Scribd
Upload
44 views

Risk Aversion - Lecture Notes

Uploaded by

quanduchung
Copyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
44 views

Risk Aversion - Lecture Notes

Uploaded by

quanduchung
Copyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 20
Microeconomic Theory Risk Aversion 1 Introduction We will now consider lotteries that have monetary outcomes. We will for- mally represent lotteries in terms of a probability density/mass function (or distribution function) on the set of monetary outcomes. Risk is defined as the variability in a monetary outcome. A degenerate lottery that gives a sure return does not involve any risk. A non-degenerate lottery on the other hand involves risk since the monetary outcomes will vary depending ‘on the state that is realized. In this chapter we will analyze the behavior towards risk of a consumer with expected utility preferences. Most individ- uals exhibit an antipathy to risk and we will be concerned primarily with the characterization of this risk aversion. We will sce that risk aversion is captured by the concavity (a second order property) of the Bernoulli utility fimetion on the set of outcomes. We will quantify risk aversion with the help of the Arrow-Pratt measure. The Arrow-Pratt measure is use- ful in examining how risk aversion varies with wealth as well as in making interpersonal comparisons of risk aversion, 2 Expected Returns and Expected Utility A lottery is formally represented by a distribution fumetion, F, (or equiva- lently a probability density/mass function, f) on monetary outcomes. Con- sider first the diserete case where the monetary ontcomes, represented by a random variable X, are drawn from a finite set X € {e1,22,..,0N)} Note that the uppercase X denotes the random variable and the lowercase x denotes a realization or outcome of this random variable. Let. f denote a probability mass function which associates probability f(x.) to the outcome x; such that f(x;) > 0 and ON, f(e:) = 1. Let F denote the associated distribution function. Then the expected return from this lottery is: EelX| = So asf(2) Q i From now on we will just write J> instead of >, in order to keep the notation simple. When the individual has expected utility preferences, then the expected utility from F is: uP) =Dueose) ®) where w is the Bernoulli utility function over monetary outcomes and U is the von Neumann-Morgenstern utility function over lotteries (or distribution functions). Now consider the continuous case where the monetary outcomes are drawn from the set (a, 8], -00 0 for z € fa, 8], f(x) = 0 otherwise, and f2 F(v)de = 1. Let F denote the associated distribution function, where 4° F". The expected return from the lottery F is BelX|= fasterde= f° 2Peite @) ‘The expected utility from this lottery is: , u(r) = [ ula) f(a)ae = [ ula) F(x)de @ We will assume throughout that the Bernoulli utility fumetion, w, is an in creasing function, i.e. u! > 0, implying that marginal utility from monetary returns is strictly positive. 3 Attitude Towards Risk An individual’s attitude towards risk can be examined by considering the individual's preference between a lottery F, and a degenerate lottery giving the expected value of F. Definition 1 An individual is risk averse if, when faced with a choice between a lottery F and a degenerate lottery which offers the expected value of P, the individual always prefers the latter. Given expected utility preferences, aan individual is risk averse if and only if for any lottery F: Lues@) 0 and an outcome y with probability 1 — p, where x < y. Then the expected return from this lottery is: Bp(X] = pe + (1 phy a The utility from a degenerate lottery that gives By /X) for sure is: u(Ep(X]) = u(pe+ (1 py) 8) The expected utility from the lottery F is: U(F) = pu(x) + (0 = p)u(y) (9) The utility from the expected return for sure utility from the lottery in Figure 1 for a conea that the chord connecting points u(x) and u(y) on the utility function w lies below u, Therefore, the utility from the expected return from sure exceeds the expected utility from the lottery when w is concave compared to the expected e utility function u. Note uw) ust Lply u(x) + (-p)uy) x px (L-ply y Figure 1: Concave Utility and Risk Aversion While most economic applications concern themselves with risk averse eco- nomic agents, in some cases they also consider risk neutral and risk loving economic agents. These can now be defined for the case