using Macroeconomic Variables 宏观因素和term spread

5.1 Yield Curve (term sread) influencing factors
强调spread的大小(y的大小)
(1)通胀&货币政策
Inflation: 主要影响short- and intermediate-term bond yield variation
monetary policy: 主要影响long-term yield variation
Monetary policy影响bond risk premium (根源在于逆周期调节)
经济扩张期 benchmark rates↑ (P↓),bearish flattening

bear(r↑P↓) flattening (短期r↑↑ 长期r↑ 变平坦)
经济衰退期 benchmark rates↓ (P↑),bullish steepening

bull(P↑r↓) steepening (短期r↓↓ 长期r↓ 变平陡峭
QE量化宽松(large-sale asset purchases.)
Fed买入长期资产,货币供给↑ bond risk premium↓ (y↓),term spread↓(变平坦)
Fed卖出长期资产,货币供给↓ bond risk premium↑ (y↑),term spread↑(变陡峭)
(2)财政政策: 赤字↑ 资金需求D↑ (发债)债券供给S↑ 导致P_bond↓ y↑
(3)Maturity structure: 政府发行长期债(P↓y↑)使长期债券y↑(Segmented Market Theory)
5.2 Int Risk Management
5.3 Trading Strategy based on Yield Curve
LM2.The Arbitrage-Free Valuation Framework 通过利率二叉树给债券定价
1. Arbitrage Opportunity
无套利定价的原理

对Law Of One Price (LOOP)的违反
套利机会
Value additivity: t=0时获利
Dominance: t=T时获利
FI的无套利定价:stripping成单个zero bond,之后reconstituion
2. Introduction of arbitrage free valuation
fundamental principle of valuation
Estimate the future CFs
Determine the appropriate discount rate(s)
Calculate the PV
对于option embedded bond,CFs存在contingent的特性(interest rate dependent)
3. Interest rate trees and arbitrage-free valuation
3.1 Binomial interest rate tree二叉树

Binomial tree构建
利率涨跌有着equal probability
lognormal tree(因为基于lognorm R.W.)
给定波动水平σ是
给定benchmark yield curve
σ↑ spread out(二叉树更离散)
σ↓ 二叉树更紧凑,极端情况下 converge to the implied 1yr forward rate.
算例

3.2 Option-free bond valuation 用二叉树估值——算例

[方法一-backward induction]将t=2的CF(principal+coupon)根据t=1的各情景利率折现到t=1,得到PV(t=1) 将t=1的PV(t=1)按照t=1的利率折现到t=0,得到PV(t=0)=P0

t=1的利率管着t=2的现金流
[方法二-pathwise valuation]对于n期的二叉树,有2^n种paths, 根据各个路径的利率进行折现,最后求算术平均(因各path等可能)

3.3 Pathwise valuation
4. Monte Carlo Method
motivation: security's CFs are path dependent(此时二叉树法失灵), 例如MBS
Monte Carlo 步骤

说明1:当计算得到的 P_理论≠P_mkt 时,使用drift adjusted model (比如每个node加上10bp)使得 P_理论=P_mkt
说明2:结果具有统计意义,而非历史结果
说明3:使用过程中使利率具有mean reversion特征,并限定upper and lower bound
5. Term Structure Models

(第一类)Eqlbrm Model 均衡模型
用fundamental econ variables推导bond均衡价格
CIR model
组成部分
det. part(趋势项):k(θ - r)dt, ensures mean reversion
k是回归的速度,θ是均值
stoch. part(随机项):σ√(r_t) 随机项会变化随着r_t变化而变化,保证了r_t>0
Vasicek model
组成部分
det. part(趋势项):k(θ - r)dt, ensures mean reversion
k是回归的速度,θ是均值
stoch. part(随机项):σ 随机项是常数,r_t有可能为负数
(第二类)Arbtrage-free Model 无套利模型
使得定价P符合obs mkt price
Ho-Lee model
组成部分
det. part(趋势项):θt, 随时间变化,使得P理论match mkt price
k是回归的速度,θ是均值
stoch. part(随机项):σ 随机项是常数
子主题
子主题
子主题
子主题
子主题
子主题
KWF model
组成部分
det. part(趋势项):θt, 随时间变化,使得P理论match mkt price
k是回归的速度,θ是均值
stoch. part(随机项):σ 随机项是常数
独特之处:与Ho-Lee的不同,r_t~logNorm
logNorm distr.的右偏特性保证了r>0
右偏的fat right tail可以很好地刻画opt embedded bond(受分布的尾部影响大)
(第三类)Multiple factor Model 无套利模型
使得定价P符合obs mkt price
Guass+ model: 分析短中长利率
The long-term factor is mean reverting and reflects trends in macroeconomic variables.
The medium-term rate also reverts to the long-run rate.
The short-term rate : central bank controlling the short end of the rate curve.
LM3.Valuation and Analysis of Bonds with Embedded Options 含权债的定价
重难点
1. Fixed-income securities with embedded options
Callable bonds:涨不上去(P↑&P>X时行权,P=X),对holder不利 P↓y↑

Vcallable = Vpure - Vcall
Putable bonds:跌不下去(P↓&P<X时行权,P=X),对holder有利 P↑y↑

Vputable = Vpure + Vput
extendible bond: 赋予holder在到期时选择延期的权利(有利于holder)
市场利率高r↑(P↓)时,不延期(把资金投向回报更好的其他项目)
市场利率低r↓(P↑)时,延期,获取比其他项目高的回报
Complex option:
Convertible bond可转债
Vconver = Vpure + Vcall_on_stock
当股价St>X时,bondholder有权利以P=X(conversion price)的价格买入股票. 有利于holder,因此P↑y↓
Bonds with estate puts 附有遗产看跌期权的债券
holder deceased,继承人有权以par value把债券卖回去给issuer Vput:holder预期寿命越短、Vput越高
Sinking fund bond 附有偿债基金的债券
发行人建立Sinking fund,以防无力偿还时
2. Valuation of callable and putable bonds

2.1 Valuation of callable or putable bond
关键点:callable涨不上去,超过strike的价格一律取X; putable跌不下去,低于strike的价格一律取X
注意,在node时全额加上coupon
2.2 Valuation of an embedded option Vput和Vcall的计算(先求Vpure)
2.3 Effect of interest rate volatility
二叉树中的利率变动
以年为单位时t=1
波动率σ变化对V_含权bond的影响

σ↑:Vcall↑(对holder更不利) Vcallable↓
σ↑:Vput↑(对holder更有利) Vputable↑
2.4 OAS calculation and explanation
OAS: opt. adj. spread行权后的spread
OAS与Z-spread对比: motivation在于 行权前后折现率有变化
OAS作用: P理论≠Pmkt时, 在node加上OAS使两者相等
callable:OAS<Z-spread call对holder不利 应给予正RP,Vcall(%) = Z-spread -OAS

putable:OAS>Z-spread put对holder有利 应给予负RP, Vput(%) = OAS - Z-spread

option-free bond:OAS=Z-spread
从RiskPremium视角看, pure(%) + opt(%) = opt embedded bond(%) [OAS] + [RP] = [Z-spread]
OAS[计算]
试错法凑数(选择题时 先试中值)
应用结论
①含权债券的估值,应当使用OAS
②A和B两个债券各方面特征相同,如果A的OAS偏高,那么A被underpriced,如果A的OAS偏低,那么A被overpriced
σ对OAS的影响

σ↑ Vcall(%)↑ = Z-spread不变 -OAS变小

σ↑ Vput(%)↑ = OAS变大 - Z-spread不变
2.5 Interest rate risk 利率变化对债券价值的影响
int变动的影响
对含权债的影响
Callable bonds:涨不上去(P↑&P>X时行权,P=X),对holder不利 P↓y↑

Vcallable = Vpure - Vcall
Putable bonds:跌不下去(P↓&P<X时行权,P=X),对holder有利 P↑y↑

Vputable = Vpure + Vput
对含权债中期权的影响
call option
yield curve faltter:r↓P↑(ITM) → Vcall↑

hight int rate:r↑P↓(OTM) → Vcall↓
put option
r↓P↑(OTM) → Vput↓
yield curve falttens r↑P↓(ITM) → Vput↑

Effective Duration (ED) and Effective Convexity (EConv)
公式
利率风险对价格影响(风险) 由duration来测度

Effective Duration
ED计算步骤
基于benchmark curve (s)和Pmkt用二叉树求出OAS,由s+OAS得到新的二叉树(现在每个节点利率为i+OAS)
基于每个节点i+OAS-△y求出PV(-)
基于每个节点i+OAS+△y求出PV(+)
不同债券的Effective Duration对比
ED(zero-coupon) ≈ maturity of the bond
ED(fixed-rate bond) < maturity of the bond
ED(floater) ≈ time (years) to next reset

ED(callable) ≤ ED(pure)
ED(putable) ≤ ED(pure)
Effective Convexity
pure bond: 涨多跌少
Callable bonds: r偏低时有负凸性negative Conv., r偏高时有positive Conv.(左半段涨不上去)

Putable bonds: 显示出正positive Conv.(全程跌不下去)

One-Side Durations
motivation:含权债券ED具有不对称性 (int rate增减对P的影响大小不对称)
(callable为例)ITM行权时(r↓P↑),现金流提前完成,ED↓
OTM不行权时(r↑P↓),现金流与pure bond一致,ED不变
callable bond: up-D > down-D

putable bond: up-D < down-D

Key Rate Duration
Option-free bond (Not trading at par): 到期日KRD最大
Callable bonds (当r↓P↑时)
CR高时,发行人不行权(持有到期),到期日KRD最大
CR低时,发行人行权(获得大额CF),行权日KRD最大
Putable bonds (当r↑P↓时)
CR低时,发行人不行权(持有到期),到期日KRD最大
CR高时,发行人行权(获得大额CF),行权日KRD最大
Option-free bond (Trading at par): 只有到期日KRD>0,其他时间KRD=0

KDR<0 (原因 zero bond 或 very low coupon)
zero bond
10yr zero coupon bond本没有五年期,但为了研究PR5(可得s5)对s10的影响 进而对P10(10年期债券价格)的影响,可以通过par bond导出s1 s2 ... s5 ... s10,如紫色部分,可知s5↑导致s10↓导致P10↑
very low coupon
3. Valuation of a capped or floored floating-rate bond
 估值流程同上,只不过考虑cap和floor对利率的约束
The capped floater
Vcap-flt = Vpure - Vcap
cap对holder不利
The floor floater
Vfloor-flt = Vpure - Vfloor
floor对holder有利
算例

capped

floored

4. Convertible bonds

4.1 Valuation and Analysis: Convertible Bond
基本概念
本质:embedded call option (call on stock 对holder有利)

Conversion ratio: 25:1 一个bond可以有权买25股
Conversion price (strike X) = bond issue price ÷ conver ratio (100元/bond ÷ 25股/bond = 4元/股)
Market conversion price = bond mkt price ÷ conver ratio
与Conver price (X) 构成套利关系
Conversion value = stock mkt price * conver ratio 立刻转股值多少钱
Straight value 普通债券(不含权)的市场价值
The minimum value
The mkt conver premium per share 在二级市场买可转债兑成股票的价格 减去 直接买股票的价格
买可转债换股 比 直接股票 更贵
Market conversion premium ratio 溢价÷直接买股票的价格(%概念)
Premium over straight value (%概念)
straight value衡量了downside risk (S不涨,最低可以获得straight value)
这种premium over straight越高,conver bond越贵(吸引力越差)
conver price and conver ratio需要适时调整
Corporate actions: stock splits, bonus share issuances, and rights or warrants issuances 导致股价S↓<<X,为保护holder利益,需要X↓
dividend payments: 发放股利 股价S↓<<X,为保护holder利益,需要X↓
threshold dividend条款:S↓股价下跌幅度超过阈值时X↓
Change-of-control events: A公司兼并了B,A股价↓,A公司的可转债holder利益受损,需要补偿
put option条款: 允许holder在特定时期将债券卖回给issuer(redemption),用现金叫Hard put,用cash, common stock, subordinated notes, or a combination of the three叫soft put
adjusted conversion price条款:让strike price (X) ↓
给予holder权利与补偿的同时,issuer利益受损,需要补偿
call option: 赋予issuer权利回购conver bond (i.e. callable convertible bond)
callable convertible bond对issuer有利,holder要求补偿
protection period
called but at a premium
forced conversion: 满足转股条件时holder优先转股,而不能被issuer call回
4.2 Valuation of Convertible Bond 含权可转债的价值
普通convertible bond
Callable convertible bond
Callable and putable convertible bond
4.3 Effects of Embedded Option 波动率σ对可转债里option价值的影响
当σ↑ Vconver_bond ↑ = Vpure 不变 + Vcall_on_stock 必增加
σ↑ V转股权↑; σ↓ V转股权↓
4.4 Analysis of Convertible Bond
比起直接买stock, 可转债的盈亏波动更小

Ps↑时,由于存在conver premium (买可转债换股 比 直接股票 更贵), 可转债换股的收益<直接买股票
Ps↓时,由于可转债有min value(不转股 straigh value)兜底, 可转债换股的亏损<直接买股票
可转债风险-收益risk-return的特征分析
S<X时 不行权,可转债为Fixed-income equivalent,呈现债券的risk-return特征
busted convertible 转债的权利没有使用
S>X时 行权,可转债为Common stock equity,呈现股票的risk-return特征
LM4.Credit Analysis Models 存在违约风险时的债券定价

1. Modeling credit risk and the credit valuation adjustment
credit risk
Expected exposure
Probability of default (POD)
Recovery rate (率%)
Loss given default (LGD) (量$)
Credit valuation adjustment (CVA) = ∑PV(Expected Loss)

Fair value of corp. bond = VND – CVA
VND: the value for the corporate bond assuming no default
prob of default (POD)
actual POD
RiskNeutral POD:通常是高估的
 -在风险中性RN定价下,分子考虑了风险(加权求期望值)、分母不考虑风险(以无风险利率折现)。无套利定价使得①=②从而 YTM-Rf=POD*(1-RecovRate) 解出POD,即R.N.POD -等式左边YTM-Rf为spread,补偿了credit risk、liquidity risk等等,但等式右边POD*(1-RR)即为Expected Loss却只衡量了credit risk,因此左边>右边,即求出的R.N.POD高于真实的POD
2. Valuing risky bonds in an arbitrage-free framework 风险债券的无套利定价

第一类估值:r相同
(1) Credit Analysis of Zero-Coupon Corporate Bond
基本思路:先求FairValue=VND-CVA (coupon=0),再求rate of return(不违约为YTM,违约为IRR),据此求出credit spread=YTMc-y_b
算例

step1. 求解FairValue

VND的计算:正常求现值
零息债券只看到期日CF
CVA的计算:CVA=∑PV(Expected Loss)其中ELt=PODt*LGDt=PODt*Exposure_t*(1-RR)

PODt的计算:上一期先存活 才能有下一期

Exposure_t的计算:P_t+coupon_t(为0)
此例无coupon,故根据时点求Pt即可
DF折现因子:1/(1+r)^t

step2. 求rate of return
无违约 YTM

违约 IRR

∑Recovery$_t/(1+IRR)^t=FairValue(价格)
  注:此例中coupon=0,若有coupon要在0所在位置补上
step3. credit spread
无违约 YTM

违约 IRR
(2) Analysis of Coupon-Paying Corporate Bond
基本思路:先求FairValue=VND-CVA (coupon≠0),再求rate of return(不违约为YTM,违约为IRR),据此求出credit spread=YTMc-y_b
算例

step1. 求解FairValue

VND的计算:正常求现值
CVA的计算:CVA=∑PV(Expected Loss)其中ELt=PODt*LGDt=PODt*Exposure_t*(1-RR)
PODt的计算:上一期先存活 才能有下一期
Exposure_t的计算:P_t+coupon_t
DF折现因子:1/(1+r)^t

step2. 求rate of return
无违约 YTM

违约 IRR
∑Recovery$_t/(1+IRR)^t=FairValue(价格)
 注:此例中有coupon
第二类估值:二叉树 (Arbitrage-Free Valuation)
(1) fixed-rate risky bond
基本思路:先求FairValue=VND-CVA (Pt需要求加权平均),再求rate of return(不违约为YTM,违约为IRR),据此求出credit spread=YTMc-y_b
算例
  题干与前一类估值相同,但存在σ=10%,故使用二叉树
step1. 求解FairValue
 
VND的计算:正常求现值

CVA的计算:CVA=∑PV(Expected Loss)其中ELt=PODt*LGDt=PODt*Exposure_t*(1-RR)

PODt的计算:上一期先存活 才能有下一期

Exposure_t的计算:avg(P_t)+coupon_t
 
DF折现因子:1/(1+r)^t
 同上一算例
step2. 求rate of return
无违约 YTM

step3. credit spread
无违约 YTM

重要结论:σ↑时,CVA↓,F.V.↑
原因:在r~logN对数正态分布下,利率变化具有不对称性
 从同一个middle rate出发,r↑↑的幅度>>r↓(上涨的多,下跌的少),导致r上涨时P↓↓,r下跌时P↑,因此r↑时对应的exposure、LGD、PV(ELt)、CVA都比r↓时的值要小,即: CVA(r↑)↓↓+CVA(r↓)↑<0 即CVA↓ 导致F.V.变高(F.V.↑=VND-CVA↓) 
(2) floating-rate risky bond
算量太大,如考到则放弃
基本思路:先求FairValue=VND-CVA (Pt和Ct都需要求加权平均),再求rate of return(不违约为YTM,违约为IRR),据此求出credit spread=YTMc-y_b
算例
 
step1. 求解FairValue

VND的计算:Backward induction 【二叉树从尾到头折现】

CVA的计算:CVA=∑PV(Expected Loss)其中ELt=PODt*LGDt=PODt*Exposure_t*(1-RR)
PODt的计算:上一期先存活 才能有下一期

Exposure_t的计算:avg(P_t)+avg(coupon_t)

P和C拥有不同的weights,因为 C的情况数比P情况数少一种
DF折现因子:1/(1+r)^t
 同上一算例
step2. 求rate of return
无违约 YTM
step3. credit spread
无违约 YTM
Floater的重要概念:Discount Margin
Floater估值时用到的折现率r=Libor+DM 相当于针对credit risk的premium

在这个算例中,F.V.=99.9047<Par=100(折价发行 CR<r)→ QM=0.5%<DM

[求解DM具体值] 试错法 在二叉树节点加上DM 使得P理论=Pmkt(市场价取F.V.值)

“在节点上+DM”反映了从不违约的定价(VND) 向考虑违约(CVA)进行调整
3. Credit scores and credit ratings 信用评估的定性方法
Credit scores 评估小型企业和个人
FISCO 费埃哲公司

子主题
Credit ratings 评估大企业和政府发行的债券和ABS
ordinal measures 不存在倍数关系
Notching:企业是A级,发行的债券是B
rating outlook (positive, stable, or negative)
Credit Transition Matrix
如何读表

[计算收益率变化]
先求各情景下的价格变动百分比: △%P = -MD*△Spread = -MD*(结果Spread-初始Spread)

再求离散期望值: E[△%P]=Prob(AAA)* △%P(AAA)+Prob(AA)* △%P(AA)+...

