投資人風險趨避的變化與金融海嘯: 行為觀點

全文

(1)國立臺灣師範大學管理學院管理研究所碩士論文 Graduate Institute of Management College of Management National Taiwan Normal University Master Thesis. 投資人風險趨避的變化與金融海嘯: 行為觀點 Change of Individual Risk Aversion and Financial Crisis: A Behavioral Perspective. 陳冠宇 Kuan-Yu Chen. 指導教授：賴慧文博士 Advisor：Christine W. Lai Ph.D.. 中華民國 105 年 2 月 February, 2016.

(2) Change of Individual Risk Aversion and Financial Crisis: A Behavioral Perspective. Abstract The purpose of this study is to investigate the change of risk aversion of household investors in the pre-and post- sub-prime financial crisis periods from a behavioral perspective. Using the 2007 and 2009 Survey of Consumer Finances (SCF) panel data, conducted by the Board of Governors of the Federal Reserve System, we examine whether the experience effects, including “snake-bit” effect and “house-money” effect, and mental accounting will influence the change of individual relative risk aversion in the pre-and post- sub-prime financial crisis periods. The “snake-bit” effect suggests that after experiencing a financial loss, individual investors become less willing to take risk. On the other hand, the “house-money” effect suggests that investors are willing to take more risk once they have experienced a gain or profit. Using the following variables to proxy for bad experience-capital losses, job losing, no income, bad investment, bankruptcy, declining in stock price, declining in real asset/other asset values, we find that bad experience during the financial crisis period will positively affect individual relative risk aversion index (RRAI, i.e., higher RRAI indicates more risk aversion) in the post-crisis period. On the other hand, using the following variables to proxy for good experience-getting jobs, an increase in income, and good investment, we find that good experience will negatively affect individual RRAI. These findings provide evidence to support that experience effects will influence change of risk aversion of household investors. In addition, this study investigates whether mental accounts are associated with varying levels of risk aversion. We find that compared to regular accounts, retirement accounts show less positive change in RRAI in the post- sub-prime financial crisis period, indicating that household investors express less risk aversion in retirement account than in regular accounts after the sub-prime financial crisis in 2009.. Keywords: risk aversion, financial crisis, investment behavior. i.

(3) CONTENTS 1.. INTRODUCTION ............................................................................................................... 1. 2.. LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT ................................................... 3 2.1 RELATIVE RISK AVERSION INDEX ..................................................................................................... 3 2.2 EXPERIENCE EFFECT ..................................................................................................................... 4 2.2.1 Snake-bit effect .............................................................................................................. 4 2.2.2 House money effect ........................................................................................................ 5 2.3 MENTAL ACCOUNTING ................................................................................................................. 5. 3.. DATA, VARIABLES AND MODELS ....................................................................................... 6 3.1 DATA ........................................................................................................................................ 6 3.2 VARIABLES- DEFINITION OF RELATIVE RISK AVERSION INDEX ................................................................ 7 3.3 MODELS.................................................................................................................................. 10 3.3.1 Experience Effect .......................................................................................................... 10 3.3.1.1 Snake-bit Effect ..................................................................................................................... 10 3.3.1.2 House Money Effect ............................................................................................................. 11. 3.3.2 Retirement and regular accounts: ................................................................................ 13 4.. RESULT ........................................................................................................................... 16. 5.. ROBUSTNESS ANALYSIS .................................................................................................. 20. 6.. CONCLUSION ................................................................................................................. 21. 7.. REFERENCE .................................................................................................................... 22. I.

(4) Tables TABLE 1 SUMMARY STATISTICS OF ASSETS ................................................................................................ 23 TABLE 2 STATISTICS OF RELATIVE RISK AVERSION INDEX (RRAI) ................................................................... 24 TABLE 3 SUMMARY STATISTIC OF MAIN VARIABLES ..................................................................................... 25 TABLE 4 REGRESSION RESULTS: SNAKE-BIT EFFECT AND HOUSE MONEY EFFECT.............................................. 26 TABLE 5 REGRESSION RESULTS: MENTAL ACCOUNTING ............................................................................... 29 TABLE 6 REGRESSION RESULTS: MULTINOMIAL LOGISTIC REGRESSION ........................................................... 30. II.