of expected utility preferences. We define it for the continuous case since the discrete case is similar: Definition 3. An individual is risk neutral if and only if for any lottery P: 5 (f sf(elde) ‘The above equality holds only for linear real-valued functions. ‘Therefore, an individual with expected utility preferences is risk neutral if and only if the individual's Bernoulli utility function is linear on the set of monetary outcomes u(Er[X]) (20) 2 f ule) f(e)ae Definition 4 An individual is risk loving if and only if for any lottery F. [sorters Su (see) The above inequality holds only for a convex real-valued function. There- fore, an individual with expected utility preferences is tisk loving if and only if the individual's Bernoulli utility function is convex on the set of monetary outcomes. u(Be[X]) ay 4 Equivalent Formulation of Risk Aversion We now discuss two equivalent formulations of risk aversion, 4.1 Certainty Equivalent Given a lottery F, the certainty equivalent of P, denoted by e(F), is the monetary amount that will make an individual indifferent. between the lottery F and the certain amount e(/). Mathematically, in the discrete and continuous cases respectively: Le@s@) = uler)) 4 [weoneyae = weer) (2) To illustrate the certainty equivalent, consider once again the lottery, F, given by (copsyo (1 —p)). The certainty equivalent of this lottery is given by u(e(F)) = pu(z) + (1 — p)u(y) (13) ‘We can now locate the certainty equivalent in the figure below by determin- ing the sure amount that gives the same utility as pu(x) + (1 — p)u(y). uw) Putx) + (I-p)uty) x o(F) px (ply. y Figure 2: Certainty Equivalent and Risk Aversion Note that. the certainty equivalent is less than the expected return from the lottery. This is an important characteristic of risk aversion; a risk-averse individual is willing to substitute the lottery with a sure return that is smaller than the expected return from the lottery. In other words, aversion to risk means that an individual is willing to sacrifice the difference, [pc + (1 ~ p)y] — e(F), to replace the risky lottery with a certain return. The following proposition presents this characterization formally in the general case Proposition 2 An individual is risk averse if and only if the certainty equivalent of any lottery F is less than or equal to the expected return from P, that is: c(F) < Ep[X] { Exile), — X discrete fExfla)de, X continuous (4) Proof: Consider the continuous case. (The proof is identical in the discrete case). Suppose an individual is risk-averse. ‘Then the utility fimction u is concave. It now follows that: 3 u(e(F)) = [ u(e)f(z)dr from the definition of e(#) A u ( f *e (aide) from Jensen's Inequality, since u is concave Since u is an increasing funetion (i.e. u! > 0), we can conclude that e(F) < [Pep (w)de = Bp(X). The proof of the converse is similar. Suppose o(F) < [.xf(e)de. Then: f[ " ula)sleyde = ulelF)) from the definition of e(F) 8 < o( [ sf(e)de) since w is increasing yielding f° wu aversion. \ie)de 0. Conversely if a(F) > 0, then e(F) < [2 cf(x)dz and it follows from Proposition 2 that the individual is risk averse. We can now gather all the above results into one statement: Proposition 4 The following statements are equivalent: (a) An individual is risk averse. (b) The Bernoulli utility function of the individual is concave on the set of monetary outcomes. (c) For any lottery F, the certainty equivalent e(F) < p(X] (A) For any lottery F, the risk premium x(F) > 0. ‘The same arguments can also be extended to the risk-neutral and risk-loving cases to obtain the following characterization results. Proposition 5 The following statements are equivalent (a) An individual is risk neutral (risk loving). (b) The Bernoulli utility function of the individual is linear (convea) on the set of monetary of monetary outcomes. (c) For any lottery P, the certainty equivalent (F) — Ep(X] (c(F) > Bp|X)). (A) For any lottery F, the risk premium m(F) = 0 (x(F) < 0). 