为何Rating的期望值总是倾向于下调?
①Prob of chg is not symmetric (Matrix显示评级下调的概率更大)
②下调时credit spread的增长更大 (spread↑则y↑P↓)
4. Structural model and reduced-form credit models 信用评估的定量方法(结论)
Structural model

从公司B/S出发,资产A<负债Liab=K时违约
违约事件是endogenous内生的(基于B/S):给出原因是公司资不抵债
Reduced-form model
从资本来源出发,分为Debt和Equity,其中Equity本质是Call on asset Debt=A(二级市场交易定价)-Equity(BSM期权定价) 最终求出Debt
 
违约事件是exogenous外生的(macroeco variable):没有给出经济学原因,认为违约只具有统计意义(surprise),但说明了何时违约
 (基于Poisson stochastic process,关键变量是default intensity,反映了每个time increment的POD)
通过类比option,可以得出V_risky-debt = V_risk-free-debt - V_put

对比 Structual VS Reduced-form

Structural model
假设
①公司资产可在二级市场交易(获取价格)assets trade in frictionless markets
②debt liabilities are zero-coupon bonds
③POD is endogenous (internal)
优点
①指出原因(公司资不抵债)the nature of credit risk.
②类比期权定价来给债务定价option analogy
缺点
①假设不切实际
②limitations in available data(非公开/无交易)
③不反应外部经济环境the business cycle
Reduced-form model
假设
POD is exogenous (external)
优点
①数据可得性好inputs are observable variables,.
②回归分析外部经济环境the business cycle
缺点
①违约有统计意义,但没有经济意义(没有给出原因)
②因为是统计模型,违约可以在任何时候发生default is rarely a surprise
5. Credit spread analysis 信用利差分析
Interpreting changes in credit spreads
POD↑使得Credit Spread↑,RR↑使得Credit Spread↓(直观理解)

POD的影响比RR的影响更大: 因为使用POD时需要叠加存活率,使用RR直接使用

The term structure of credit spreads 信用利差的期限结构
credit curve:credit spread为纵轴,t为横轴的曲线
Key drivers
Credit quality:信用等级高的,credit curve平坦;信用等级低的,credit curve陡峭(也可能向下倾斜——)
Financial conditions:经济好,risk↓,credit spread↓
Market supply and demand:
D>S,P↑y↓,credit spread↓
frequently traded securities对credit curve影响大
infreq traded bond的bid-ask spread越大,说明流动性差
Company-value model:分析影响POD的因素,POD↑ risk↑ credit spread↑
如果P_bond接近recovery value,则说明即将default (原因:如果市场价为公允价值,则反映了扣除CVA后的价值)
分析term structure of credit spreads 的重要考虑因素
(1)选择适当的benchmark rate
gov bond需要满足:流动性好、on-the-run、期限M匹配的
如果期限M不匹配:使用interpolate插补,或者用benchmark SR(swap rate)
(2)选用适当的debt来代表公司
选用senior unsecured debt来分析
不适用含权债embedded option、优先求偿债lien、unique provision的债券
6. Credit analysis for securitized debt 针对[证券化]债券的信用分析
ABS信用分析的考虑因素
(1)characteristics of the asset portfolio 资产池的特征
投资组合的颗粒度granularity of the portfolio:资产池的资产数量多(颗粒度大),则从组合角度进行分析;资产池的资产数量少,则对每一个资产(eg loan)进行分析
底层资产的相似度homogeneity:资产池各资产相似度高,则从组合角度分析;相似度低,则逐一分析
(2)credit exposure of the origination and servicing of ABS ABS相关方能力带来的影响(能力的强弱是一种信用敞口)
(3)the structure of a collateralized or secured debt transaction 证券化债务的结构(eg 破产隔离保护下,F.V.↑,TYM↓,spread↓)
bankruptcy remoteness破产隔离
credit enhancements增信
担保债权covered bond
A用Bond作为抵押,向B借贷,Bond所有权在A、CF流向A,A除了直接还款、还可以用bond现金流还款——“dual recourse”
LM5.Credit Default Swaps (CDS) 信用违约互换

1. CDS fundamentals
CDS fundamentals
原理:CDS本质是一种针对债务违约的保险产品,其互换特征体现为protection buyer每期支付fixed coupon(由CDS spread确定)给protection seller

相关方:bond issuer 叫做ref. entity,holder叫做protection buyer,保险公司叫做protection seller
标准化CDS spread(即保费)为投资级1% 投机级5% (交易时如果卖价为其他数字,则多退少补)

场外市场欲订立3%spread保费(应付)的合约,但已有的标准化产品投资级为1%, 于是签订标准化1%合约(实付),buyer额外在期初额外补上按2%spread计算的保费
场外市场交易、有行业协会:International Swaps and Derivatives Association (ISDA) 交易双方遵守 ISDA Master Agreement
notional amount:赔付额度
易错易混:protection buyer is short and protection seller is long
标的资产是bond,P↓时,protection buyer (bond holder) 获益(获得赔付金)=short side
Structure and features of credit default swaps: 实物交割、现金交割

Common types of credit events
Bankruptcy
Failure to pay
Restructuring: 在美国,重组不算credit event
2. Types of credit default swaps
(1) single-name CDS

①ref. entity的债券中同等级pari passu&更高等级债券违约,则prot buyer可以申请赔付
②赔付的时cheapest-to-deliver(CTD)原则:ref. entity的相同级别债券中,按照cheapest(loss最大)的额度进行赔付
(2) Index CDS (CDX): 对一堆标的资产的CDS

易错易混:CDX buyer is long credit exposure and CDX seller is short credit exposure (注意是信用敞口的多空——发生credit event→buyer获得赔付(long side) seller支付赔付(short side))
①总保额平均分配给n个标的资产 (100million cover 100个bond,每个最大只能保1m)
注意,如果其中一个bond违约(最大赔付额1m),赔了0.7m, 那么整个保单剩余赔付额notional principle=100-1=99m

②标的资产中相关性ρ↑,集体违约risk↑,spread↑(保险机构的成本高,要求高保费)
③实战中,保险公司会定期update这个index中各个资产的情况(on-the-run series),投资者也要根据标的资产组合的变化签订新的CDX(这个动作叫roll)

(3) tranche CDS: 与ASB tranche类似,按信用等级分级偿付
3. Pricing and valuation of CDS
Basic Pricing Concepts
本质:支付额为对credit risk的补偿,即为credit spread
影响因素:POD、LGD越高,收费越高
CDS Pricing定价 Conventions
“两条腿走路”
Premium leg 每期保费
Protection leg 赔付额
Pricing定价原理
事后角度:价格(一次性付清保费)=PV(赔付额)-PV(每期保费)

事前角度:价格(一次性付清保费) = (Credit spread - Fixed coupon) * Duration

Quoted Price 报价方式 (站在buyer角度)

upfront pmt是buyer支付的额度,如果是收到2%退费, 则支付了-2,那么Price of CDS = 100 - (-2) = 102
Valuation估值 after Inception of CDS
利润来源:spread↑,账面浮盈profit↑(以更低的保费获取相同的保额)
Termination of CDS: 通过进入一个反向合约,将账面浮盈实现
% change in CDS price
4. Applications of CDS
(1) Naked CDS(不持有bond): 公司信用↓(认为要违约),买入CDS;公司信用↑(认为不违约),卖出CDS
(2) long/short trade: 两个同等级债券的credit sprea不一致,买低卖高
(3) Curve trade: 利用credit spread进行CDS多空交易(画图分析)

(4) basis trade: 单个bond和CDS各自的credit spread存在套利空间
算例: bond CS > CDS CS

算例: bond CS < CDS CS

口诀:哪个收益高就用哪个,用的时候同方向 如果是买(bond)那么也要买CDS 如果是卖(CDS)那么也要卖bond
(5) arbitrage trade: 一组bonds和CDX各自的credit spread存在套利空间

买低卖高
(6) 其他策略: 任何发行都伴随着信用风险,根据情景判断
例子: 杠杆收购LBO中,A公司发债 risk↑(投资者应该买入基于A公司的CDS)
 
Quantitative Methods
LM1 Basics of Multiple Regression and Underlying Assumptions

1. Linear regression
1.1 Simple linear regression

图示

1.2 Multiple Linear Regression

应用1:量化投资

应用2:债务分析

应用3:宏观分析

2. Multiple Linear Reg. Assumptions
Linearity: relationship between Y and X is linear.
Homoskedasticity: Var(ε)=σ
Independence of errors: Cov(εi,εj)=0
Normality: ε~N
ε~iid N(0,σ)
Independence of X(与ε):
5a. Cov(ε,X)=0 Independent variables are not random.
5b. X'X不满秩(非奇异矩阵) X不具有multiple linearlarity.
多元回归特有
3. Detection of violations: Diagnostic plots
对假设的违背进行建设
• Scatter plots

检验线性性
plot Y X
• Residual plots 检验homoske

error term在上下界内波动→ Var(ε)=σ √
error term从左到右上下交替出现,没有持续>0或<0→ Cov(εi,εj)=0 √
plot hat_Y ε
error term Q-Q图连续且为直线,横轴为理论norm分布,纵轴为实际分布 (符合随着分位数增长而等比例变化)→ ε~N √

Q-Q plot原理: 理论上的分位数分布与实际的分位数分布应该同比例变化

5.a 随着X变化,斜率基本为0→ Cov(ε,X)=0 √

plot X ε
5.b 相关系数ρ(Xi,Xj)≠±1→ X'X不满秩 √

plot Xi Xj
4. Regression process

identifying the Y and X (连续、不连续);

selecting the appropriate regression model;
testing if the assumptions behind linear regression are satisfied;
examining goodness of fit;
making needed adjustments
LM2 Evaluating Regression Model Fit and Interpreting Model Results

1. Measures of goodness of fit
• ANOVA table

例子

TSS=ESS+RSS

方差:公式遵循 "中心化平方和/自由度df"

MSR=RSS/k 回归的均方(√解释)
 MSR=RSS/k
均方是方差的概念,不是标准差
MSE=ESS/(n-k-1) 误差的均方(×解释)
MSE=ESS/(n-k-1)
均方是方差的概念,不是标准差
Var(Y)=TSS/(n-1) 因变量的方差(总的方差)
概念辨析:ss/df均方(方差)、(ss/df)^(½)标准差(波动性)、variation偏差
• Coefficient of Determination 决定系数 R²
R²能说明什么:total variation中能被解释的那部分variation占比是多少
R²不能说明什么
cannot tell the coefficients are statistically significant
cannot tell biases in the estimates (coefficients and predictions)
cannot tell whether the model fit is good
好模型可能R方很低:CAPM
坏模型可能R方很高:overfitting and biases
• Adjusted R²: adjusted for df
公式
特点
第③点:根据新加变量的|t|><1判断对Adj.R²的影响
 注意,Adj.R²↑不能作为新变量significant的依据(因为只知道|t|>1,不知道是否显著,所以需要另算)
R²和AdjR²是衡量,不能作为Model fit的指标
而需要F-test,AIC,BIC
• AIC & BIC
motivation: 好的模型是在准确度和简洁度parsimonious之间权衡

公式:AIC,BIC数值越低、fitting better
前半部分关注准确度(ESS↓ 准确↑) 后半部分关注简洁度(#k↑ 复杂↑ overfit)
 AIC推荐model4(关注准确性),BIC推荐model1(惩罚过多的自变量)
两者差异:n>8时(绝大多数分析)时, BIC有greater penalty(ln(n)>2)、更强调简洁

AIC用于prediction 更关注准确度,牺牲一定的简洁度
BIC用于goodness of fit 更关注简洁度,牺牲一定的准确度
AIC BIC的数值无绝对意义,必须在比较中得出结论
2. Significance test for regression coefficient
• Joint hypothesis tests
2.1 Tests of a single coefficient: 单个解释变量显著


2.3 General linear F-test: 所有解释变量联合检验、模型整体显著


2.2 Joint F-test: 部分解释变量联合检验


算例

General-F(MSR/MSE)本质上是Joint-F(restr-unrestr) 的一种特例:相当于把k个自变量都施加restr

• General linear F-test
3. Predicted value of Y
Two sources of uncertainty
Model error: 误差项 ε contains uncertainty
Sampling error: Uncertainty in 自变量X forecasts
Confidence interval [考察辨析] 宽度为2t*Sf(标准误std error)
计算二级不考
model error变大,Sf↑,C.I.变宽 (误差越大,估计的准确度月底)
LM3 Model Misspecification
1. Model Specification: • Misspecified Functional Form
1.1 Model Specification 应该具有特质
economic reasoning
parsimonious
perform well out of sample 外部有效性
functional form should be appropriate 二次型、指数型ect
satisfy regression assumptions
1.2 Misspecified Functional Form 模型设定错误的后果、检验、修正
Omitted Variable

后果
biased estimates
如果cov(X1,X2)=0(遗漏X2), 截距项b0 biased但b1 estimated correctly
如果cov(X1,X2)不等于0(遗漏X2), b0 and b1 biased
serial corr.: 因为ε'=b2X2+ε, X2自身具有相关性(t期与t-1期)使得ε序列相关
heteroskedasticity: ε'=b2X2+ε使得Cov(X1,ε')≠0→X1变化则Var(ε')≠0
inconsistency: 无法依靠n↑来实现min S.E.
检验:plot residual and 遗漏的X2 (如相关→确实omitted X2)

修正:在模型中加入omitted variables
Inappropriate form of variables 本来是二次关系却用一次函数刻画

后果
heteroskedasticity
检验
plot Y X
修正
根据真实的关系构建回归方程,eg 对数回归
Inappropriate variable scaling 变量的测度不统一
后果
heteroskedasticity: 单位不同而成为outlier

Multicollinearity: 变量之间高度相关

检验
检查变量之间的单位统一(万元、万元、亿元)
修正
common-size analysis
Inappropriate data pooling 结构性变迁导致前后时间的数据不能在一起回归
后果
heteroskedasticity
serial corr.
检验
structural breaks test
修正
sub-sample regression
2. Multiple Regression Assumption Violations

2.1 Heteroskedasticity Var(ε)≠σ
问题来源: Model Misspecification
omitted variables,
incorrect functional form,
incorrect data transformations,
extreme values of independent variables
异方差的两种类型
Uncond.l heteroske. Var(ε)

Cond.l heteroske. Var(ε|X) 对数据分析威胁严重

后果(方差类)
无影响
有影响

S.E.↓ t统计量↑ 更容易rej H0 [容易犯第一类错误]
S.E.↑ t统计量↓ 更容易accept H0 [容易犯第二类错误]

S.E.↓ t统计量↑ 更容易rej H0 [容易犯第一类错误]
S.E.↑ t统计量↓ 更容易accept H0 [容易犯第二类错误]
子主题
检验
Residual scatter plots

Breusch-Pagan χ² test (BP-test)

H0: 没有异方差 (!没有)
统计量χ²=n(R_resid)²>C.V.(或者p-value<α)时rej H0,认为存在heteroske
修正
White-corrected S.E.
获取robust S.E.
金融市场中,hetero的存在表明 有效市场失灵,可以通过σ²赚取超额收益

2.2 Serial Correlation Cov(εi,εj)≠0
serial corr的两种类型
Positive serial correlation: 惯性

first order serial corr: Cov(ε_t,ε_t-1)≠0
second order serial corr: Cov(ε_t,ε_t-2)≠0
Negative serial correlation: negative effect

后果(方差类)
incorrect S.E.

进而影响t-/F-stat检验的不可靠
inconsistency:当自变量含有Y的滞后项Y_t-1, Y_t-2,... 此时scatter plot上点较为集中 MSE↓ S.E.↓系数更容易显著

如果自变量不含Y滞后项则仍可以consistency
检验
Durbin-Watson (DW) test (只考虑first oder序列相关)


Breusch–Godfrey (BG) test 可以对p阶自相关检验

修正:针对S.E.进行修正
serial correlation consistent S.E. 仅修正serial corr.
serial correlation and heteroskedasticity adjusted S.E. (HAC) 同时修正serial corr和heteroske
H:异方差 A:自相关 C:一致性
Newey–West S.E. 协方差矩阵对角线权重不唯一的一种HAC
robust S.E.
2.3 Multicollinearity X'X满秩(奇异矩阵)
后果(均值类)
无影响
有影响
incorrect estimates of b0, b1

inflated S.E. 同时t-/F-检验失效

S.E.↑ t统计量↓ 更容易accept H0 [容易犯第二类错误]
检验
Classic method

单个变量t检验不显著,但整体方程F检验显著,R²偏大
Check Pairwise Correlations 两两变量相关性(皮尔逊相关)
局限在于:只能描述两个变量之间的相关,无法捕捉3个及以上联合相关
Variance inflation factor (VIF)
VIF越高则多重共线性越严重(VIF>10表明多重共线性十分严重)
修正
Excluding one or more regressor (删去共线变量)
Using a different proxy (把共线变量替换为其他代理变量)
Increasing the sample size (n↑)
LM4 Extensions of Multiple Regression

1. Influence Analysis
Influential Observation 具有强影响的观测值
high-leverage point:x太大

outlier:y太大

当远离回归线时才会产生影响(slope coeff、goodnes of fit)
Detecting Influential Data Points检验强影响点
Single linear regression: Scatterplot
Multiple linear regression: Quantitative way (注释有总结)

Leverage:用于high leverage points, hii=standardized[Xi-X_bar] > 3(k+1)/n
其中(k+1)/n是各个hi的均值
Studentized residual


考试考察判定法则 注意df=n-k-2
Cook's distance (Cook's D) 同时检验了leverage(高x)和residual(高y)
2. Dummy Variables
dummy variable, or indicator variable
共有n中类别,只需设定n-1个dummy即可

coeff.的含义:
截距项b0是参照组的y值:eg第四季度销售额 (一季度m,二季度n,三季度k,四季度不设dummy)

系数b1是基于参照组值得出的y值:第一季度销售额相对第四季度销售额

检验:无法rej [H0: b1=0] 说明存在季节性(一季度和四季度销量无差别)

Different Types of Dummy
Intercept Dummy

例子

Slope Dummy

例子

Dummies in Both Slope and Intercept

例子

3. Logistic Regression
Qualitative 定性自变量
logistic transformation:将变量进行log-odds处理

逆运算:P=1/(1+exp(-Y))
Logistic Regression
Logistic regression (logit) 用于二值变量回归
参数估计方法:由于distr不满足OLS、存在fat tail 转而用maximum likelihood estimation (MLE)
coeff解释
Slope: 自变量x变动1单位时,log-odds即ln[p/(1-p)]变动多少单位, 再通过log-odds逆变换求出对p的影响
Intercept: 所有自变量x=0时,log-odds即ln[p/(1-p)]=1
Hypothesis test
Test for single coeff:与OLS回归的检验一致
Test For model fitness: Likelihood ratio (LR) test 似然比检验