(5) 1. Introduction Modern financial economic theory is based on the assumption that the representative agent in the economy makes decisions according to the axioms of expected utility theory. However, rational models seem to have trouble explaining all that we see in financial markets. Hence the literature of behavioral finance draws on results from psychology to motivate the behavior of the agents in financial markets. However, one important challenge in the empirical study of behavioral finance is the availability of data on individual investor levels. The Survey of Consumer Finances (SCF), conducted by the Board of Governors of the Federal Reserve System, is normally a triennial cross-sectional survey of U.S. families, but over 1983–1989 and 2007–2009 periods, the survey collected panel data. The survey data include information on families’ balance sheets, pensions, income, and demographic characteristics. Information is also included from related surveys of pension providers and the earlier such surveys conducted by the Federal Reserve Board. The 2007-2009 SCF panel data, in particular, enable us to examine the change of individual investor behavior due to the sub-prime financial crisis and subsequent recession and whether any behavior change can be attributed to factors documented in the literature of behavioral finance. The purpose of this study is to investigate the change of risk aversion of household investors in the pre- and post- sub-prime financial crisis periods from a behavioral perspective. Using the 2007 and 2009 Survey of Consumer Finances (SCF) panel data, we examine whether the experience effects, including “snake-bit” effect and “house-money” effect, and mental accounting will influence the change of individual relative risk aversion in the pre- and post- sub-prime financial crisis 1.

(6) periods. In particular, the “snake-bit” effect suggests that after experiencing a financial loss, individual investors become less willing to take risk. Hence when faced with a gamble after already losing money, investors subject to snake-bit effect generally choose to decline the gamble. On the other hand, the “house-money” effect suggests that investors are willing to take more risk once they have experienced a gain or profit. The underlying reason is that gamblers refer to the feeling of experiencing a gain as playing with the house’s money. After winning a big profit, amateur gamblers don’t fully consider the new money as their own. Since gamblers don’t fully integrate their winnings with their own money, they act like they are betting with the casino’s money, and hence are willing to take more risk. Using the following variables to proxy for bad experience-capital losses, job losing, no income, bad investment, bankruptcy, declining in stock price, declining in real asset/other asset values, we find that bad experience during the financial crisis period will positively affect individual relative risk aversion index (RRAI, i.e., higher RRAI indicates more risk aversion) in the post-crisis period. On the other hand, using the following variables to proxy for good experience-getting jobs, an increase in income, and good investment, we find that good experience will negatively affect individual RRAI. These findings provide evidence to support that experience effects will influence the change of relative risk aversion of household investors. According to Richard Thaler (1999), “mental accounting is the set of cognitive operations used by individuals and households to organize, evaluate, and keep track of financial activities.” In general, household investors mentally place their money in different accounts for various purposes, including expense, retirement, and for the future, etc. This study investigates whether mental accounts are associated with 2.

(7) varying levels of risk aversion. We find that compared to regular accounts, retirement accounts show less positive change in RRAI in the post- sub-prime financial crisis period, indicating that household investors express less risk aversion in retirement account than in regular accounts after the sub-prime financial crisis.. 2. Literature Review and Hypothesis Development 2.1 Relative Risk Aversion Index William and Victor (1992) present how to examine the hypothesized relationships between risk tolerance and given variable indicates, such as wealth and age, in their paper-Asset Allocation and Individual Risk Aversion. Relative risk aversion decreases as one rises above the poverty level and as one growing up under 65. After age 65(retirement), risk aversion increase with age. The study looked at four classes of assets-personal property, real estate, bond and risky assets. The coefficient of the relative risk aversion is measured by the ratio of risky assets to wealth. The paper named Risk Aversion Revisited (1983) presents an empirical investigation of the demand for risky assets of individual using data from the SCF. According to this paper, two asset categories, riskless and risky assets, are identified. Risky assets include stocks, bonds, mutual funds, real estate except primary housing, equity in own business, and loans held. However, Friend-Blume (1975) recognizes that investment in primary house is not consistent with the assumption of equation, which is mentioned in this study, so housing is treated as a riskless asset. On the other hand, the sample is divided in two wealth classes with net worth of $12500 as the common class boundary. They are labeled the “lower wealth” and the “upper wealth” class. If household belongs lower wealth class, with increasing wealth, relative risk 3.

(8) aversion increases. The age variable is as important as key wealth variable, so this paper also shows someone would expect a negative age effect on risk-taking behavior if human capital is relevant. 2.2 Experience Effect Memory is reconstructive. If people had encountered some experience and they have somehow been written to the brain, when you encountered the similar situations, people may grab the memory from brain as a judgmental basis. The results of the matters will rewrite in your brain and be a new memory. For example, magicians always do this. At the beginning, we know this is that, when people witness an event and receive misleading information about it, this misinformation is often consolidated into their memory. Memory is not only reconstructive, but variable. People can bring to mind certain very positive or negative memories easily and quickly. Thomas and Diener (1990) explain that “the reason for this seems to be that events are remembered more vividly when they arouse emotions.” 2.2.1 Snake-bit effect Experience effects are using the past memories as a basis of judgment. Since pleasant memories make you happier than unpleasant ones, it is not surprising that we are following the emotions to make decision-making. If you have ever experienced the bad things, and that bring lots of pain to you, when you meet the similar situation again, and you will adopt the aversive decision, it is so-called “snake-bit effect”. Richard and Eric (1990) also recognized that “if prior losses are facilely integrated with subsequent outcomes, we would expect decision makers to be risk seeking; however, because integration is not automatic, an initial loss might cause an increase 4.