5 Absolute Measure of Risk Aversion In many economic applications, we would like to have a measure of the consumer's risk aversion. We can then meaningfully investigate the change in risk aversion due to a change in some parameter such as wealth, w. Since risk aversion is equivalent to the concavity of the Bernoulli utility function, a natural candidate to measure the degree of risk aversion at w seems to be the curvature of the utility function as measured by u”(w). However, the second derivative by itself is not a valid measure of risk aversion. Recall that a consumer's expected utility preferences remain the same under any linear transformation of utility, Therefore, the Bernoulli utility function, v(w) = au(w) +b, a> 0, represents the same preferences between any two lotteries as u. But: vi(w) = au'(w) v"(w) = au"(w) Ifa #1, then v"(w) ¢ u"(w). This contradicts the fact that v should exhibit the same degree of risk aversion at w as u. In order to have a quantitative measure of the degree of risk aversion which is invariant to affine transformations of the Bernoulli utility function, we define such a measure at wealth w as follows: u"w) wl (17) It is called the Arrow-Pratt measure of absolute risk aversion. It is easy to verify that all Bernoulli utility functions which are affine transfor- mations of each other will have the same Arrow-Pratt measure of absolute ra(w) = risk aversion. For example, for v(w) = au(w) +, a> 0: rw) ww Vw) aww) Wey g ra(w) = — We now need to provide some justification for interpreting the expression given by (17) as quantifying the aversion to risk, We know that a risk averse individual is willing to pay a positive amount (the risk premium) to avoid a lottery and replace it with a sure return, We will now show that the Arrow Pratt measure is directly related to the risk premium that a risk averse individual is willing to pay to avoid a small but actuarially fair (zero expected return) lottery. Consider an individual with wealth w. The individual faces a lottery F that is “small” in the sense that the different monetary values « that are realized under this lottery, positive or negative, are small in magnitude, Therefore lottery F causes small fiuetuations in the individual's wealth around the given level w (equivalently, the lottery F involves only a small risk). The lottery is actuarially fair in the sense that its expected value is equal to zero: EplX] f xF'(x)dr = 0 (as) Further, the lottery has a constant variance equal to o?, Recall that the variance of a random variable with distribution function Fis defined as: Ep [X — Ep(X]? (19) Since Ep[X] = 0, it follows that the variance is Ep[X?] ‘The risk premium associated with F, x(F), is the amount that an individual is willing to pay out of current wealth to substitute the lottery for a sure return. In other words, it is the amount that makes the individual indifferent between the sure amount w —7(F) and the lottery F. ‘Therefore: u(w —x(F)) = Ep(u(w + X)] (20) We will expand both sides of the above equation using a Taylor series approx- imation.! Since the risk premium is a constant amount that will be small TS where the restriction that the risk should be “suuall” comes in, ‘The Taylor series approximation is only valid in a small neighborhood of the wealth level w. 10 in magnitude (because the risk involved is small), terms such as (x(F))?, (x(F))? etc, will be very small. Therefore we can ignore terms that are of second order or higher when we take a Taylor series expansion of the left hand side: uw —x(F)) = u(w) — (PF) (w) (21) We take a second order approximation of the right hand side and ignore terms that are of third order or higher since they will be extremely small in magnitude (given that the realizations of X are small): Epluw+X)} = Be |u(w) + Xu )+ hxral(w) uw) + ww) BEX] 4 Suto) BelX?] ue(w) + 5 (22) Equating (21) and (22): (23) Therefore, the Arrow-Pratt measure of absolute risk aversion is directly proportional to the risk premium a risk averse consumer is willing to pay to avoid a small but actuarially fair lottery. In particular, the Arrow-Pratt measure is equal to twice the risk premium the consumer is willing to pay to avoid a unit of variance for very small risks. ‘The Arrow-Pratt measure is a local measure of risk aversion (or, quantifies risk aversion “in the small” in the terminology of Pratt 1964) because it is directly related to the amount that a risk averse individual is willing to forego to replace very small risks with @ sure return, Putting it differently, it is a local measure because it, quantifies risk aversion only at a given level of wealth, The measurement of risk aversion could be different at other wealth levels because w(w) and w"(w) would typically vary with wealth. rt 6 Risk Aversion and Wealth An interesting question is whether an individual's risk aversion increases or decreases with the level of wealth. We can classify utility functions