与F检验类似,但用logL而非ESS

H0: restr model (b2=b3=0) H1: unrestr model
统计量 LR = -2 * [ logL(restr) - logL(unrestr) ]
logL的性质

LR>C.V.时,rej H0 认为unrestr model更好(b2 b3确实有用)
算例

LM5 Time-Series Analysis

1. Trend models
Linear trend model

适用于数据点显示 uncorrelated errors
Log-linear trend model

适用于按照指数增长,增长率g=exp(b1)-1

trend model存在局限性:无法解决serial correlation
2. Autoregressive models (AR)
AR(p) model:
2.1 Chain rule of forecasting 递推法预测
2.2 Assumption
假设一:协方差平稳 Covariance-stationary series
具有mean-reversion 以AR(1)均值为Xt=b0/(1-b1)

协方差平稳内容
期望值constant&finite: E[Yt]=Y_bar
方差constant&finite: Var(Yt)=σ
协方差constant&finite: Cov(Y_t,Y_t-k)=constant
违背后果:随机游走R.W.(单位根Unit Root)

Random Walk
Random Walk with a drift
检验:DF-test
t-test:直接对b1系数检验,但由于存在了serial corr,b1本身的dirstr不服从t-distr,不可靠
Dickey-Fuller test (DF test)
H0: g=0(存在单位根且non-stationary);H1:g<0(不存在单位根且stationary)
*DF原假设H0是不可欲的(H0: 有随机游走/单位根)(DF-EG也是H0不可欲), 其他检验都是可欲的(比如BP-test H0:没有异方差 BG-test H0: no ser. corr.)
注意H1必须是g<0(保证b1<1否则发散)

DF_t>CV(修正后的新表) rej H0 认为不存在单位根且stationary(可以直接用AR)
修正:一阶差分first-difference
如检验后仍存在unit root则再次差分
假设二:无序列相关 No autocorrelation
无序列相关 cov(εi,εj)=0
违背后果:随机游走R.W.(单位根Unit Root)
S.E.错误
DW统计量(检验序列相关)错误
检验:对residual autocorr 进行t-test

【考试】样本T=n-p
S.E.=1/√T
修正:季节性调整seasonality
比如p=1,2,3都不存在序列相关,但p=4存在,则需要在模型中加上X_t-4

算例

假设三:不含条件异方差 No conditional heteroskedasticity
自回归的error term存在条件异方差: Var(ε|X)≠σ

用ARCH(1)来刻画 AutoRegCondHeteroske
volatility clustering
违背后果:同heteroske的后果
检验:简化版ARCH(1)

多元回归时用BP-test
ARCH(1)[σ²用ε²替代]
H0: a1=0(不存在CondHet);H1:a1=0(存在CondHet)
t-stat>CV rej H0 认为有条件异方差ARCH(1) error
修正:广义最小二乘GLS
通过上一期ε²_t-1找出本期σ²_t的形式,修正协方差矩阵到满足OLS的条件,即可开展正常OLS手续

2.3 评价模型表现 RMSE
预测效果的表现
In-sample forecast errors
Out-of-sample forecast errors
定量评估方法
衡量指标:样本内误差的S.E. (简称S.E.of Error)越小越好
只用于in-sample
均方根差Root mean squared error (RMSE)
属于S.E.类的概念
定性评估方法
关注earlier sample VS later sample period(例如汇率制度的regime变化)
relatively shorter and longer sample periods(较短时段和较长时段)
2.5 其他自回归模型
MA model
ARMA model
p是自回归项数,q是移动平均项
3. Regression with more than one time series
两组时间序列进行OLS回归Yt=b0+b1Xt+ε

DF-test 检验单位根情况

Yt Xt 都没有单位根:可以OLS
Yt Xt 两者其一有单位根:不能OLS
Yt Xt 两者都有单位根:如果存在coint关系,则可以ECM处理后回归
DF-EG test: 检验coint协整关系
H0: no coint H1: coint
同DF-test,H0不可欲
DF-EG> C.V.(调整后的新表) if rej H0, we can use multiple regression (ECM)
LM6 Machine Learning

0. Overview of machine learning
0.1 Defining machine learning

ML与reg的异同

Terminologies
因变量Y:target variable目标变量
自变量X:features特征变量
Hyperparameter: 程序员人为设定的参数(eg features数量(k个)的选取)

0.2 Types of ML 概述
Supervised learning (SL) labelled data
告知分类 机器总结规律
类型
Regression model:用于连续Y

Classification model: 用于分类Y(定性的)
Binary classification(1 or 0)
Multicategory classification(债券评级)
Unsupervised learning(SL)
机器先总结规律再分类
discover structure within the data总结规律
类型
Dimension reduction降维:

①reducing the # of X (防止出现overfitting)
②尽可能保留方差即信息的variation(防止出现underfitting)
Clustering聚类:
进行分类sorting observations into groups (clusters)
 优先股在风险-收益性质上更接近bond,因此要归为FI
Deep learning.
模拟人思维的方式
Deep learning:处理复杂任务
reinforcement learning:从错误中修正模型
0.3 Challenges of ML
Underfitted: not capture the relationships in the data.
Overfitting过拟合:模型对于degree of specificity过多地incorporate noise,导致无法很好地拟合out-sample-data
SL监督式学习更容易出现overfitting
measure how well a model generalizes 测度模型泛化的程度
Training sample(训练集) 样本内数据
Validation sample(调适集) 样本外数据
Test sample(测试集) 样本外数据
关于fitting的三类错误
Bias error:(主要因素)模型太简单 用训练集都underfitting
Variance error:模型太复杂 用样本外数据就失效 overfitting
权衡取舍:Fitting curve
 模型复杂度:the number of features, terms, or branches以及线性非线性
Base error:模型自有的 通过权衡fitting后仍无法消除误差 residual errors due to random noise
overfitting的修正
complexity reduction降低复杂度:施加penalty(变量数k↑ penalty↑)
cross validation交叉验证:
滚雪球式的rolling验证

k-fold cross validation分为k组互相验证

三种ML:掌握机制和适用条件 (注释有归纳图)

1. 监督式学习 Supervised Learning Algorithms
1.1 Penalized regression惩罚回归: OLS回归中加上penalty term

SL+regression
线性模型:
超参数λ↑ 惩罚↑
λ=0时,LASSO=OLS
非线性模型:Regularization正则化模型
1.2 Support vector machine (SVM)支持向量机: 用hyperplane将向量空间线性二分

SL+classification
内容:通过对数据进行切割(最大化最小距离maximize min distance,从而取到maximum margin),以实现max Prob{作出正确预测}

特征
本质:对数据进行linearly分类
目标函数:用直线分割max margin, 在[wider margin]和[classification error]之间权衡取舍
曲线分割为soft margin
适用条件:small- to medium-size 但同时 complex high-dimensional data sets.
high-dim数据容易用平面(直)分割 low-dim则需要用soft margin

缺点
只能分为2组
落到margin内部的new data无法分类
1.3 K-nearest neighbor (KNN 即 K近邻算法) 对最相似的数据(nearest)进行分类
SL+classification
内容:通过设置hyperparameter 划定近邻范围,识别数据点属于那种类型
优点
数据可划分分为多组
可以对new data进行分类
缺点
超参数k对结果有重大影响:过小(high error rate)、过大(dilute the result)、偶数(识别失败)
很难定义similarity:空间距离还是数据特点相似?
应用
bankruptcy prediction
credit rating assignment
1.4 Classification and regression tree (CART) algorithms 决策树 适用linear也可non-linear
SL+regression+classification
算法类型与结果
离散Y:使用Classification trees
分类结果:少数服从多数majority points
图例中3个NoDiv,2个Div
连续Y:使用Regression trees
回归结果:所有的terminal nodes取均值
决策树
术语
起点:initial root node
终点:terminal nodes(圆圈)
决策点:decision nodes
features:自变量X
阈值cut value:在起点、决策点 对X设定阈值进行classification
特征
在initial root node实现widest separation(min classification error)
经过每个decision,对feature space的划分越来越细致(减少within-group error)
一直划分到分叉已经无法减少classification error时,决策树停止
子主题
优点:visual explanation,不像其他算法运算过程是blackbox
缺点:本质上是SL 容易overfitting
修正
regularization criteria设置正则条件(k↓)
pruned 剪枝
1.5 Ensemble learning集成学习 and Random forest随机森林 集成学习(综合了前4中ML算法)
各种模型的noise相互抵消 会有lower average error rate
SL+regression+classification
类型
aggregation of heterogeneous learners 异质化模型:相同数据,不同算法
aggregation of homogenous learners 同质化模型:不同数据,相同算法
不同数据的获取:boostrapping(bagging)重抽样

Random Forest随机森林 (是一种同质化模型)
内容:基于重抽样获得的sample sets,分别进行CART(决策树),多个决策树=>森林
步骤
①重抽样获取data bagged,产生多个决策树
②每个决策树随机使用subset of features(X)
降低k↓ 部分克服overfitting
③wisdom of crowd:对每棵树的结果少数服从多数
优点
protect against overfitting
reduces the ratio of noise to signal (ratio=noise/signal)↓
缺点:仍存在blackboxes,哪棵树对结果的影响大、影响小?
2. 非监督式学习 Unsupervised Learning Algorithms
2.1 Dimension reduction降维:PCA主成分分析
降维k↓ + 尽可能保留variation
UL+regression
PCA步骤

根据相似性构建uncorrelated composite variables
所得特征向量Eigenvectors (composite variable) = Ci 互相线性无关
获取每个Ci的特征值Eigenvalue (类似R²):每个Ci的解释力度
保留能解释85%—95% of the total variance的主成分C1,C2,C3...并reg against Y
优点:降维k↓ 模型复杂度↓ 克服overfitting
缺点:black boxes,输出结果cannot be easily labeled or directly interpreted
2.2 Clustering聚类: 将相似变量聚合cohesion为cluster,不同变量拆分separation
UL+classification
k-means clustering k均值聚类 划分组数后,通过中心点聚类
k均值聚类的机制

设定超参数k=划分组数
随机选取k个中心点(可能会在边缘)
按样本点与中心点距离来分出类(红蓝)
将中心点移动到类的中心,再次聚类,按距离重新划分红蓝
直到再聚类也不改变类内的数据点时,停止,划分完毕
优点
适用于very large data sets
很好地刻画patterns in high dimensional data
缺点
需要人为设定hyperparameter(k)
通过尝试a wide range of k来确定最优k
hierarchical clustering 层次聚类 自上而下或自下而上聚类
类型
agglomerative (or bottom-up) clustering 聚集型聚类(自下而上)

divisive (or top-down) clustering分裂型聚类 (自上而下)

3. 神经网络(介于监督非监督之间) Neutral networks
3.1 模拟人的神经传导过程

术语
输入层an input layer仅一层
隐藏层hidden layers(黑箱,可有多层)
输出层an output layer仅一层
3.2 ANN人工神经网络:从例子说开

对layers设置hyperparameters,规定各层的#nodes
[输入层]的每个nodes收到input values后, 通过两种函数进行加工、输出
整合运算summation operator:通过赋权将input value整合成total net input
权重为Synaptic weights,是随机数
激活函数activation function:将total net input输出到下一层(隐藏层), 成为第一层隐藏层的输入值
像一个like a light dimmer 调高调低光(tot. net input)的强度
建模(向前计算forward propagation):第一层隐藏层接收到输入值(i.e.激活函数的输出值)后,按相同逻辑传导至下一层隐藏层
算例

修正模型(调整synaptic weights权重):
向后计算Backward propagation:向前计算的逆运算,从output出发求input,从而得到synapticweights
超参数调整Revision of hyperparameters:一开始是人为设定的结点数,现在使用out-sample-data进行调试,确定各结点数
3.3 神经网络NN的应用
Deep Neural networks (DDNs) 深度神经网络 隐藏层多于20层
特征:隐藏层多于20层(#hidden layers>20)

应用场景:image, pattern and speech recognition problems
Reinforcement learning (RL) algorithms 强化神经网络 在大量试错中修正算法
特征:从大量trials and error中获得反馈(通过试错获得强化)
例子:AlphaGo
LM7 Big Data Projects
1. Big Data Introduction:
主要特征:五V
Volume 数据量的大小
Variety 数据来源的广度(除了表格呈现外,还包括文本、图像...)
Velocity 数据产生和被收集的速度
Veracity 数据可靠性
Value 数据价值
Data Analysis Steps

1. Conceptualization of the modeling task.
2. Data collection
3. Data preparation预处理 and wrangling清洗
4. Data exploration数据挖掘(ML之前的人工分析)
5. Model training
2. Big Data Analysis: Structured Data
可用表格呈现的numeric data
2.1 Conceptualization of the modeling task.

2.2 Data collection
internal data
external data
2.3 Data preparation预处理 and wrangling清洗

Data preparation (cleansing)数据清洗
Incompleteness error数据缺失
Invalidity error数据范围不合理
Inaccuracy error答非所问
Inconsistency error内容不一致
Non-uniformity error格式不统一
Duplication error重复数据
outlier处理异常值
trimming (truncation)截位:剔除异常值
winsorization缩尾:大于上/下界a/b的数据一律取上/下界
Data Wrangling (Preprocessing)数据预处理
Transformation数据转换

Extraction: 基于已有变量新建变量
Aggregation: 变量加总
Filtration: 删除不必要的行(研究美国居民,删除加拿大居民数据)
Selection: 删除不必要的列(id和姓名 删除姓名)
Conversion: 数据的性质或单位进行转换(某个观测值的CAD转换为USD)
Scaling变量缩放
Normalization归一化
优点:对数据分布无要求
缺点:敏感于outlier
Standardization标准化
优点:能减缓outlier影响
缺点:需~Norm Distr
2.4 Data exploration数据挖掘(ML之前的人工分析)

Exploratory Data Analysis (EDA) 探索性数据分析:基本的统计学指标、可视化
serve as a communication medium
Summary statistics
Visualizations

histgram
box-whisker plot
heat map
scatter plot
Feature Selection (FS) 特征选择:删除冗余变量k↓防止overfitting
与降维dim reduc的异同
相同:reduce #features(k↓)
不同:FS不产生新变量,单纯剔除变量
Feature Engineering (FE) 特征工程:把不利于ML处理的X转化为利于处理的X,防止underfitting
optimize and improve X:基于已有变量新建变量(生日→年龄)
新建变量有利于ML分析(年龄是定距变量)
模型好坏取决于FS和FE
2.5 Model training

Method Selection 选择合适的机器学习方法

Supervised or unsupervised learning.
Type of data. (numerical, text, speech, image)
Size of data.
Number of feature and number of observation
Performance Evaluation 性能评估
同样适用与unstructured data
对于binary classification模型的评价方法
(1)Error analysis:混淆矩阵Confusion matrix
精度Precision (P) = TP/(TP + FP)
横看
精度Precision越高越好
如果Type(I) error越高,则精度越低
检索率Recall (R) = TP/(TP + FN)
竖看
召回率Recall越高越好
如果Type(II) error越高,则召回率越低
准度Accuracy = (TP + TN)/(TP + FP + TN + FN)
准度Accuracy越高越好
F1 score = (2 * P * R)/(P + R) 属于调和平均值

用于unequal class distribution
F1 score越高越好
(2)Receiver Operating Characteristic (ROC)
基本原理: false positive rate (x-axis) 和true positive rate (y-axis)之间权衡取舍
under the curve (AUC)衡量模型表现的好坏:面积越大越好(曲线越convex越好)
AUC close to 1.0:perfect prediction
AUC of 0.5:random guessing
对于regression 模型的评价方法
(3)Root mean square error (RMSE): 样本外ε的S.E.,越低越好
Model Tuning 模型调整:对参数和超参数调整
Parameters调整:由模型基于training sample进行最优化
Hyperparameter调整
对于单个hyperparameter:使用Confusion matrix计算F1 Score越高越好
对于多个hyperparameters:
grid search网格搜索:列网格阵,将超参数成对划分,哪一对的confusion matrix计算出来的F1 score越高,则哪一对超参数最好

Ceiling analysis上限分析

3. Big Data Analysis: Unstructured Data
以文本分析为例
3.1 Text problem formulation. eg研究正面/负面情绪
3.2 Data collection (curation).
3.3 Text preparation预处理 and wrangling清洗
Text Preparation (Cleansing)

Remove HTML tags.删除网页标签
Remove punctuations.删除标点符号(可以用DollarSign这样的文本代替符号$)
需要加annotation(告诉计算机这里删除了)
Remove numbers.删除数字
需要加annotation(告诉计算机这里删除了)
Remove white spaces.删除空格
Text Wrangling (Preprocessing)
a token = a word
tokenization
第1步 normalization
Lowercasing.小写化
Removal of stop words.删除停顿次
Stemming词干提取(用于缩短单词长度)
Lemmatization词形还原(用于缩短单词长度)
第2步bag-of-words (BOW)
N-gram:涉及到顺序重要的词组(eg 天安门 不能拆/调顺序),用下划线链接。两个单词组成bigram,三个单词组成trigram
最终形成词包BOW
第3步 document term matrix (DTM). (词频矩阵,可用数据表格形式呈现)
3.4 Text exploration数据挖掘(ML之前的人工分析)
Text Exploratory (EDA) 探索性数据分析:基本的统计学指标、可视化
Summary statistics
Term frequency词频
Co-occurrence两个或以上的词同时出现
Visualizations:云图word cloud
Feature Selection (FS) 特征选择:删除次要冗余token防止overfitting
处理对象
High-frequency words:is a the 等stop word
Low-frequency words:与主题无关的
FS methods特征选择的方法
Frequency: 高频且有意义的单词
A total of 44,151 non-unique tokens are in the 2,180 sentences
术语
Term frequency (TF) at the corpus level (collection frequency, CF)
used for vocabulary pruning to remove noise features (太高频=无意义is a the,太低频=专有名词或罕见词组)
CF=单词出现次数/语库总字数
TF (Term Frequency) at sentence level
(常用)句内TF=句子中出现该单词的次数/句子中的总单词数
DF (Document Frequency) 用以区分出无意义的单词如is the
DF=出现该词的句子数/总句子数
越低越好:单词越特殊、越含有意义信息
IDF (Inverse Document Frequency) : IDF = log(1/DF) 越高越好
衡量指标:TF-IDF = TF * IDF
TF-IDF of token = 出现频率*意义程度,若数值越高,则越应该保留
Chi-square: 单词含义和文章主题的独立性
H0: 单词独立与文章
如果χ²>C.V., rej H0,, in favour of 单词不独立与文章(应保留)
χ²↑ token越好
Mutual information: 衡量token对文本类别区分的贡献
MI接近1,越有利用于区分文本类别;MI接近0,越难以用于区分文本类别

Feature Engineering (FE) 特征工程:把不利于ML处理的X转化为利于处理的X,prevent underfitting
Numbers 重要数字要改回来(eg“iPhone/number4/ is better than iPhone /number1/”这里的数字有意义)
N-grams 标记单词构成的特定词组
Name entity recognition (NER)专有名词打标签(“合肥”是地名,不是两个胖子合在一起的)
Parts of speech (POS)根据词性打标签(Microsoft是一个名词,1947是一个date)

模型好坏取决于FS和FE
3.5 Model training
Method Selection 此例用logistics regression

以发生的概率P作为Y的解释,P>threshold则为正面态度
threshold value设定:使模型accuracy最高时的Prob(optimal)
Performance Evaluation 性能评估
使用ROC法进行评估

Model Tuning 模型调整:对参数和超参数调整
出现overfitting则用正则化LASSO进行调整
Equity
LM1 Equity Valuation: Applications and Processes
1. Value Definitions and Applications
Intrinsic value
Alpha(abnormal return)
Going-Concern Value 永续经营价值
Liquidation Value 清算价值
Fair Market Value(公允交易下的价值,等于intrinsic value)
Investment Value(考虑投资并购后的协同效应)
在有效市场下,Liquidation value < intrinsic value = fair market value < investment value
Applications of Equity Valuation
Stock selection
Inferring (extracting) market expectations
Evaluating corporate events
Fairness opinions
Evaluating business strategies and models
Communicating with analysts and shareholders
Appraising private businesses
Share-based payment (compensation)
2. The Valuation Process
基本流程
Understand the business;

Forecast company performance;

Select the appropriate valuation model;

最重要Convert the forecasts into a valuation(估值计算);

Apply the valuation conclusions.

对于集团公司的估值方法:Sum-of-the-parts valuation(breakup value or private market value)
子公司分开估值再求和 Va+Vb+Vc+Vd>Vabcd
Inefficiency of internal capital markets
Endogenous factors
Research measurement errors
估值模型选择
Consistent with the characteristics of the company being valued
Appropriate given the availability and quality of data
Consistent with the purpose of valuation
LM2 Discounted Dividend Valuation
三种估值模型的适用条件对比
DDM股利折现模型
存在dividend payments
div policy is clear 且 div与公司earnings相关
从minority shareholder视角估值
FCF自由现金流模型
not have a dividend payment
div PMT not related to earnings且 FCF与公司earnings相关
从controlling shareholder视角估值
RI剩余收益模型
not have a dividend payment
在未来FCF为负数
财报透明 and high-quality earnings
1. DDM Formula
2. The Gordon Growth Model
Assumptions
D1: expects to pay a dividend, D, in one year
g: dividends will grow at a constant rate
r>g: growth rate (g) is less than the required rate (r).
GGM formula
Relationship between GGM and GDP:可用GDP增长率作为growth rate g
如果远高于GDP增长率,则用use a multistage DDM
GGM中g的正负没有限制
g=0时,变为fixed-rate perpetual preferred stock
D: dividend on such preferred
r: discount rate (capitalization rate)
g<0时,也可以刻画公司越来越差的情景
GGM中的收益率r拆解
公式变形求r
Estimating a Required Return
增长率g:rate of value or capital appreciation (also called the capital gains yield).
Share Repurchases: GGM考虑了股票回购,因此有回购时,也可用于估值
implied g 与高估低估
If implied g > expected g → overvalued
If implied g < expected g → undervalued
Present Value of Growth Opportunities (PVGO)
全收益分红假设

E1/r:no-growth value per share
PVGO = V0 - E1/r
P/E ratio based on forecasted earnings
注意与justified(fundamental) P/E相区分
优点
适用于stable, mature, dividend paying companies
适用于valuing market indices(用大盘指数估值)
Easily communicated and explained
Used to determine g, r, and PVGO
Supplement other, more complex valuation methods
缺点
sensitive to r&g
不适用于non-dividend stocks
不适用于unpredictable growth patterns (use multi-stage model)
3. Multistage Dividend Discount Models
TerminalValue_m = D_m+1/(r-g_L) m为stage1最后一期,还要折现m期
Two-stage DDM

Three-stage DDM

H-Model

在增长率上修正Half(triangle) V0 = [D0*(1+g_L+(t/2)*(g_S-g_L))]/(r-g_L)
Multistage Model
Spreadsheet modeling用Excel开展DDM,优势
any pattern of dividend growth
project different growth rates
inherent flexibility and computational accuracy
4. Growth Phase, Transitional Phase, And Maturity Phase
不同阶段的公司特征 和匹配的估值模型
The Financial Determinants of Growth Rates: sustainable growth rate
作为sustainable g,应当选用maturity时的相关数据
逻辑:收益留存率retain rate(节流) * 净资产收益率ROE(开源)
GGM中的增长率g是一系列指标 dividends, earnings, and stock value的增长率

求解g值:PRAT model
计算题可能需要
profit margin (P), retention rate (R), asset turnover (A), and financial leverage (T).
5. Equity Analysis
Ratio analysis
DuPont model
计算题可能先求ROE
ROE = tax retention rate * interest burden * operating margin * asset turnover * leverage
LM1 Free Cash Flow Valuation
1. Introduction to Free Cash Flows
使用free cash flow (FCFF/FCFE)的原因

FCF的计算原理:先求FCFF再求FCFE
扣除现金形式的operating expenses(包括Tax但不包括Int)
先求FCF to firm(包含债权人),因此不用扣除利息Int
扣除短期和长期资本投资支出WCinv和FCinv
得到FCF to firm
扣除利息费用Int(1-t),扣除本金偿还principal rePMT并加上新增债务principal borrowed(合称:加上净债务借款net borrowings)
得到FCF to equity
适时选择FCFF or FCFE
优先使用FCFE:easier and straight forward to use
前者失效时选择FCFF:当高负债(高杠杆)导致negative FCFE 和 融资结构D&E此起彼伏波动 changing capital structure时,使用FCFF
估值原理:选取适当disct rate进行GGM折现
FCFF
增长率g:FCFF_t = FCFF_t-1 * (1+g)
折现率:使用WACC(包含股权和债务)
DDM型估值:Firm Value = FCFF1/(WACC-g) =FCFF0*(1+g)/(WACC-g)
FCFE
增长率g:FCFE_t = FCFE_t-1 * (1+g)
折现率:使用r_e(仅包含股权,多种计算来源)

DDM型估值:Equity Value = FCFE1/(r-g) =FCFE0*(1+g)/(r-g)
Equity Value = Firm Value - MktValue(Debt)
2. FCFF and FCFE Valuation Approaches
算例
 注意:①WCinv中cash类和debt(付息)类的识别,②GrossValue(GV)一般指gross fixed assets 
先算FCFF:知1求3
再算FCFE:知1求4
在4种FCFF后-Int*(1-t)+NB即可
若干指标不由FinReport直接给出
• NCC (Non-cash charges) Adjustments

计算FCF时,带来收益的一并减去
①D&A:added back
②Impairment of intangibles:added back
③Restructuring charges
expense:added back
income:subtracted
④non-operating activity
losses:added back
gains:subtracted
⑤Amortization of LT bond
discount:added back
 利息费用要算114元(按照12%利率去融资)[NI记-114],但实际只付给债务人100元的真金白银[CF记-100],那么相差的14即为摊销Amortization,不用以cash形式支出,故在求FCF时要加回【CR<r:少交现金,FCF↑】
premium:subtracted
⑥Deferred Taxes
DTL(当前少缴税手头富余):added back
DTA(当前多缴税):subtracted
• WCinv Calculation
WCinv本身是增量△WC
Working Capital=(Current Asset−Cash)−(Current liability−Debt)
CA是被占用作oprt活动,因此不含cash, cash equivalents
CL是用于oprt的债务,不含融资行为的付息债务如notes payable, and current portion of long-term debt
• FCinv Calculation
是净额(net amount)概念,即求difference
No disposal不处置固定资产:
CapExp=GV末-GV初=(BV末+AD末)-(BV初+AD初)=BV末-BV初+DepExp
GV:gross value 原值(购买价值) BV:book value 净值(账面价值) AD:Accum.Depr.累计折旧 原值-累计折旧=净值
GrossValue(GV)一般指gross fixed assets
算例

Disposal处置固定资产:
CFA II 一般不考
CapExp=GV末-GV初+GV_disposal=BV末-BV初+BV_disposal+DepExp

Proceeds=G/L+BV_disposal

FCinv=BV末-BV初+Depr-G/L
算例

• Net borrowing Calculation
定义式(无法从财报获取指标):NB = 长短期new debt - 长短期debt repayments
计算式:NB = 长短期debt末 - 长短期debt初
包括note payable、long term debt等等项目
• Preferred Stocks issue 导致NI的口径差异
NI_total = NI_comm + Div_pref
求FCFF时,只给NI_comm,要加上Div_pref
求FCFE时,只给NI_total,要减去Div_pref
• Forecasting FCFE with Target Capital Structure 给定DR求解NB并FCFE
Target debt ratio (DR): NB=△Cap*DR
推导过程:从NB=△Cap*DR出发,代入Cap=Equity+Debt-Cash求得NB,算出FCFE=FCFF-Int*(1-t)+NB

公式
理解
FCinv-Depr是长期资产的新增净投资,WCinv是短期资产新增净投资,两者相加*DR(由新增债务融资来cover的部分)
理解
FCinv-Depr是长期资产的新增净投资,WCinv是短期资产新增净投资,两者各自*(1-DR)代表由新增股权融资来cover的部分(花出去的费用)。由NI减去两者,代表NI中剩下真金白银留在手中的部分
(FCinv-Depr)又叫NetFCinv
(FCinv-Depr)中减去的是Depr而非NCC,即默认其他NCC项目不变
3. Other Issues in Free Cash Flow Analysis
Proxy for FCF:EBITDA、NI都是poor proxy (因为是权责发生制,但FCF是收付实现制)
Transactions between公司和股东:不会影响FCF
eg现金分红,回购,增发
杠杆率变化leverage chage:对FCFE有影响,对FCFF无影响
4. Free Cash Flow Model Variations
Single-stage model

Two-stage model
算例

以2016或2017为stage1终点,结果一样,2016可以少算一年
计算注意:GGM的标准形式,用D1更标准

5. Sensitivity Analysis in FCFF and FCFE
探究单个input变化对final valuation的影响
可以发现forcasting error
FCF的base-yr value对计算结果影响很大
6. Nonoperating assets and firm value
FCFF算出的FirmValue只是经营性部分(operating),不包括非经营性FCF
做题看选项,相差一个题目中提到的整数(即为non-operating)

LM2 Market-Based Valuation: Price and Enterprise Value Multiples
1. Price Multiples
1. Price and enterprise value multiples in valuation
Price multiples对股价估值:包括P/E P/B P/S P/CF P/Div
Enterprise value multiples对公司估值:EV/EBITDA
估值逻辑:The Method of Comparables、The Method Based on Forecasted Fundamentals
估值原则:在比较重判断under-/over-valued
估值原则:将multiples与benchmark相比较. 越低越好(被低估要买入)
例外:D/P(Dividend yield) or E/P 越高越好(回报高要买入)
五种价格倍数
• P/E
优缺点:
优点:
Earnings is the chief driver of investment value
P/E ratio is widely recognized and used
判断long-run average returns:偏高 未来下跌,偏低 未来上涨
缺点:
EPS can be zero, negative, or insignificantly small
nonrecurring components
managers may distort EPS
类型
trailing P/E = P0/E0
EPS的选择注意
1) prefer to use Diluted EPS

2) Adjustments for Nonrecurring Items 剔除非经常性项目
3) Adjustments for Business-Cycle Influences (防止Molodovsky effect) normalizing EPS

historical average EPS
没有考虑 size of biz
average ROE

4) Adjustments for Comparability with Other Companies 统一会计口径
5) Dealing with Extremely Low, Zero, or Negative Earnings 采用E/P即earnings yield
又叫做inverse price ratio.用于meaningful ranking
forward P/E = P0/E1 基于盈利预测
估值Valuation Based on Forecasted Fundamentals
justified P/E
含有justified代表要用于估值模型
如未来Earning有重大变化用Leading P/E,比如acquisition
用DCF 类model,求解P/E
P/E与r负相关,与(1-b)和g正相关
实证检验

估值Valuation Based on Comparables 比较对象
peer group of companies
industry or sector
representative equity index
average past P/E for a stock
P/E to Growth (PEG) Ratio
①低PEGs&高P/E are more attractive
growth stock一般有高P/E,但高P/E不一定是成长股(有可能是被炒的)
②must be used with care
Assumes a linear relationship 但P/E和g的关系非线性
没有考虑:风险、增长持续期
• P/B
BV:Book Value 普通股common share的每股收益
Common-share Equity = Tot Equity - Senior Equity
(Equity_comm)/(#comm_shares) = book value per share
优缺点:
优点:
BV almost always>0
BV more stable than EPS
适合liquid assets为主的公司(交易频繁→BV接近MV)如金融机构
可用于 not going concern公司估值
判断long-run average returns
缺点:
没有考虑Size differences
Influenced by accounting choices受会计标准
BV≠MV due to inflation(BV<<MV)/ technology(BV>>MV)
没有考虑表外资产的英雄昂
Justified P/B

一般只考虑trailing P/B
ROE↑时,P/B↑;ROE和r的spread↑,P/B↑
Excluding intangible assets 如商誉
• P/S
优缺点:
优点:
Meaningful even for distressed firms 可用于财务困境的公司
Sales revenue not easily manipulated
not affected by revenue recognition
Sales are positive even when EPS is negative
Not as volatile as P/E ratios
判断long-run average returns
Useful in valuing mature, cyclical, and start-up firms适用于各阶段公司
differences in long-run average returns.
缺点:
没有反应成本、利润(high sales rev ≠ high profits)
人为因素扭曲 distort sales
但是比EPS的扭曲要少,因为是top line而非bottom line
Justified P/S

E/S:Earning/Sales为利润率profit margin
• P/CF
没考过
优缺点:
优点:
CF less subject to manipulate than EPS
More stable than P/E
避开了problems of quality of reported earning
Empirical evidence supported
缺点:
Difficult to estimate true CFO
FCFE better but more volatile & 经常为负
不同口径的CF
Total CF:
无视了 noncash revenue and changes in net working capital
Adjusted CFO
FCFE:
EBITDA:Used in enterprise value-to-EBITDA ratio
• Price to Dividends and Dividend Yield
优缺点:
优点:
Dividend yield is a component of total return
Dividends are a less risky component of total return than capital appreciation
缺点:
only one component of total return 没有反应return的全部信息
dividend displacement of earnings (trade off 当前收到股利 and 不分红用于公司发展得到未来高股价、股利)
relative safety of dividends的假设: 当前收到股利比不分红得到未来高股利股价更安全
类型
trailing P/D = P0/D0

g↑时,D/P↓
r↑时,D/P↑
forward P/D = P0/D1 基于股利预测

当D=0时用div yield:DivYld=D/P
2. Enterprise Value Multiples: EV/EBITDA
EV= MV(common stock) + MV(preferred equity) + MV(debt) – cash and investments
MV=MVIC=MV(Equity)+MV(Debt)
EBITDA = NI + I + T +D&A (与盈利能力无关)
优缺点:
优点:
可用于Co. with different degrees of financial leverage(剔除了Int的影响)
EBITDA useful for valuing capital-intensive business
EBITDA usually positive even when EPS is not
缺点:
If WC↑, EBITDA will overstate CFO

FCFF is more strongly linked with valuation theory than EBITDA
3. International Considerations When Using Multiples
国际比较中,适合用P/CF,因为P/CFO and P/FCFE will generally be least affected by accounting differences
4. Momentum Valuation Indicators 属于技术分析
Momentum indicators动量指标:当前价格的时间序列TS_t/过去价格时间序列TS_t-k,positive momentum表明看涨
Unexpected Earnings
Standardized Unexpected Earnings (SUE)
Relative strength (RSTR) indicators相对强度指标:一支股票的表现与其历史表现或一组股票的表现相比较
5. Averaging Multiples: Weighted Harmonic Mean
站在portfolio视角分析乘数
mitigate the impact of large outliers
算例

LM3 Residual Income Valuation
1. Concept of Residual Income 属于超额收益的范畴
Residual income (RI) ∈股东

Economic value added (EVA) ∈公司
EVA即为公金里的EconProfit
TotCap = MV(D) + MV(E)
Market value added (MVA)

MV=MVIC=MV(Equity)+MV(Debt)
2. Residual Income Valuation Model
估值逻辑:将intrinsic value of equity分解为两部分
B0:Adjusted current book value of equity
∑RI_t/(1+r)^t:Present value of expected future RI
RI model 估值公式
特点
对第二部分(DCF)不敏感(RI可以<0, r&g的选取可以不准确):less sensitive to terminal value estimates. V0主要由B0确定,DCF部分的影响小
Difference with DDM & FCFE
假设不同
DCF的占比不同:RI只局部DCF,DDM&FCF是全局DCF
Calculation of RI:基于BV计算RI
3. Single-stage Model 永续增长模型
single-stage RI model
来自
反推 implied g
Assumption: intrinsic value = P_mkt
与justified P/B的关系(推导)