in terms of constant, increasing or decreasing absolute risk aversion over some subset of wealth if r4(w) remains constant, increases or decreases respectively with wealth over this subset. ‘Constant absolute risk aversion (CARA) Tncreasing absolute risk aversion (TARA) | 24050 Decreasing absolute risk aversion (DARA) | “#27 <0 Example 1: The Bernoulli utility function: u(w) =e 4b, a0 displays constant absolute risk aversion (CARA). The utility function: u(w) = a+ bw — cw, b aber cme displays increasing absolute risk aversion (IARA). The utility function: displays decreasing absolute risk aversion (DARA). Of the three possibilities, DARA is the most realistic restriction on the preferences of a risk averse individual. It is consistent with the observation that individuals are generally less averse to risk at higher svels of wealth, 7 Comparing Risk Aversion Across Individuals The Arrow-Pratt measure can also be used to compare risk aversion across individuals. Consider individuals 1 and 2 with Bernoulli utility functions 1 and up over monetary outcomes respectively. In many economic applica- tions, we are interested in comparing the optimal decisions of two individuals where one, say individual 1, is always more risk averse than individual 2 We, therefore, need a definition of what it means for individual 1 to be al- ways more risk-averse than individual 2. Given different: Bernoulli utility 2 functions for the two individuals, the most obvious way is to compare the Arrow-Pratt measure of absolute risk aversion for the two individuals at the same level of wealth: Definition 5 Given two individuals, 1 and 2, with strictly increasing and concave Bernoulli utility functions u; and uz respectively, individual 1 is more risk averse than individual 2 if and only if for every w: 2 = r'(w) (24) Using the above definition, we can obtain other equivalent characterization results for comparing risk aversion across individuals. Note that since both uy and uz are strictly increasing functions, we can find some strictly increas- ing real-valued function such that: = d{ur(w)) (25) m(w) Differentiating (25) yields: uj (w) = 6'(ua(w))uea(n) Differentiating the above a second time yields: 8"(waw)) [ur(w)] Dividing the above expression by uf, and simplifying yields: 8" (ua(w)) ‘9 Fata) | ©) 2 uf (w) = 6" (ualw) uw) Face) = A) — 9 [ It follows from the above expression that r}(w) > r4(w) is equivalent to 6" <0, ic. ¢ is concave. This gives us the following characterization result: individual 1 is more risk averse than individual 2 if and only if the Bernoulli utility function of individual 1 is more concave than that of individual 2 in the sense that u is a concave transformation of uz Recall that the Arrow-Pratt measure is a local measure of risk aversion, ic. it applies to small risks (lotteries that cause small fluctuations around the current level of wealth). Therefore Definition 5 compares the risk aversion of two individuals for small risks. However, in many applications we would 13 like to make global rather than local comparisons of risk aversion. In other words, we would like to compare the risk aversion of two individuals for large risks as well (i.e. for any arbitrary lottery including those that involve large fluctuations around the current level of wealth). It turns out that even though the Arrow-Pratt measure is @ local measure, it allows us to compare the risk aversion of two individuals for any lottery. This is made precise in the following result: Proposition 6 Consider two individuals, 1 and 2, with strictly increasing and concave Bernoulli utility functions u, and uz respectively. ‘Then the following statements are equivalent: (a) For every wealth level w (b) us is @ concave transformation of uz (c) For every lottery F, the certainty equivalent of individual 1, ex(F), is less than or equal to the certainty equivalent of individual 2, c2(F). (A) For every lottery F, the risk premiurn of individual 1, m(F), is greater than or equal to the risk premium of individual 2, 72(F) Let us examine Proposition 6 before proving it. Part (a) of the statement makes a local comparison of the risk aversion of two individuals. It states that individual 1 is more risk averse than individual 2 when both are subject to small (and actuarially fair) risk, ie. small fluctuations around the current wealth level, The other three statements make global comparisons because they are not restricted to small risks. Part (b) is a global comparison because it states that individual 1's utility function is more concave than that of individual 2 at any wealth level and thus applies to large risk as well that cause large fluctuations in wealth. Part (ec) is a global comparison of risk aversion because it considers arty lottery F and is therefore applicable for all risk, It states that a more risk-averse individual would always be willing to accept a smaller amount to substitute the risk (whether small or large) with a sure return, Similarly part (d) is a global comparison because it shows that a more risk averse individual is always willing to pay a greater rT risk premium for both small and large risks. ‘The importance of Proposition 6 is that it shows how a measure “in the small” allows making inferences about risk aversion “in the large” or global scale. Proof: To prove the proposition, we have to show that (a) = (8) + () > (d) = (a). We have already shown above that (a) = (b). We will now show that (b) = (c). Consider any lottery F and note from the definition of certainty equivalent that: ‘ m(a(F)) = fot F"(w)dw 3 sale) =f value Using our result that u; is a concave transformation of uy tm (e(F)) 2 f u(w)F"(w)dw definition of ex(F) ye o(ua(w))F’(w)dw since wy = e(u2) A 3 ° ( [ s(u) (ude) [ wF"(w)dw — oF) = m(F) (27) Therefore, for the same lottery F, the relatively more risk-averse consumer has a greater risk premium, 15 Finally we have to prove that (d) => (a). Since (27) holds for all lotteries, it also holds for lotteries that involve a small and actuarially fair risk with variance a? around wealth w. Tt now follows from (23) that: a, 2m(F) , 2m(F) ri (w) proving the result. i 8 Relative Measure of Risk Aversion The Arrow-Pratt measure of absolute risk aversion considers those lotter ies where the uncertain monetary outcomes are independent of the current wealth of the consumer. In other words, the random variable refers to ab solute changes to the current level of wealth. If the current wealth is wo, then wealth changes according to wy +6 where 6 is # random variable with distribution funetion F. However, there are many lotteries whose uncertain monetary outcomes are gains or losses as @ percentage or proportion of current wealth. In other words, if the current wealth is to, then wealth changes according to éw» where é is a random variable with distribution function F. To measure risk-aversion to such lotteries that affect wealth proportionately requires an alternative measure referred to as the Arrow-Pratt measure of relative risk aversion: rp(w) = = wra(w) (28) Analogous to the absolute risk aversion case, we have the following charac- terization of relative risk aversion: Constant relative risk aversion (CRRA) | 287 =0 Increasing relative risk aversion (IRRA) | [327 30 ‘Decreasing relative risk aversion (DRRA) | 7 <0 Empirically we observe DARA and CRRA in the real world. Example 2: Consider the following Bernoulli power utility funetion: uw af S asl afo 4 togw, a=0 16 ‘The above fimetion displays decreasing absolute risk aversion but constant relative risk aversion. il Once again we can justify this measure of relative risk aversion in the fol- lowing way. Consider @ lottery which causes small proportionate changes in wealth equal to éw, where 6 has distribution function F. Tt is assumed that the lottery is fair, ie. expected value of Ep{6] is equal to zero. Further, the variance of this lottery, Ep[6"] = 0?. Define the proportional risk premium of F, x°(F), as the maximum proportion of wealth w that the consumer is willing to forego in order to substitute the lottery F for a sure return. Then: u(w — 7°(F)w) = Ep[u(w + 6w)] (29) Expand both sides of the above equation using a Taylor series approxima tion. Since the proportional risk premium is constant and small in magni- tude, we take a first order approximation of the left hand side: u(w—1?(F)w) = ul = wa (P)u'(w) Taking a second order approximation of the right hand side: Ep(u(w + 6w)) = Bp |u(w) + dwu!(w) + 3euul(w) Noting that Ep[6] =0 and Ep(6*) , equating the two sides we get: (30) ‘Therefore, analogous to the case of absolute risk aversion, the Arrow-Pratt measure of relative risk aversion is directly related to the proportional risk premium a risk averse consumer is willing to pay to avoid a small but fair lottery. References [1] JAV. Pratt (1964) “Risk Aversion in the Small and in the Large”, Econo- metrica 32, 122-136. 7

You might also like