注意:含justified表明与GGM相联系
4. Multistage Models
Motivation:现实中 公司发展成熟后ROE↓ over time 并趋近r_e
估值逻辑:
Intrinsic value = ①BV + ②PV(interim high-growth RI) + ③PV(continuing RI)
图示: 情况3高频考点
情况1:永续年金型
情况2:③PV=0
情况3:ROE↓并趋近r,等价于g<0的GGM,设1+g=w<1

w为持续因子,代表以g(=1-w<0)的速度下降
情况4:由PVt=Pt-Bt得来

算例

Step 1: 计算BV_0 per share
Step 2: 计算RI model在(0-T)时期的价值
Step 3: 计算TV at time T
RI模型评价
优缺点
优点
TV所占比重不大(DCF的参数误差对估值的影响小)
数据来源可靠readily available accounting data
no dividends or 负RI仍可用
CF unpredictable仍可用
反映真实盈利能力focus on economic profitability
缺点
NI manipulation by management
inputs require significant adjustments
前提是clean surplus relation,违背时model失效
假设了cost of debt capital完全反应在IntExp之中,实际上不完全反应
适用场景
适用于
not pay dividends/ div unpredictable(DDM失效)
cash flows are negative(FCF失效)
uncertainty exists in forecasting terminal values(TV不稳定,DCF类失效)
不适用于
clean surplus假设失效时
clean surplus:Equity长期来看只受R/E影响,而不受Cap.和OCI影响
B0、ROE不确定时
5. Accounting and International Considerations 使用RI model需要关注
前提假设clean surplus relationship的违背来源: Cap.和OCI的不稳定
 售衍老外估对OCI的影响
balance sheet adjustments for fair value BV不一定等于fair value 需调整
表外资产和负债 significant off-balance sheet A and Liab
若干项目BV≠FV需要调整

intangible assets;
Intangibles recognized at acquisition:广告花费记为expense而非Asset;如果acquirer overpays,则未来RI下降
R&D expenditures:unproductive R&D expenditures要求调低RI;mature company的ROE如果能反应R&D expense则无需调整
nonrecurring items:overestimates and underestimates of future RI

aggressive accounting practices:需要评估accounting policies是否激进
international considerations:考虑会计口径,调整后再比较
LM4 Private Company Valuation
1. Private and Public Company Valuation: Similarities and Contrasts
Company-specific factors
Stage of life-cycle:非上市公司处于早期
Size:非上市公司更小
Overlap of shareholders and mgmt:管理层即为股东
Quality and depth of management:管理深度不够(自己管理,不聘请职业经理)
Quality of financial and other information:财务数据质量差 且 难以获取(非公开)
Pressure from short-term investors:可集中关注长期投资
Taxes:为了避税,故意underreport净利润
Stock-specific factors
Liquidity of equity interests in business:流动性差
Potential agreements restricting liquidity:股东协议,限制出售股票,低流通性
Concentration of control:集中在少数人手中
2. Reasons for Valuation
Transactions encompass events
Private financing
IPO
Sale in an acquisition
Bankruptcy
Share-based payment (compensation)
Compliance-related valuation: law or regulation
Financial reporting:eg商誉减值核算
Tax purposes
Litigation-related valuation
damages, lost profits, shareholder disputes, and divorce.
专题:6种values
Fair market value:市场上,双方信息对称且自愿交易形成的公允价值
Market value:当前市场上的价格
Fair value (financial reporting):财报中的公允价值
Fair value (litigation):法律法规界定的公允价值
Investment value:在内在价值的基础上,包含了支付给协同效应的价格
Intrinsic value:真正的正确的价格
3. Private Company Valuation Methods
3issues+3methods
第一步:3 issues 调整
1 Normalized Earnings 例题

剔除nonrecurring, non-economic
剔除故意↓NI的项目(虚高的cost&expense)
剔除real estate from the operating(与经营无关的房产)
具体需要removing any revenues and expenses
剔除的原因是: 房产与经营活动different levels of risk and expected future growth
2 Strategic and Nonstrategic Buyers

strategic transaction:考虑synergies,需要将revenue↑&cost↓
financial transaction:no material synergies 不做调整
3 Estimating Cash Flow:分scenarios分析
每个scenario下有各自的折现率(反应本情景下的风险水平)
使用probability-weighted average求出estimated scenario values
使用FCF的原因:非上市不发Div(DDM失效),非上市不满足earning surplus假设(RI失效),FCFE容易受cap structure波动影响,所以使用FCFF
确定合适的Discount Rate
CAPM:很可能不合适
Expanded CAPM:多因子 CAPM a premium for small size and company-specific risk.

影响非上市公司r的众多因素

Build-up method:将CAPM中 令β=1
第二步:3 methods 估值
1 Income Approach (本质是DCF类)
FCF法: 即2-stage model,和前面的模型完全相同
CCF法:capitalized cash flow method (single stage)
(r-g)和(WACC-g)叫做capitalization rate
适用场景:
smaller private company
no projections are available
market pricing evidence or transactions is limited
EEM法: excess earnings method
Firm Value = WorkingCap + FixedAsset + IntangibleAsset

三步求解FirmValue(例题)

①求无形资产I.A:Earnings-E(WC)-E(FA)=EE (excess earning)
总盈利=WC贡献的盈利+F.A贡献的盈利+I.A贡献的盈利
②DCF永续折现 求I.A的价值
③加总求和Firm Value = WC + F.A + I.A
2 Market Approach (本质是PriceMultiple类)
关键在于选择benchmark
使用倍数法 需要调整
要求倍数reflect both expected risk and growth
再评估Risk and growth assumptions
考虑stock-specific factors
不同规模的private company采用不同指标
Large mature: MVIC/EBITDA multiples
Market value of invested capital (MVIC) =MV(Equity)+MV(Debt)
Small: net income multiples 市净率
Extremely small: revenue multiples 市销率
三大benchmark
Guideline public company method 可比上市公司GPCM
估值步骤

step1 算出可比公司的relevant pricing multiples
step2 进行调整reflecting the relative risk and growth
risk premium
control premium:V(control_share)>V(minority_share)

优缺点
优点:large pool of Co.; info available
缺点: 在风险和增长的调整上 考虑comparability and subjectivity
Guideline transactions method 可比交易GTM
估值步骤
算出可比交易的relevant pricing multiples 不需要调整control premium,因为可比交易已经control
优缺点
优点:Synergies;Contingent consideration;Noncash consideration
缺点: Availability of data;Date of data
Prior transaction method 可比过往交易PTM
没考过
估值步骤
使用可比过往交易的relevant pricing multiples 既可以直接使用actual price paid or the multiples implied from the transaction
适用于给minority equity interest估值,而非control
3 Asset-Based Approach (ABA) 又叫cost approach
Fair Value: Equity = Asset -Liab
不准确1:资产A核算困难,特别是IntangibleAsset
不准确2:容易低估FV,因为在going concern下不考虑PVGO
适用场景
Natural resource companies;
Financial companies, such as bank;
Holding (investment) companies, such as real estate investment trusts (REITs) and closed end investment companies (CEICs);
Asset容易明确计量(mkt频繁交易)
Very small businesses with limited intangible value or early stage companies.
不用考虑PVGO
4. Discount / Premium in PE Valuation
Motivation:给control interest估值时,若benchmark是minority interes),则需要对minority的价值进行折价调整
子主题
Discount for lack of control没有控制权导致的折价
Discount of lack of marketability流通性差导致的折价
题目给定
Total discount
折扣具有叠加作用

FinStatmAnalysis
LM1 Intercorporate Investments
1. Overview:不同类型股权投资的核算方法
2. Investments in Financial Assets: IFRS 9
对B/S和I/S的影响
IFRS 9 Classification

Amortized cost (Debt only):满足两个条件
Business model test:收取合同现金流
Cash flow characteristic test:收取合同现金流仅仅包括本金、利息
FVPL: FV through profit or loss (D & E)
Debt——held-for-trading or amortized cost results in accounting mismatch
accounting mismatch

Equity——held-for-trading must be classified as FVPL; others may be classified as either FVTPL or FVOCI, irrevocable
FVOCI: FV through OCI (D & E):持有意图不明确是trading 还是 HTM
Reclassification改变分类
Equity不允许
Debt允许 only if 满足business model(持有获取合同现金流)
Amort→FVPL:将Amort调整为FV,造成的G/L计入I/S
FVPL→Amort:将FV调整为Amort,造成I/S中的的G/L计入账面价值
注意:FVOCI不能调整
3. Associate:Equity Method
Assoc:具有significant influenced的实体
typical ownership interests 20%-50%
criteria for significant influence

财务计量方法
B/S
CV of investment = cost of investment + (adj. accum. NI – accum. div) * %owned

cost of investment: 通常是以超过BV的价格收购 =BV + FVapprec + GoodWill 三部分组成

FVapprec=(FV-BV)*%
算例
adj. accum. NI:将收购目标公司的NI从BV转为FV Inv影响COGS; F/A影响Depr; I/A影响Amort 算新NI
因为是按FV收购,需统一口径
以Depr为例:根据T公司的BV和FV算出Depr(BV)=4k[现折旧] 和 Depr(FV)=7.5k[应折旧],折少了 要多折,则Depr.adj.=-3.5k,新NI=旧NI-3500

算例

Transactions with Assoc:NI含不符合经营利润的部分 剔除transaction with associates
Upstream (associates卖货给investor): 记录在associate's I/S(例子是unrealized G)

Downstream (investor卖货给associates): 记录在investor's I/S(例子是realized G)

注意realized要考虑货物的几分之几没有被真正售出
I/S
Gain = NI * %owned
Associate - Fair Value Option 直接以FV计量
US GAAP:可以对所有entities 直接用FV计量
Under IFRS:主要适用于金融公司(VC, mutual funds, unit trusts...)
一旦采用FV option,则不可转为其他会计核算方式irrevocable
Associate - Impairment
IFRS:Recoverable Amount(标准) < Carrying Value 则需要减值
RcovAmount=max{disposal value直接卖价, value in use未来现金流折现}
US GAAP:FV < CV 则需要减值
FV的衡量(优先到劣后)
mkt value
找comparable公司
valuation估值
两种会计标准下均不可转回 impairment loss
4. Controlling interest investment: Acquisition Method
核心考点
parent用BV,sub用FV
Acq公司100%合并T公司
B/S并表:Asset、Liab分别合并,Equity不合并(防止重复计算)
少数股东MI配平
注意1:Acq公司的BV中,收购Investment不计入Asset
注意2:T公司的FV中,FV(Equity)不合并,故只有Acq公司的FV(Equity)
注意3:Acq的Invest - T的FV(Equity) = GoodWill
理解:股权收购
例子

I/S并表:将D&A expense从BV调到FV,并得到新NI,再用于合并
算例

商誉GW=实付-应付
partial GW:<100%收购时算出的GW
full GW:=100%收购时算出的GW
full GW只是在100%收购下算出的值,并不要求真的100%收购,eg US GAPP要求GW一律用full GW计算
说明MI和GW的例子
不含GW的收购(Purchase=FV)
含partial GW的收购(Purchase>FV)
含full GW的收购(Purchase>FV)
4. Joint ventures:一般用Equity Method 某些情况下用proportionate consolidation
本质上是controlling乘上百分比:based on share of the assets, liabilities, income, and expenses of the joint venture.
不用报告MI: minority owner's interest
6. Effect of the methods
trade-off:控制程度低 但财务指标好看
再审视equity method:存在的问题
① 账面收益E(equity method)> 实际收益E(passive investment),
②不并表One-line consolidation,导致investee的经营业绩无法体现在investor上
7. Business combination
Business Combination
Merger兼并 A + B = A 不并表
Acquisition收购 A + B = (A + B) 并表
Consolidation A + B = C 新表,不并表
SPE or Variable Interests Entity: 被证明有控制权关系时 并表
Acquisition Method
Goodwill
Impaired减值测试: carrying value > fair value
impairment loss在I/S中单列
Under IFRS (single step approach)
IFRS:Recoverable Amount(标准) < CV(GW) 则需要减值
RcovAmount=max{disposal value直接卖价, value in use未来现金流折现}
如果GW减到0以下,则按比例扣减other non-cash asset

Under U.S. GAAP (two-step approach)
Step 1: FV(GW) < CV(GW) 则需要减值 Step 2: 计量减值损失 impairment loss = impliedFV(GW) - CV(GW)
impliedFV(GW) 此刻再收购的FV(GW)
如果GW减到0以下,则不作处理

US GAAP要求Full goodwill; IFRS 允许partial or full goodwill.
Negative Goodwill is Recognized as a gain
Goodwill & MI process
GW算例

full goodwill
partial goodwill
Bargain Acquisition:收购价<FV(Equity) 不计入GW,而是 gain in profit or loss
Accounting treatment演变史:Pooling-of-interests method → Purchase method → acquisition method
8. SPE and VIE
US GAAP, variable interest entity (VIE): 要想不并表,三者缺一不可
Under IFRS, 只要显示出对SPE的控制权,就要并表
LM2 Employee Compensation: Post-Employment and Share-based
1. Post-retirement plan overview
typical post-retirement plans
Defined-contribution pension plan (DC):风险由employee承担

Defined-benefit pension plan (DB):风险由employer承担

Other post-retirement benefits (OPB): DB的医保版本
2. Illustration of DB
Key terminology
算出PBO可用倍数求ABO
PMT=退休前工资*n%,n退休时工龄
PBO算例

Annual unit credit (benefit) per service year = Value at retirement/Years of service
Annual unit credit:只需求出工龄1yr的PV(retire),即可用倍数算出工龄nyr的PV(retire)和相应PBO

算法1:已知工作1yr的PV(m=1)工作M年退休,求工作x年时的PBO
PV(m=1)即为AUC,指站在退休时(工龄M年,开始领钱),以工作1yr作为退休金的累计工龄数计算的,从退休到死亡共N年所有CFs的现值(PMT=基数*M%, I/Y ,N=退休到死亡, FV=0)
step1:倍数PV(m=x)=x*PV(m=1)
step2:站在m时点上,还有M-x年退休 折现PBO(m=x)=PV(m=x)/[(1+r)^(M-x)]
算法2:给定工作年限N和lump sum value at retirement LumpSumValue/N=AUC
算例

Current Service Cost (for 1 year) = AnnualUnitCredit/[(1+r)^(M-m-1)

之所以折现(M-m-1)是因为CSC本身是增量 (公司过了这一年,在存量PBO上增加)
Other Post-employment Benefits (OPB 形式类似与DB的医疗保险)
Healthcare inflation rate:医疗费的增长率(先快后慢,因此用多阶段DCF)

定性考察:benefit obligation↑ and periodic cost↑的影响因素
近期增速 g_S↑
长期永续增速 g_L↑
g_S向g_L的转换时点推迟 (高g持续更长)
财务报表列示
3. Balance sheet presentation
列示:Funded status = FV of the plan assets - PV of the Defined benefit obligation
B/S中唯一以净值列示的科目
deficit:列示为net pension liability.
surplus:列示为net pension asset.但有上限(审慎原则)

Funded Status of a Pension Plan
精算假设
Discount rate
Assumed rate of increase in compensation
Vesting:eg 0-10yr工龄领50%养老金,10yr以上领100%
DB的基数确认(不同方法)
Pay-related plan 用当前工资作为PMT,即ABO
Final-pay plan 用退休前工作作为PMT,即PBO
Final-average-pay plan
Career-average-pay plan
3. Income Statement presentation
IFRS
IFRS的OCI的项目统称Remeasurement
OCI中的科目同向变化
US GAAP
E[R]是收益,抵减成本,因此要(加负号)
OCI中的科目: Unamort of PSC (记为负号) Unamort of 纯A.G/L 按+G/-L Unamort of A[R]-E[R] (记为负号)
改变精算假设的影响 Pension Assumption
4. Pension disclosure 分析师视角下的处理
Assumptions假设不同则公司不可比较
Sponsoring company披露的假设参数
Analyst should evaluate
Net Pension Liability (or Asset) 净值列示掩盖Ast&Liab
在财报附录notes to adjust a company's assets and liabilities for the gross amount
Total Periodic Pension Costs IFRS and U.S下 Periodic pension costs计入I/S or OCI
Net periodic pension cost (i.e. total period pension cost TPPC) = △FudingStatus - Contribution
减去Contr的原因:B/S I/S OCI中的项目都能相互抵消,除了Contr和Cash
算例【包含BP TPPC I/S OCI的计算】

benefit paid?

Total periodic cost?

Period pension cost in P&L and OCI if under IFRS?

Period pension cost in P&L and OCI if under GAAP?

(US)Pension costs被全额计入oprtin exp 但实际有∈financial nature的部分
财报披露的:被记如P&L(i.e. I/S)的pension costs(叫做pension expense),在财报中列入operating expenses,whereas the interest component and asset returns are both non-operating.
pension costs: 一笔进入I/S作为expense,一笔进入OCI
Operating income + full amount of pensions costs reported in the P/L (pension expense) - service costs 财报的经营性收入 加回(错误计算的)pension_exp 减去(真正算经营性收入的oprt exp)CSC
同时 Int_exp 加上(漏记的)Int_Cost
同时 Int_inc 加上(漏记的)A[R] 即作为non-operating income.
注意,不是加回E[R]
算例

(US)Cash Flow被全额计入CFO 但实际有∈CFF的部分
情况一:Periodic contributions实缴 > total pension costs应付
over-contribution:本质是还给员工的欠款 CFO(+) CFF(-)
情况二:Periodic contributions实缴 < total pension costs应付
under-contribution:本质是向员工的融资 CFO(-) CFF(+)
算例

5. Share-based compensation
主要类型
Equity settled:Stock options; Stock grants;
Cash settled: Stock appreciation rights; Phantom shares.
不发真的股票,但享受股票的盈利分红
Disadvantages
激励效果有限:manager对市场影响小(eg熊市下股价难以上升)
扭曲激励:manager变得risk averse or excessive risk-taking
稀释股权:dilute the shareholders’ interests
Stock options – call option
成本计算是在grant date(5.1则记为8个月)到vesting date之间
记为compensation expense
Stock grants直接发股票
LM3 Multinational Operations
1. Foreign currency transactions
Foreign currency transactions:以外币结算的交易
时间节点:交易日≠结算日 产生transaction exposure

transaction date交易日
F/S date财报日(产生T.G/L)
settlement date结算日(产生T.G/L)
不同交易方向下, 汇率变动的影响
Analytical issues and disclosure analysis
FX G/L 列示在 I/S(不同方法下进I/S或者OCI)
不明确在operating or non-operating
2. Classification of currency 货币折算的币种
3. Translation of foreign currency F/S 外币报表折算
3.1根据情景选择折算方法
3.2折算方法:考试通常只考 单个科目的计算,不考全表
算例
   
Current Rate Method (T.G/L进OCI,先I/S后B/S)
用于属地型子公司
Cumm.Transac.Adj是OCI的二级科目,curr method下 unrealized G/L on mone&non-monetary Ast&Lib进入这个CTA科目
Temporal Method (T.G/L进I/S,先B/S后I/S)
用于整合型子公司
3.3特殊科目的处理 Several special items
R/E:retained earning
Current Rate Method:T.G/L进OCI,R/E(年末)通过BASE法则求解

Temporal Method:由于T.G/L不进OCI,E=R/E+Cap.(仅两部分构成)倒求R/E

Treatment of T.G/L:
Current Rate Method:T.G/L进OCI中的CTA科目

Temporal Method:T.G/L进I/S

3.4Exposure to translation
定性判断汇率对 G/L的影响方向
A-L=E>0,风险在资产A
外币↑导致G
外币↓导致L
m.A-m.L>0时,风险在货币性资产m.A
外币↑导致G
外币↓导致L
m.A-m.L<0时,风险在货币性负债m.L
外币↑导致L
外币↓导致G
3.5 financial ratios under current rate method
3.6 Comparison of ratios under 2 methods

核心:1.明确外币↑↓方向(定顺序),直接标价法下,汇率数值↓表示外币贬值(更少的本币去换外币); 2.明确ratio科目(定构成);3.两法下比较(定大小)
如果是Temporal(用hist rate计量inv),在FIFO下,ending inv是近期买入的,受近期汇率影响大,在I/S中用avg rate就不再精确
外币depr(Hist>Avg>Cur)情况下,FIFO ending inv是近期买入的 数值低 LIFO ending inv是早期买入的 数值高
4. Translation of entity's F/S in a hyperinflationary economy
Hyperinf Economies:Cumulative 3-year inflation > 100% (average 26% per year)

折算方法:
Under US GAAP:直接使用Temporal method
Under IFRS
restated for inflation
B/S items
对于Monetary assets and monetary liabilities,采用current rate
对于Non-monetary assets and non-monetary liabilities,按照inf-adj调整BV,在采用current rate
对于shareholders' equity,按照inf-adj调整BV(Foreign exchange risk),在采用current rate
I/S items
All I/S items are restated (from the dates when the items were originally recorded) to the B/S date
例子 Rev

不论是current rate还是temporal,由汇率变动的net G/L都进I/S
5. Tax Implications of Multinational Operations
法定税率statutory tax rate 通常不等于 有效税率effective tax rate

Effect of disallowed expenses:如罚款的不可抵扣的费用
Effect of exempt income:免税收入
Effect of taxes in foreign jurisdictions:海外子公司按照当地缴税
Effect of recognition of prior losses:往年亏损递减今年税收
6. Revenue Growth Issue and Foreign Exchange Risks
问题:关注因Foreign exchange risk而造成的sales growth(有可能是虚假的盈利)
解决措施:disclosures about the effect of exchange rates on sales growth in the MD&A
LM4 Analysis of Financial Institutions
1. The difference between financial institutions and non-financial institutions
Introduction
serve as intermediaries
Types of financial institutions

Feature
systemic importance. 特别是银行
assets are predominantly financial assets. BV随市场价格波动,敞口大
Financial contagion具有传染性
2. Financial regulations of financial institutions
银行监管Basel III 的目的
Improve ability to absorb shocks
Improve risk management and governance
Strengthen banks’ transparency and disclosures
highlights of Basel III
Minimum capital requirement最小资本充足率
Minimum liquidity 短期流动性(30day以内)
Stable funding 长期流动性(1yr以上)
Other regulatory institution
金融稳定委员会The Financial Stability Board
国际银监会The International Association of Deposit Insurers
国际保监会The International Association of Insurance Supervisors (IAIS)
国际证监会The International Organization of Securities Commissions (IOSCO)
3. CAMELS model
Capital Adequacy 资本充足率
Risk-weighted Assets

核心资本充足率>4.5%,一级资本充足率>6%,资本充足率>8%
Asset Quality 主要是贷款质量
Primary assets of bank
Loans
Creditworthiness借款人的信用程度
Adequacy of adjustments for expected loan losses预期损失
Financial assets

中国的贷款质量监测模型

算例

Management Capabilities 主要是风险管理能力
Earnings Sufficiency
衡量盈利能力的指标
Unbiased accounting estimation
Sustainable earnings.
provide adequate ROC(NI-ROC>0真赚钱)
Fair value are used to measure financial assets

Liquidity Position
Basel III requirements for liquidity 定量指标
现金流出CF(-) one-month liquidity needs in a stress scenario
standards set a target minimum of 100%.
现金流出CF(-) one-yr liquidity needs in a stress scenario
standards set a target minimum of 100%.
定性指标
Concentration of funding融资集中度
Contractual maturity mismatch期限错配:借短 贷长(贷款利率更高,为获利,增加loan,风险↑)
Sensitivity to Market Risk
sensitivity of earnings to different market conditions
hypothetical situation存贷利率相等

In reality 贷款利率>存款利率,引发期限错配

4. Other factors for banks
Other Factors Considered
Government support
Government ownership
Mission of banking entity
Corporate culture
Analytical Considerations
Competitive environment
Off-balance-sheet items
Segment information
Currency exposure
Risk factors
Basel III disclosures
5. Factors considered for insurance company
Revenue
Type
Property and casualty (P&C)财险
products
Property insurance财产险
Casualty insurance意外险
distribution
direct writing直销:自己的员工卖
direct writing是一笔固定成本
agency writing代销:找中介卖
agency writing缴纳佣金即可
Earnings characteristics
Ratios in analyzing
The underwriting expense ratio is an indicator of the efficiency of money spent on obtaining new premiums.
The loss and loss adjustment expense ratio indicates the degree of success an underwriter has achieved in estimating the risks insured
Investment Returns:Conservatively invest保守投资;steady-return稳定回报, low-risk assets低风险;shunning low-liquidity避免低流动性。(决策时流动性因素的优先级高)
Liquidity:high degree of liquidity;low-risk, steady-return types of financial instruments
Capitalization:
没有国际标准
Target min cap adequacy ratio
Life and health (L&H)寿险
products
distribution
direct writing直销:机构对机构B2B
agency writing代销:自己员工卖+找中介卖
Earnings characteristics
Main expenses:
Life insurance: benefit payments
Other types of insurance: contract surrender (early cancellation)
Profitability measures
Return on assets (ROA)
Return on equity (ROE)
Growth and volatility of capital
Book value per share (BVPS)
Investment returns:由于理赔相对可预计relative predictability of claims,因此对liquidity的要求没有P&C那么高,seek the higher returns offered by riskier investments
Liquidity
less important to life insurers因为理赔可预测+long term nature
最近more important to life insurers
Capitalization
没有国际标准
Target min cap adequacy ratio
Analysis focus
LM5 Evaluating Quality of Financial Reports
1. The quality of financial reports
quality of F/S
Financial reporting quality
Earnings (results) quality
Two basic points
basic conceptual framework
合规且决策相关 GAAP-compliant and decision-useful
results (earnings) are high quality (adequate level of return)
GAAP:accounting principles(IFRS,US GAAP)
Decision-useful information:Relevance and faithful representation
High-quality earnings:adequate level of return
Earnings management:利润操纵
Biased choices:Aggressive激进 Conservative保守

Potential Problems
定性分析
Classification Issues
I/S
B/S

标注
C/F
Mergers and Acquisitions Issues
问题
one-time boost may not be sustainable
Acquirer 和 target company各自诉求不同
Accounting choices
解决方案:normalize
Diverges from Economic Reality
问题
Revisions to ongoing estimates
Sudden increases to allowances and reserves
Large accruals for losses
解决方案:normalize
定量分析
财务造假模型:Beneish Model
临界值M-score exceeding –1.78,即认为财务造假

DSR: changes in the relationship between receivables and sales could indicate inappropriate revenue recognition.
GMI: deterioration in margins could predispose companies to manipulate earnings.
AQI: change in the percentage of assets other than in PPE and CA could indicate excessive expenditure capitalization.
SGI: managing the perception of continuing growth and capital needs from actual growth could predispose companies to manipulate sales and earnings.
DEPI: declining depreciation rates could indicate understated depreciation as a means of manipulating earnings.
SGAI: an increase in fixed SGA expenses suggests decreasing administrative and marketing efficiency, which could predispose companies to manipulate earnings.
Accruals: higher accruals can indicate earnings manipulation.
LEVI: increasing leverage could predispose companies to manipulate earnings.
Limitation
只能表示一部分partial representation of economic reality
公司管理者会调整指标以适应模型managers have learned to test the detectability of earnings manipulation
破产Z模型:Altman's Z-score
<1.81 indicated a high prob >3.00 indicated a low prob 1.81 and 3.00 were not clear
Limitation
A single-period static model fails to incorporate the changes over time
uses accounting data not market based data
2. Earnings
unbiased
Sustainable 本节重点
衡量方法
时间序列回归coeff on current earnings
Mean reversion in earnings:extreme earnings should not be expected to continue indefinitely
Earning decomposition: Aggregated Accruals
earning分为cash(优质)和accrual(次级)两类
B/S法
C/F法
Adequate
LM6 Integration of FSA Techniques
3. cash flow
The characteristics of high-quality cash flow
Positive OCF
OCF derived from sustainable sources;
OCF adequate to cover capital expenditures, dividends, and debt repayments.
OCF with relatively low volatility
CFO很难被操控,但是仍有可操作的空间
Timing issues:把未来的CF(+)提前,把现在的CF(-)推后,从而boost CFO
Classification issues:将CFF CFI归入CFO
4. balance sheet
completeness
考虑没进B/S的表外业务 Off-balance-sheet obligations

并表方法的选择会影响指标:Intercorporate investments, profitability ratios under equity method would be higher than under acquisition method
unbiased measurement
clear presentation
5. Sources of information about risk
Financial statements
Auditor’s report
Notes to financial statements
Management Commentary (MD&A)
SEC form NT
Financial press
LM7 Financial Statement Modeling
1. Approaches for Equity Valuation Models
• Revenue Forecast Approaches
• Forecast COGS
• Forecast SG&A
• Forecast Financing Cost
• Forecast Income Tax Expense
2. Forecasting Balance Sheet Items
3. Return on Invested Capital (ROIC)
4. Competitive Position Based on a Porter's Five Forces
5. Forecast Sales and Costs Subject to Price Inflation and Deflation
6. Effects of Technological Development
7. Forecast Horizon
8. Sales-based Pro Forma Company Model
Economics
LM 1 Currency exchange rates: Understanding Equilibrium Value
1. Bid-Offer Spread
Notation
Exchange rate (P/B)

Spot exchange rate
Forward exchange rate
Dealer 报价spread的影响因素
interbank foreign exchange market
The size of the transaction(量太大则dealer风险高,需要补偿)

The relationship between the dealer and client(交易商市场的竞争)
Interbank 报价spread的影响因素
The currency pair involved:主要货币对的流动性好,spread小
The time of day
Market volatility:Geopolitical events (e.g., war, civil strife), market crashes, and major data releases (e.g., US non-farm payrolls)
2. Cross Rate and Triangular Arbitrage
Cross Rate
计算技巧:相乘同边,相除交叉,灵活取倒数
Triangular Arbitrage
例子

3. Forward Discount or Premium
forward points = F - S
基于spot,加上forward point可得Forward rate

4. Mark-to-market value
process
进入反向合约
选择匹配的all-in forward rate
计算cash flow at the settlement day
折现到当前(注意折现率必须为现金流货币的折现率)
算例

5. The International Parity Relationships
长期内汇率的影响因素
Interest Rate Parity 汇率和利率
Covered Interest rate parity
根据情况适时判断

The day count convention
Uncovered Interest Rate Parity
名义利率反映了预期汇率变动
低(高)利率国家,长期会升(贬)值

UIP的前提是有enough risk-neutral investors,因此短期内一般不成立
Forward Rate Parity: UIP&CIP都成立时
unbiased forecast
Purchasing Power Parity 汇率和通胀
题目表述

Absolute PPP:Determined entirely by the ratio of their national price level.
Relative PPP:
Ex-ante version of R.PPP: 通胀高的国家将会贬值
Ex-post version of R.PPP:
International Fisher Effect 利率和通胀
结论
Summary of International Parity Relationships
CIP:名义利率差=forward premium/disct

UIP:名义利率差=汇率变化率

CIP&UIP:forward rate parity

ex ante PPP:汇率变化率=预期通胀差

Int'l Fisher effect:名义利率差=预期通胀差

ex ante PPP and Fisher effects 推出UIP

所有都满足的话:
6. FX Carry Trade
前提:UIP不成立
交易内容:long高利率货币(investing curr.),short低利率货币(funding curr.)
效果
算例

FX Carry Trade的特征
Property:large-scale trades
Return:reward is the gradual accrual of the interest rate differential
Risk:汇率变化突然转向会带来巨大损失
Crash risk: non-normal distribution of returns
7. The impact of Balance-of-Payments Accounts
Current account
The flow supply/demand channel:
出口>进口,本币需求↑,本币升值;之后,本币升值导致进口↑出口↓,本币贬值,消除initial imbalance
影响这种调整恢复balance的因素:initial gap between imports and exports;response of import and export prices to △ER;response of import and export demand to changes in △ER
The portfolio balance channel
The debt sustainability channel:debt is rising to unsustainable level导致本币贬值
Financial account: 短期内,对汇率影响大 Investment/financing decisions are usually the dominant factor
利率高,则国际资本流入,本币升值;利率低,则国际资本流出,本币贬值
8. Exchange Rate Determination Models
间接干预汇率
Mundell-Fleming model
Implicit assumption:短期内,货币供应量变化只影响r,不影响price
各种情况
Flex+HighMob
Flex+LowMob
Fixed+HighMob
The monetary approach
Implicit assumption:短期内,货币供应量变化只影响price,不影响r
Pure monetary modelL货币供应M↑,通胀↑,并导致ER变化
缺点:RelativePPP在长期不一定成立
Dornbusch overshooting model
The asset market (portfolio balance) approach
长期中,debt ratio过高会导致本币贬值
9. Exchange Rate Management: Intervention and Controls
both a blessing and a curse
increase economic growth and asset values
asset price bubbles
Emerging market 在控制汇率方面做的更好
因为FX reserve holdings足够多
10. Warning Signs of A Currency Crisis
liberalized to allow the free flow of capital
large inflows of foreign capital (relative to GDP
banking crisis
fixed or partially fixed exchange rates are more susceptible currency crisis
Foreign exchange reserves tend to decline
the terms of trade often deteriorates before crisis
Broad money growth and the ratio of M2 to bank reserves
Inflation tends to be significantly higher
LM 2 Economic Growth
1. Preconditions for Growth
Factors affecting economic growth

Factors limiting economic growth

2. Importance of Potential GDP
Potential GDP:maximum amount of output an economy can sustainably produce without inducing an increase in the inflation rate
Grinold-Kroner (2002) decomposition of ROE
change in shares outstanding, ΔS
一方面 Net buybacks, or nbb 净回购使得E[R]↑
另一方面 the relative dynamism of the economy, rd 中小企业贡献的增长
Growth in potential GDP:超过会下降,低于会上升。潜在GDP的增长会导致更高的实际利率和更高的资产回报率

Potential GDP Growth and Fixed Income
信用提升improves the general credit quality of fixed income securities.
逆周期调节Monetary policy decisions are affected
经济变差时政府债务增加Government budget deficits
3. Production Function
Cobb-Douglas production function
Properties
diminishing marginal productivity
α≈0说明存在显著的diminishing marg produc
constant returns to scale.
Growth Accounting
growth accounting equation
Labor productivity growth accounting equation
Growth rate in potential GDP = long-term growth rate of labor force + long-term growth rate in labor productivity
Output per worker (labor productivity)
growth in labor productivity
Two sources of growth in per capita output
Capital deepening
MPK=MCK

Technological progress
4. Factors Affecting Economic Growth
1. Natural resources
ownership and production of natural resources is not necessary
“resource curse”.
may fail to develop the economic institutions
suffer the Dutch disease(本币升值 出口竞争力↓)
2. Labor supply factors
Population growth
Labor force participation
Net migration
Average hours worked
Human Capital
3. Capital
ICT (information, computers, and telecommunications equipment)
network externalities, which are largely captured in TFP.
Non-ICT(non-residential, transport equipment, and machinery)
result in capital deepening
Technology
Public Infrastructure
5. Economic Growth Theories
Classical growth theory: 人均产出增长是暂时的
Neoclassical growth theory:增长源于外生的技术进步
Long-term steady state growth rate
eqlbrm:output-to-capital ratio is constant
sustainable 人均产出增长率
sustainable 总产出增长率
Steady State

达到稳态后,capital deepening不再是持续增长的来源
在稳态之下,且Marg product of capital > MC时,capital deepening才有用
Implications
Capital Accumulation:影响level of output 不影响长期增长率
Capital Deepening vs. Technology:技术进步(TFP)才是增长的根源
Convergence:增长达到稳态
Effect of Savings on Growth:储蓄Saving能影响R&D,但不能影响TFP
Endogenous growth theory
不一定达到converge to a steady state rate
no diminishing marginal returns to capital
储蓄Saving能影响R&D,进而影响TFP
R&D and for human capital may have large positive externalities
6. Convergence Debate
Neoclassical才有, Endogenous没给出结论
Absolute convergence
条件:access to the same technology
结果:增速趋同(水平不趋同)convergence of per capita growth rates
Conditional convergence
条件:参数相同same saving rate, population growth rate, and production function
结果:相同人均产出水平y和相同ss增长率
Club convergence:
条件:institutional changes
结果:converging to the income level
发展中国家和发达国家的趋同路径
capital accumulation and capital deepening
adopt technology
7. Growth in an Open Economy
borrow or lend funds in global markets
shift resources into industries in which they have a comparative advantage
have access to a larger, global market
import technology
forcing companies to produce better products, improve productivity
LM 3 Economics of Regulation
1. Economic Rational for Regulation
• Reason for Economic Regulation
1. Informational frictions
Lack of access to information and inadequate information
Asymmetrical information
2. (-tive)Externalities:Systemic risk and financial contagion
3. Weak competition
4. Social objectives
• Regulation of Financial Markets
maintaining the integrity of markets
Disclosures
Prudential supervision
• Regulation of Commerce

• Antitrust Regulation
required for mergers and acquisition
Competition and antitrust laws
2. Regulations and Regulatory Tools
• Classifications of Regulations
Statutes立法机构
Administrative regulations or administrative law行政监管
Judicial law司法造法
• Classifications of Regulators
legislative bodies
government agencies
independent regulators:不是政府机构,但拥有法律和政府授权权威
do not rely on government funding
self-regulatory bodies(SRB)
standard SRB: 属于私有机构private agencies,没有授权,只对会员有约束力
例如CFA Institute
self-regulating organizations (SROs)
被政府机构或机构承认和授权,包括强制执行的权力
funded independently
non-SROs
courts
• Regulatory Interdependencies
Regulatory capture theory
Regulatory competition:不同监管机构争相给出宽松条件
Regulatory arbitrage
• Regulatory Tools
Price mechanisms
Restricting some activities
Mandating some activities
Providing public goods
Financing private projects
Bail-in tools: 2008后规定,确保破产由股东和债权人 承担
3. Cost-Benefit Analysis of Regulation
assess the overall benefits and costs
Regulatory burden
Net regulatory burden
two types of costs.
Intended consequence
Unintended consequence
Unanticipated implementation cost, e.g. hire more lawyer
不是说完全预料不到的,而是can be identified prior
Indirect cost
Analysis of Regulation
Assessment of the likelihood of regulatory change
Assessment of the impact of regulatory change on a sector
Impact on revenues
Cost impact
Business risk
Derivatives
LM1 Forward Commitments
1. Forward

Principle of Arbitrage-free Pricing
No-Arbitrage Principle
Forwards Arbitrage
Cash-and-Carry Arbitrage 持有标的物

Reverse Cash-and-Carry Arbitrage

T-bill (zero-coupon) forwards
forward price (FP)
Forward value (V_long)
Equity Forward and Futures Contracts
Equity Forward Contracts
forward price (FP)

Forward value (V_long)
Ft(T)是指站在t≠0看FP F0(T)是指站在t=0看FP
Equity Index Forward Contracts
复利cont.comp.
复利risk-free rate
复利dividend yield
=div/price 股息率
forward price (FP)

Forward value (V_long)

类比万能公式

Fixed-Income Forward and Futures Contracts
Forward Contracts on Coupon Bonds
forward price (FP)

PVC指的是coupon,属于benefits所以减去
Forward value (V_long)
Interest Rate Forward and Futures Contracts(FRA)
标的资产:FRA的underlying为LIBOR(单利)
作用:对冲利率波动的风险,borrower选择long FRA(担心LIBOR↑,所以支fix收float);lender选择short FRA(担心LIBOR↓,所以支fl收ffix)
settlement:

Forward Contracts on IntRate (FRA)
Pricing

Valuation

标注
第一步:求解FRA,注意去年化
第二步:求解PV支在结算时的价值 (用FRA-settlement rate)去年化为(Loan持续期)/360
第三步:把PV支从结算时,折现到当前(用从t到T的利率),并且去年化为 (T-t)/360
第四步:求解PV收,现金流发生在合约结束、Loan开始那一刻
第五步:V_long = PV收-PV支
2. T-bond Futures

T-bond Futures Contracts
Quoted futures price

S0是quoted bond price,属于clean price
算例1( 不考虑FVCoupon)

3. Swap
Interest Rate Swap Contracts

Pricing a plain vanilla swap
Notation

作用:多笔现金流,相当于一系列FRA

结算:注意fl rate在前一期决定;注意利率去年化

pricing
valuation
老考纲:fix和float计算,为cross curr swap做铺垫
新考纲考法:只对fix swap rate计算(老SR和新SR)
Currency Swap Contracts

Pricing a currency swap
两种货币互换的结构

作用:降低成本(期初交换本金,各期利息现金流不冲销 因为货币不一致)

pricing
但凡涉及fix的货币,一律用IRS的方式求解
valuation
注意求V时,统一货币
注意floating rate在期初即确定
例子:涉及各种情况

例子:“固定-固定”型

Equity Swap Contracts

Pricing an equity swap
pricing
settlement
涉及本金,最后一期c+1
valuation
LM2 Contingent Claims
1. Binomial Model——The Expectations Approach
Put-call parity for Euro: C0+S0=P0+X/(1+r)^T

考点:构造合成;套利策略

synthetic call = put + stock − bond
synthetic put = call + bond − stock
synthetic bond = put + stock − call
synthetic stock = call + bond − put
The one –period binomial stock model
assume that u = 1/d. 保证了stock价格中枢稳定
Expectation approach
风险中性概率
推导:risk-neutral

value of an option

Delta(hedge ratio) 给定opt,用#股票来对冲


算例

The two –period binomial stock model

Euro option:算例

American option:在每个节点权衡是否行权

Interest rate binomial model 利率天然具有均值回归特性

标的是LIBOR用单利
assume that πu=πd=50%. 保证了int均值回归
与FRA的区别
①是option而非commitment,因此可以不行权
②Int option的settlement发生在Loan期末,无需折现 FRA的settlement发生在Loan期初(合约的期末)要折现

Interest rate call option: 约定的X成为最高利率(int cap)

Interest rate put option: 约定的X成为最低利率(int floor)
2. Black-Scholes-Merton Model
本质是Continuous-time option pricing model

underlying assumptions
underlying asset price ~ geometric Brownian motion process
asset price has a lognormal distribution

rf: risk-free rate is known and constant
σ: volatility of the underlying asset is known and constant
markets are frictionless
no cash flows on the underlying asset 可以放松
valued are European options
BSM formulas 定量计算

BSM formulas 定性理解
call: 融资买股票Borrow money to invest in stock
put: 卖股票投资国债Buying the bond(or lend) with the proceeds from short selling the underlying
Options on dividend paying stock or currency
考虑支付div
如果有分红div,会影响S0,d1,d2
考虑投资外币
3. Option Greeks and Implied Volatility
BSM five inputs:S, rf, σ, T, X
Option Greeks
Theta的特例: Euro Put

The Option Delta
=N(d1) ∈ (0,1)
推导

=N(d1)-1 ∈ (-1,0)
其他条件不变的情况下,只要T发生改变,Delta就会随之改变

① Deep OTM: Delta→0
② Deep ITM: Delta→|1| (call是1 put是-1)
当underlying S向着ITM发展时 单个option的hedging效果越明显 (n个S需要至少用n个option)

Dynamic hedging
为实现delta-neutral,对于每股S需要1/Delta份option (对于每份option需要delta股S)
T变则Delta变,当Delta最不稳定时,需要频繁调整仓位,造成higher transaction cost 当ATM时,Delta变化最大,即Gamma=△Delta/△S最大
期权的Gamma是非负的,股票的Gamma是0

算例

role of volatility
Historical volatility

implied volatility:用于计算Vega

4. Binomial Model——The No-arbitrage Approach
5. Black Option Valuation Model
Option On Futures期货期权
定价公式
理解思路:现金流概率折现
算例

Interest Rate Option利率期权
定价公式
理解思路:现金流概率折现
注意:settlement和NP都发生在T时刻,不用折现, 不像FRA那样发生时刻不同需要折现
算例

Swaptions互换期权
Underlying: Forward swap rate (即站在t=1看Loan开始时的利率)
定价公式
算例

Alternative Investments
LM1 Overview of Types of Real Estate Investment
1. Basic Forms of Real Estate Investment
1.1 Basic Forms
按用途划分
按标的划分
equity investor:ownership interest
拥有所有权 property controls
REIT除外, that control is delegated
expect a higher rate of return 因为求偿权靠后
return由两部分组成:income stream + capital appreciation
debt investor:position of lender
Mortgage按揭贷款买房:loan funds to acquire RE
并且用RE作为collateral
求偿权靠前
similar to other fixed-income
按是否公开交易划分
例题

1.2 Characteristics of Real Estate Investment
1.3 Risk factors of Real Estate Investment
1) Property Demand and Supply
2) Valuation
3) Property Operations
<1> Management
<2> Lease provisions:租金条款
<3> Leverage:more leverage for the equity holder
<4> ESG Considerations
<5> Obsolescence:淘汰、过时
<6> Ongoing market disruption
<7> Other risk factors
2. Economic value drivers, role in portfolio, and risk/return of real estate investments relative to stocks and bonds
2.1 Economic Drivers 房价的影响因素
2.2 Role of Real Estate in Portfolio投资房地产的原因
1) Current income
2) Capital appreciation
3) Inflation hedge
4) Diversification:因为与传统投资工具的不相关,可以降低总组合风险
特别注意:private RE比public RE的效果更好(负相关)
5) Tax benefits

2.3 Real Estate Risk and Return 风险和收益特征
Bond-like:债券特征:legal agreements requiring the tenant to make periodic payments
Equity-like:股票特征:tenant续租 或 landlord改租 都不确定,且下一期租金也会改变
风险和收益特征介于两者之间between the risk and return profiles of stocks and bonds
2.4 Classifications
Residential properties 商用住宅地产(美国公寓都是出租的)
Non-Residential properties 商用非住宅地产

2.5 Investment Characteristics by Property Type
core property types(low risk):High-quality, well-leased office, retail, industrial/warehouse, and multifamily
hotel(higher risk):highly correlated with the business cycle
Common Types of Leases
Gross lease:房东承担物业水电费(租金不含物业水电费)
Net lease:租户承担物业水电费(租金含物业水电费)
Expense reimbursement: 房东和租户共同承担费用
Sale-leaseback: 售后回租(本质上是融资行为)
Contractual increases in rents:rent bumps, rent step-ups, or step-up rents每期租金按比例上浮
Long-term leases:根据市场趋势判断
market rents↑ 签订短期有利
market rents↓ 签订长期有利
Investment Characteristics
Office
Demand depends on

Length: varies globally

Industrial and Warehouse
Demand depends on

Length: long-term net leases

Retail
Demand depends on

Length: varies

Percentage lease = fixed$ + 浮动%

Multi-Family
Demand depends on

Length

3. Considerations in Analysis and Due Diligence
Market review
market trends
expected additions to supply and space absorption rates
tenant preferences
Lease and rent review
Compare the tenant rents with market rent forecasts and lease length
Review the lease expiration timeline
history of rental payments, late payments, and any defaults
Review costs of re-leasing
Review costs and incentives
expenses are capitalized and amortized
Review documentation
Review copies of bills for operating expenses
Review multiple years of audited financial statements
Look for evidence of overstated income
Property inspections and service agreements
Conduct an environmental inspection
Conduct a physical/engineering inspection
Review service and maintenance agreements
Conduct a property survey
Legal documentation and tax compliance
Conduct a title search by reviewing the ownership history
Verify that the property is compliant with zoning laws, environmental regulations, parking ratios
Verify that property taxes, insurance, special assessments
4. Indexes
used to measure
1) Appraisal-Based Indexes
Quarterly data 按季度更新数据
Return calculation
优点:无交易时仍可使用
缺点:
2) Transaction-Based Indexes
based on actual transactions 有交易就更新,无交易用上一期价格
repeat sales index 用前几次交易的价格更新
hedonic index 只用上一次交易的价格更新
The regression methodology:基于历史数据,回归得到趋势项,对那些没有交易的房产的价格进行调整
优点:lead appraisal-based index.
缺点:noisy (includes random elements in the observations)
3) Real Estate Security Indexes
本质是股票或股票指数 real estate equity security indexes
REITs and REOCs
LM2 Investments in Real Estate through Private Vehicles
1. Private Real Estate Valuation Approaches

估值
估值的结果是highest implied land value

• Income approach
单期GGM:direct capitalization method
NOI:本质上是CF(真金白银),由Inc减去费用得来(但不减depr.)类似于EBITDA
一旦使用NOI就表明是Gross Lease 因为NOI=Rent-Exp.(房东承担费用)
Stabilized NOI: 某年的NOI出现例外时

先不考虑exception,求出V0
在减去多算部分的现值
对于Gross Lease 计算 Cap rate = discount rate – growth rate
可由 current yield (going-in cap rate)求解
也可通过comparable transactions得来
对于Net Lease来说,叫做All risks yield (ARY)
Gross Income Multiplier:另一个视角看, 把1/CapRate当作具有放大效果的乘数
Value = Gross income * Gross income multiplier
缺点在于没有考虑:Vacancy rates; Operating expenses.
公式
多期GGM:discounted cash flow (DCF)
terminal value
注意是terminal cap rate

算例

• Cost approach
Replacement cost 作为基准
Adjustments 针对各个方面开展调整
支付价格不应高于cost approach求出的价格

• Sales comparison approach (market approach)
Recent property sales
Adjustments
支付价格不应高于similar sales的价格
2. Private Market Real Estate Debt

Max debt 的两个指标
• DSCR偿债覆盖率
DS既包括principal又包括int
• LTV负债率
算例

Equity dividend rate (cash-on-cash return)
• Equity dividend rate

• Leveraged IRR 上杠杆后的收益率
PMT=Inc-Int,FV=Sales-Debt,PV=-自有资金Equity,N=yr

• Unleveraged IRR 纯用自己钱的收益率
PMT=Inc,FV=Sales,PV=-自有资金Equity,N=yr

LM3 Investments in Real Estate Through Publicly Traded Securities
1. Types of Publicly Traded Real Estate Securities
1) Real estate investment trusts (REITs)

Equity REITs: own real estate
2) Real estate operating companies (REOCs)

持有real estate的企业
3) Mortgage-backed securities (MBS)

本质上属于ABS
子主题
2. Advantages and Disadvantages of REITs
3. Net Asset Value Approach
3.1 基本原理:based on asset value estimates, price multiple comparisons, and discounted cash flow(DCF).
1) NAVPS (net asset value per share)


2) BVPS (book value per share)

3.2 Important Considerations
1) Real estate stocks trade at premiums and discounts to NAV.
2) NAV implicitly treats company as an individual asset or static pool of assets静态估值.
3) NAV estimates quite subjective when illiquid and few transactions
3.3 Further Observations: REITs and REOCs should trade at premiums to underlying NAVPS. 原因
(1)流动性好stocks have liquidity on a day-to-day basis
(2)管理能力强attract above-average management teams
4. Relative Value (Price Multiple) Approach
P/FFO & P/AFFO
P/FFO, P/AFFO
Funds from operations (FFO)
类比CFO
Adjusted FFO (AFFO)
类似FCF
Comparisons of FFO and AFFO
AFFO体现分红能力,但variation大
优缺点
Advantages of P/FFO & P/AFFO

Disadvantages of P/FFO & P/AFFO

EBITDA
EV/EBITDA multiples
EBITDA/EV (inverse of the multiple) 可用来逼近 cap rate (NOI/MV)
三个multiples之间差异
1) Expectation for growth in FFO and AFFO
2) Risk associated with the underlying real estate
3) Risks associated with the company’s capital structures and access to capital
5. Private Vs. Public: Comparison
LM4 Private Equity Investments
1. Introduction to PE
Limited partnership

Classification of Private Equity
基金角度估值
2. Valuation Techniques In PE Transactions
Valuation Characteristics: VC vs. Buyout

前四种估值方法(不考计算)

key considerations
value of control,
impact of illiquidity
country risk
Buyout买控股权,VC买少数股权
2.1 Sources of Value Creation in PE
1) Re-engineer the private firm to generate superior returns
2) Financial leverage and the ability to access credit markets
3) Better alignment of interests between private equity firm owners and the managers of the firms they control.
2.2 Balance of Rights and Obligations
1) Corporate board seats 保证PE控制权
2) Noncompete clause 竞业限制
3) Preferred dividends and liquidation preference:
4) Reserved matters
5) Earn-outs(mostly in venture capital)对赌协议
3. Characteristics of Buyout and Venture Capital
4. LBO Model and VC Method
Characteristics of attractive LBO targets
4.1 LBO Model
计算IRR而非value
provides the maximum price that can be paid
input parameters

Value creation
算例

4.2 VC Method
采用增资方式
return on investment (ROI)
与持有时间无关 与revenue有关
算例

算例2

Stage Financing阶段特征:早期risk高
Option Pools:估值时应假设所有opt都行权 on a fully diluted basis
算例

5. Exit Routes: Return Cash to Investors
Type of exit routes
1) Initial Public Offering (IPO)
2) Secondary Market
3) Management Buyout (MBO): 成熟的公司才行
4) Liquidation
Exit timing
LP角度估值
6. Risks And Costs of Investing in Private Equity
Risks of Investing in PE
1) Illiquidity of investments
2) Unquoted investments
3) Competition for attractive investment opportunities
4) Reliance on the management of investee companies
5) Loss of capital
6) Government regulations
7) Taxation risk
8) Valuation of investments
9) Lack of investment capital:
10) Lack of diversification
11) Market risk
Costs of Investing in PE
1) Transaction fees
2) Fund setup costs
3) Administrative cost
4) Audit cost:
5) Management and performance fees
6) Dilution
7) Placement fees:
7. Private Equity Fund Structure and Terms
7.1 Economic terms of a PE fund
1) Management fees represent an annual percentage of committed capital (sometimes calculated on paid-in capital)
2) Transaction fees
3) Carried interest:赚到的钱分成给GP
算例

4) Ratchet对赌协议
5) Hurdle rate

6) Target fund size
7) Vintage year生成时期,同一时间线才可比较
8) Term of the fund
7.2 Corporate Governance terms of a PE fund
1) Key man clause
2) Disclosure and confidentiality
3) Distribution waterfall
waterfall算例

4) Clawback provision 保护LP
5) Tag-along, drag along rights
Tag-along rights随售权:VC收购target时,小股东有权和大股东一起卖
Drag-along rights拖售权:当exit时,大股东有权要求小股东也卖
6) No-fault divorce:75%上的LP同意即可无条件解聘GP
7) Removal for cause
8) Investment restrictions:minimum level of diversification
9) Co-investment

7.3 Due Diligence
exhibit a strong persistence of returns
performance range between funds varies widely
Liquidity in PE is very limited, and thus LPs are locked in for the long term.
7.4 Private Equity Fund Valuation
Net asset value (NAV) = MV(Asset) - MV(Liab)
Ways of asset valuation

Undrawn LP commitments:当初认缴X,现在已经paid in cap有Y,则还剩X-Y未缴纳,不纳入NAV而计为unfunded liabilities
8. Performance Evaluation of PE Fund
算例

8.1 Analysis of IRR since Inception
优劣势
8.2 Analysis of Return Multiples
TVPI=(Cum.Distr+NAV_after)/Cum.Inv.Cap.
优劣势
8.3 Qualitative analysis
LM5 Introduction to Commodities and Commodity Derivatives

子主题
子主题
子主题
子主题
子主题
中心主题
LM1 Exchange-Traded Funds: Mechanics and Applications

1. ETF Mechanics
• Creation/redemption process
• Arbitrage
• Advantages of the ETF mechanics
• Trading and settlement
2. Understanding ETF
• Expense ratio
• Tracking Error
• Trading costs
• Tax issues
• Total Costs of ETF Ownership
• Types of ETF risks
3. ETFs in Portfolio Management
LM2 Using Multifactor Models

1. Arbitrage Pricing Theory (APT)
2. Multifactor models
3. Application
• Return attribution
• Risk attribution
• Portfolio construction
LM3 Measuring and managing market risk

1. Understanding VaR
2. Estimate VaR
3. Extensions of VaR
4. Other key risk measures
5. Applications of risk measures
6. Using constraints in market risk management
LM4 Backtesting and Simulation

1. Objectives of backtesting
2. Backtesting process
3. Problems in a backtest
4. Historical scenario analysis
5. Simulation
6. Sensitivity analysis
LM5 Economics and investment markets

1. Framework for the analysis of financial markets
2. The discount rate on real default-free bonds

3. The discount rate on nominal default free bonds
4. Credit premiums and the business cycle
5. Equities and the equity risk premium
6. Commercial real estate
LM6 Analysis of Active Portfolio Management

1. Value Added
2. Decomposition of value added
3. The Sharpe Ratio & Information Ratio
4. Constructing optimal portfolios
5. Information coefficient & Transfer coefficient
6. The basic fundamental law
7. The full fundamental law of active management
LM7 Trading Costs and Electronic Markets
1. Costs of trading
Effective Cost

VWAP

Implementation Shortfall
2. Advantages of Electronic Trading Systems
3. Market Fragmentation
4. The Major Types of Electronic Traders
5. Low latency
6. Impact of Electronic Trading
7. Risks of Electronic Trading
8. Real-Time Surveillance for Abusive Trading Practices
Corp Finance
LM1 Analysis of Dividends and Share Repurchases

1. Dividends: Forms and Effects on Shareholder Wealth and Issuing Company's Financial Ratios
2. Dividends Policy and Company Value: Theories
3. Factors Affecting Dividend Policy
4. Payout Policies
5. Share Repurchases: Methods and Financial Statement Effects
6. Analysis of Dividend Safety
LM2 ESG Considerations in Investment Analysis

1. Ownership Structures and Their Effects on Corporate Governance
2. Evaluating Corporate Governance Policies and Procedures
3. Identifying ESG-related Risks and Opportunities
4. Evaluating ESG-related Risks and Opportunities
LM3 Cost of Capital: Advanced Topics

1. Cost of Capital Factors
2. Estimating the Cost of Debt
3. The ERP
4. The Cost of Equity
LM4 Corporate Restructuring

1. Corporate Evolution, Actions, and Motivations
2. Evaluating Corporate Restructurings
3. Modeling and Valuation
4. Evaluating Investment Actions
5. Evaluating Divestment Actions
6. Evaluating Restructuring Actions
中心主题
主题
1. Proceedings

2. The Panel
3. Comparisons of AMC and Code and Standards

4. Ethics & Professional Standards
I. Professionalism
I(A) Knowledge of the law;

知法
must&should
applicable
direct governing (无需成为法律专家了解所有法)
守法
more strict
cross-border biz 需要考虑social culture(伊斯兰)
sale-place也要考虑当地法律和情况 (不要轻易移植A国产品到B国)
违法
know确认违法
报告给:supersivor compliance employer
行动上:dissociate(peer不管) 不行就 quit(BOSS不管)
suspense怀疑违法:consult专业人士
一定要对方给出clear的结论
自己也要judge
I(B) Independence and objectivity;
Best Practice: Reject gift that could be expected to compromise独立客观性
modest and normal gift is OK only if 不是为了影响独立客观性
Gift from corporate: 评估对独立客观性的actual effect和in the eyes of clients
Gift from clients: Client's gift should be disclosed, if not = violate I(B)
when possible 在接收前披露
if not possible 在接收后披露
具体场景
Buy-Side clients
Buy-Side Clients may try to pressure sell-side analysts
Portfolio manager threaten or engage in retaliatory practices
Fund manager relationships and Custodial Relationships 选择外部经理人和第三方托管机构
Investment banking relationships IBD和Res之间
Firewall建立防火墙:各自直属领导不交叉;两部门奖金不挂钩(但与公司整体挂钩,就算公司整体50%是IBD也可以);物理隔离(电梯不能经过两部门,咖啡厅不能共用)
仍可以合作:conflicts are adequately and effectively managed and disclosed,需要合规部作为桥梁
定期审阅合规手续:regularly review policies and procedures
Performance Measurement and Attribution 业绩度量与归因
偏离所宣传的策略 stray from their mandate
私自调整组合的成分alter the construction of composite indices
Public companies被上市公司俘获
被要求写有利于该公司的研报
Credit rating agency opinions
被要求给出有利于该公司的评级
Influence during the Manager Selection/Procurement Process选择外部经历和采购
Requesting contributions to a favorite charity or political organization 给老板妻子的慈善协会捐钱来谋取招标目的
Issuer-Paid research付费研报
不能付超过flat fee以外的钱(不能plus XXX),否则是贿赂
研报中要fully disclose potential conflict of interests 以及 compensation(研报费)
Travel Funding
Best practice: always use commercial transportation
不接受paid travel arrangement(除非矿区太远)
不接受reimbursement
Social Activities
审慎参加(大多数情况下不参加),评估对独立客观性的actual effect和in the eyes of clients
I(C) Misrepresentation;

一小三大
typo error:correct更正+及时
Omission
performance:composite construction
model
点名limitation(qualifying); assumption
必须含有negative scenario(可以只有negative)
Untrue statement
吹牛
个人:学历、资质造假
公司:"提供所有服务"、夸大产品
采用的3rd服务 info第三方信息不披露
同时违反V(A)Diligence
core compentence是3rd提供的 不披露
应当选合适的benchmark:if没有,用cash plus + disclose
valuation method:consistency 不能因为用另一种方法更高就换
Reference
Data
权威机构(统计局 Citi HSBC):raw data不用标引用,但二次加工(图表)要引用
其他一律需要引用
Idea
原作有error,修正后呈现在自己的报告里,也要引用
原作只有framework,填充后呈现在自己的报告里,也要引用
单位离职同事的作品,可以直接用、不引用(因为其成果是公司资产)
引用:要具名,不能famous expert
引用媒体上的发言:①找出original出处 或②引用media name + author name
I(D) Misconduct
必须要与 工作相关
dishonest conduct
abbsence of appropriate conduct and sufficient effort 没有尽职尽责、没有充分付出(研报随便写就给出结论)
同时违反V(A)Diligence
conduct that damage trustworthiness or competence negatively affects ability to perform prof. activities
过量饮酒(abuse/introxicated)影响到工作成效(compromised)
drink是中性词
personla bankruptcy 主观恶意导致的破产才违反I(D)且要disclose
贷款上学财务规划不好破产→不违反I(D)
例外:个人信仰+非暴力导致的违法→不违反I(D)
环保主义、种族平等
II. Integrity of Capital Markets
II(A) Material nonpublic information;
Material Nonpublic
Material有2个要求
reliable:insider,relatives,有业务关系的人
competitor's estimation也不行,必须是evidence
clear impact:对股价有明确影响
Nonpublic:只要是内幕就不准交易,哪怕研究验证应该交易也不可以
selective disclosure不叫公开信息 因为知情人的范围被限制
Mosaic Theory
只说用mosaic还不够 必须点明是M&N的info
使用mosaic时必须save document
否则同时违反V(C)Record retention
Social Media
哪怕付费也算是public info(不违规)
Industry expert
获取行业专家的general insights不是违规
获取特定产品研发信息专家的insights属于违规
Investment Research Reports
知名分析师的观点,哪怕是要付费获取,也算public(不违规)
Recommendation
Firewall&Inter-dept. comm.
IBD和research部门之间的沟通要经过compliance officer
Personal trading limitations.
Firm:consider restrictions or prohibitions
Employee:make periodic reports
Proprietary trading procedures 自营业务(两种不违规的例外)
Market maker:无需stop,但要passive to mkt making
Risk-arbitrage trading: best to stop; if not stop 需要证明公司内控完善
II(B) Market manipulation
衡量标准
Info-based: Dissemination of false or misleading
Transaction-based: Members or candidates knew or should have known
三个例外
To increase liquidity: Future Exchg+New contract+Member+Agreement+Disclosure
Tax purposes: selling then buying back
利用市场机会获利(而非主动去干预、扭曲市场来获利): exploit a diff in market power, info, or other inefficiency
III. Duty To Clients
III(A) Loyalty, prudence and care
loyal to client
Fiduciary
fiduciary relationship:投资分析师受托为客户提供投资建议(√),下属为上司服务(×)
trade execution(而非advisory)不算fiduciary
Prudent man rule:higher standard of loyalty and extra care
注意,对client,loyalty要higher
Identifying Client
Individual: investment manager managing the personal assets
Beneficiary: pension plans or trusts 关注的是the ultimate beneficiaries
Trust:委托人父亲,受托人CFA,受益人女儿 如果CFA根据父亲(而不是女儿)来给投资建议 同时违反III(A)&V(A)Diligence-Suitability
Mandate: managing a fund to an index or an expected mandate 关注的是stated mandate而非具体的客户
按照宣称的风格来投资即可
Investing public: investing public as a whole, the goals of independence and objectivity of research
Client's Portfolio
realistic and suitable, appropriate.
站在总组合的角度分析投资 total portfolio
Soft Commission
soft dollar/commission应当100%地用于client,不能给CFA
Manager买Broker的服务,broker为了留客,就给manager好处费(soft commission),但由于买Broker的服务本身是为了客户,因此好处费要用于client
CFA支出费用如果超过normally pay to purchase goods or services,超出部分也应用于服务client
Directed brokerage:客户有权指定manager购买哪家broker的服务
先“best execution”
后“best price”
Proxy Voting Policies
Keep client informed the Policy info
A cost–benefit analysis:CFA权衡后决定有无必要去投票
Recommendation
Regular account information
submit to each client, at least quarterly, an itemized statement
Client approval
CFA ask client if uncertain about the action
client's written approval.
Diversify
应当从总组合来考虑投资:the basic characteristics of the total portfolio
有规定时就不用从总组合出发:比如行业股票plan guidelines or the account objectives.
III(B) Fair dealing
Inv Recommendation
要求同时告知(技术限制是允许的)
发生material changes时(buy变sell),①需要inform all client ②对于受之前评级影响的客户,要particular care
失联也要发(哪怕对方没有阅)
shorten the time frame (例如先brief summary)
limit number 限制信息扩散范围
Inv Action
Allocation 下单交易
block trade 大单交易
policy disclose
fair
错的就是错的:如果unfair且disclose 也属于违反III(B)
same price
Order 客户要求的订单:pro rata on the order size
对于投资经理discretion做投资时, 允许pro rata on acct size
Oversubscribe 超额认购:(优先满足客户的认购额)经理的family acct和immediate acct不允许参加认购
Round-lot和Odd-lot:取决于公司Policy,如要求round-lot则必须按最小交易量调整
附注:family a/c和immediate a/c
亲戚的账户family a/c
normal fee-paying:付费就当做普通客户一视同仁,不付费就视为经理个人账户
beneficiary owner:经理是账户受益人就视为经理个人账户,不收益就视为普通客户一视同仁
例外:pooled account视为一视同仁的普通客户账户
immediate a/c
under one roof(夫妻或室友)的账户应当最后交易
III(C) Suitability
投资经理遵守IPS
Understand Client:包括RRTTLLU
written IPS
必须充分了解才能写IPS,不允许才对话30min就写
不允许用tailor-mode 标准化模板(哪怕是根据模板再改进)
update IPS
至少annually
material change
先改IPS 后交易
comply with IPS:应该站在total portfolio考虑并diversified,除非有特殊要求
Unsolicited trades:是否符合IPS?
符合IPS:执行交易
不符合IPS:判断是否impact IPS?
Yes:问客户Review IPS
客户同意:review&执行交易
客户不同意:经理人要seperate account
No: 执行交易
基金经理遵守Mandate
无需对individual investor负责,但要对stated mandate负责
Recommendation
与客户失联:要记下来自己去联系对方的记录
suitability test policies:首先符合IPS(diversified&fitness),其次符合risk profile,最后符合E[R]
III(D) Perfomance presentation
业绩展示fair, accurate, complete
不能cherry-picking
与I(C)mispre相同
过去的业绩≠保证未来业绩
weighted avg return
terminated account:保留已终止账户的历史业绩
simulated results/back-test results:回测业绩可以用,但要not link to actual results
brief:presentation可以是brief,最好能加reference(基本的数据)
不遵循GIPS但遵循III(D)
遵循GIPS视同遵循III(D)
simulated model:可以用,但与actual区分
prior entity:前东家处做的业绩,要与现东家做的业绩区分
业绩口径:报告gross of fees和net of fees都可以,但不要求报告after tax(因为税收很复杂)
maintating data record
否则同时违反V(C)record
III(E) Preservation of confidentiality
对于 former current prospective 都要遵守
3种披露情景
The information concerns illegal activities
Disclosure is required by law
对于“即使违法也不能披露”的法律,就不能披露,如瑞士银行
client permits disclosure of the information.
Electronic Information and Security
不能在社交媒体上泄露
CFA Institute调查时的后门条款
PCP来检查时,如果法律不禁止,可以披露给PCP用于investigation
employees who are also working for the client
A和B都同时服务于张三,那么AB之间可以共享张三信息,不用保密
IV. Duty To Employers
IV(A) Loyalty
Independent practice 竞争性行为
在职场景
上班:comply with policy
下班:
副业或兴趣:时间、经历、业务内容 不能与工作冲突
if 冲突,必须有written consent才能做 (题目只说“申请了”但没说书面同意,就违规)
开公司(同行业):允许做准备工作(盖章/租场地/下班约现同事加入),不允许solicit client
Leaving the employer
离职场景
离职前:get permission才能带走
client list:
only name不算做client list,必须是client+contact info
记住a few 是允许的,但记住2000个就算刻意背诵(不允许)
研报/model
带走前必须有written consent from employer(而非supervisor)
跳槽到新公司,model without record, 想要record:①permission ②public source补全
V(C)record: 使用model(参数)时 必须配合record(数据)
前东家要求 model√ record×,CFA私自copy record带走 那么同时违反IV(A)loyal和V(C)record
performance history
跳槽到新公司,figure without record, 想要record:①permission ②public source补全
V(C)record: 使用figure(业绩)时 必须配合record(交易记录)
not allowed to solicit client
离职后:想solicit former client
是否签署Non-compete agreement(NCA)?是否内含有禁止solicit条款?
在A国签署NCA后去B国工作,B国视NCA为illegal 如果CFA去solicit,仍然违反
获取contact info:①with permission ②public source
Use of Social Media
Best practice:工作账号与私人账号分开
if混用公私账号,应该discuss with employer
Whistleblowing
动机必须是:①cap mkt integrity ②interest of client
不能用于个人争端personal issue(哪怕是发现别人行为不端,害怕别人报复而去举报)
Nature of Employment
只有employee 才受制于前述规定
雇佣关系:full-time,part-time,paid-intern,unpaid-intern
非雇佣关系independent contractor:通过oral或written方式来表明身份
IV(B) Additional compensation agreement
Def of Add.Compensation
条件:①基于本职工作所获得的 ②来自non-employer的
除非有written consent from all parties,否则不能接受那些产生conflicts of interest的好处
除了商务宴请
written report包括:①the nature ②amount ③duration of compensation
IV(C) Responsibility of supervisors
可以delegate事权, 但责任仍在supervisor
Use of Social Media
Best practice:工作账号与私人账号分开
if混用公私账号,应该discuss with employer
compliance procedure
必须有compliance procedure,如果没有
如果已经是supervisor,违反了IV(C)
如果刚刚promote(未上任),必须decline in written until completed
制定appropriate(根据size/complexity)不允许standard version
written compliance system & periodic review
detection system
发现potential violation
initial investigation(而非lead)
place limit(限制员工行为)
training all staff
reward system
financial
non-financial
V. Investment
V(A) Diligence and reasonable basis
Defining 尽职尽责:自上而下地分析
Using 2nd or 3rd Research
Criteria
选取第三方来协助时,需要determine soundness & adequate due diligence
选择自己的朋友作为第三方(不允许)
对于internet source,需要greater level of review
timely and consistent review第三方
Using Quantitatively Oriented Research
必须understand :parameter,assumptions,limitation
不能一只半价、know,而要understand
Developing Quantitatively Oriented Research
understand the technical aspects
将module汇总后要整体检验
使用样本外数据检验outside the observable databases
结果必须包括negative scenario
I(C)mispre同样要求含有悲观情景
Selecting External Advisers and Subadvisers
必须都满足,不能拉一个list 根据业绩看一下就得出选哪家
第四条“言行一致”,不能只看公司stated有compliance就选用
第三条,要评估其业绩的quality
Group Research and Decision Making
可以不认同小组研报结论,但如果认同分析过程就可以署名
在自媒体上发自己不认同小组结论→违规
不能将hot issue作为分析的起点。但可以分析一堆股票后,得出hot issue
V(B) Communication with clients
总纲
披露投资的basic format and general principles,以及material change
重大改变包括:key person,model,strategy&mandate,limitation,investment process
基于reasonable judgement,给出significant limitation and risk
区分开fact 和 opinion
will be属于fact
Informing Clients of the Investment Process
changes to the investment process
basic characteristics of an investment
the impact each investment
expertise of external advisers
Different Forms of Communication
采用new technologies去通知时,需要treat all clients fairly
采用in capsule form简报去通知时,client request就要回应
III(B)fair dealing 同样是brief+detail
Identifying Risk and Limitations of Analysis
significant risks and limitations
Risks
the use of leverage
general market-related risks
use of complex financial instruments
counterparty risk, country risk, sector or industry risk, security-specific risk, and credit risk.
limiting fators
Liquidity
Capacity:超过100w策略就失效了
Report Presentation
必须包括analysis elements和conclusion(只有conclusion不允许)
分析内容可以这一部分详细,另一部分简略,只要stipulate the limits
对于quant analysis 需要提供reference material
Separate opinions from facts
statistical conjecture≠事实
V(C) Record retention
Records Are Property of the Firm
包括hard copy or electronic form
想拿走必须要有consent
要使用record:①permission ②public source
Local Requirements 保存时间
当地如有法律要求,遵循法律优先
如果没有法律要求,按照CFA要求 保存7年
VI. Conflicts of Interest
VI(A) Disclosure of conflicts
full&fair disclosure with plain language
分析师inherit分析对象公司的股票
disclose就可以继续分析
best practice:assign another analyst
Disclosure to Employers
披露个人全部持仓
①personal trading
②outside board membership
Disclosure to Clients
披露客户的投资标的
corporate financing or market making relationship;
Security holding:包括stock、option,但不包括large diversified mutual fund
Directorship:自己上司与对方公司管理层的交叉任职
Individual relationship
Disclosure of fee
只要是fee(包括non-standard fee)都要披露给科客户
Cross-Departmental Conflicts:IBD与research隔离
同时违反I(B)indpd&obj
之前投资VC的标的上市,需要向客户披露
incentive fee
例如:公司采用long-term strategy,业绩采用short-term compensation 内在冲突影响客户利益,需要disclose to client,如果employer不允许→需要quit job
VI(A) Priority of transaction
Priority
Client>employer>individual
如果没有提及employer交易,可以client>individual,相隔十天半个月
>:相隔十天半个月,不能是几分钟
Personal Trading
beneficiary owner:如果是一视同仁型账户,属于client账户
管理父母的账户(fee-paying),为了避嫌,其他客户交易完有剩才交易 violate VI(B)priority 因为没有一视同仁
aligned interests
liquid好(blue chips,国债)→可以同时
liquid差(PE, small cap)→client优先
Recommendation
Limited participation in equity IPOs
Restrictions on private placement(PE)
Establish blackout/restricted periods
Disclosures of policies
VI(C) Referral fees
disclosure
推荐方和接收方都要告知客户
nature of compensation
事先告知
Recommendation
Firm may completely restrict such fees
provide the employer regular (at least quarterly) updates
其他regular update 1.Quarterly: statement of transaction 2.Annually: IPS
VII. Responsibility as Members
VII(A) Conduct as members and candidates
cheating
exam information
可以表达观点(难易)
泄露题目范围:考了什么、没考什么
以公谋私
Improperly using an association with CFA Institute
VII(B) Reference to CFA institute, designation
CFA citation
书写格式
adj, not n.
Always capitalized.
CFA level I
C.F.A
Don’t alter to create new words or phrases.
Mustn’t be used as part of the name of a firm.
Shouldn’t be given more prominence (e.g. larger, bold) than the charter holder’ name.
不允许expected date of exam completion
Level III passed ≠ CFA member
引用头衔
CV中:CFA, 2001, CFA Institute (optional Charlottesville, Virginia, USA)
business card、letterhead
顺序无所谓 CPA, PhD, CFA
CFA Institute membership
①annually PCS
②membership fee annually
Referring to candidacy
candidacy:从registration到收到exam results
statement:只能跟自己比,不能跟别人比("远胜于他人"&"competency and future investment results")