(9) in risk aversion, particularly when the second choice does not offer the opportunity to break even.” Therefore, if investors who had investment experience no matter what is good or bad, are not so rational when making investment decisions. H1.1: Household investors who had bad investment experience will express larger positive change in relative risk aversion (RRAI) than those who do not in the post- sub-prime financial crisis period. 2.2.2 House money effect If you use the prior gains to make the next decision, it makes you are willing to assume greater risk. Using the jargon of the casino, you are betting with “house money”. Based on Richard and Eric’s viewpoints (1990), “after a gain, subsequent losses that are smaller than the original gain can be integrated with the prior gain, mitigation the influence of loss aversion and facilitating risk-seeking.” In other words, house money effect is that investors are willing to take more and greater risks when they are gambling or investing with profits. Each household experienced some good matters or good investment experience in 2008, and gain some profit in their portfolios. It is like you get the house money from casino. Based on house money effect, it makes them take more risks. H1.2: Household investors who had good investment experience will express smaller positive change in RRAI than those who do not in the post- sub-prime financial crisis period. 2.3 Mental Accounting People use mental accounting as a method to make decision-making manageable. According to Richard Thaler, “mental accounting is the set of cognitive operations 5.

(10) used by individuals and households to organize, evaluate, and keep track of financial activities.” In general, many people place their money in different purpose, e.g. expense, retirement, and for the future etc. But these are mental concepts but not real accounts. Each household has their own mental accounts, and every investor is going to set different functions to different mental accounts. It has its own relative risk aversion index. Under the influence of mental account functions, changes of RRAI are based on your account function setting. H2: Compared to regular accounts, retirement accounts show less positive change in RRAI in the post- sub-prime financial crisis period.. 3. Data, Variables and Models 3.1 Data The data of this study is 2007-2009 Survey of Consumer Finances (SCF) panel, which is designed by Federal Reserve Board (FRB). According to surveying the aftermath of the storm, the SCF is a triennial cross-sectional survey, which is conducted by the FRB. However, there is an earlier history of the collection of panel data. For instance, the collection of wealth data started with the 1962 Survey of Financial Characteristics of Consumers and 1963 Survey of Changes in Financial Characteristics of Consumers. No further SCF panel interview had been conducted till the re-interview in 2009 with the participants of the 2007 SCF. According to “Surveying the Aftermath of the Storm: Changes in Family Finances from 2007 to 2009”, the information of data is as follows: “The 2007 SCF provided detailed information on all aspects of household finances, and most of this information is collected at the level individual items. 6.

(11) Generally speaking, the interview time war between 75 and 90 minutes, but when interviews for some participants with complicated finances is requiring up to four hours and several sessions from time to time. Because 2008 is a time of economic and financial turmoil that most families had no experienced, the 2009 SCF follow-up interview focused on a smaller set of variables that are most useful for understanding the nature of the changes experienced by families during the financial crisis. The 2009 re-interview also collected changes families made, plan of making in their portfolios, and key positive and negative events for the family between 2007 and 2009. On the other hand, the panel questionnaire design is possible to construct parallel estimates for all of the most important aspects of wealth in both 2007 and 2009.” 3.2 Variables- Definition of Relative Risk Aversion Index Asset allocation decision-making is a part of finance. Individuals provide their wealth into different portfolios by their individual risk preferences and degrees of risky aversion. How can we measure the investors’ preference of risk? The coefficient of the Arrow-Pratt relative risk aversion is measured by the ratio of risky assets to wealth. Following Friend and Blume (1975), we use relative risk aversion index(RRAI), which is proposed by William B. Riley Jr. and K. Victor Chow, for the kth investor as follows:. 𝑅𝑅𝐴𝐼𝑘 = (1-𝑟𝑖𝑠𝑘𝑦 𝑎𝑠𝑠𝑒𝑡𝑠/𝑤𝑒𝑎𝑙𝑡ℎ)=(1-𝑀𝑃𝑅/𝑅𝑅𝐴). (1). According to the paper- The Demand for Risky assets, we separated assets in different sorts. Assets are classified into financial and non-financial assets. Assets 7.

(12) include risky-free asset and risky asset, and risky asset includes mixed-risk assets and pure-risk assets. Because of using RRAI to estimate degree of investors’ risky aversion, the risky asset to net worth term is separated into two ways to measure. One kind is pure-risk assets (riskyp), such as stocks, miscellaneous assets, investment of real estate and so on, which is divided by net worth. We take year 2007 for example to define the following RRAI. RRAIs of year 2009 are defined in the same way.. 𝑝 𝑝 𝑅𝑅𝐴𝐼𝑖,𝑡=2007 = (1-𝑟𝑖𝑠𝑘𝑦𝑖,𝑡=2007 /𝑤𝑒𝑎𝑙𝑡ℎ𝑖,𝑡=2007)=(1-𝑀𝑃𝑅𝑖,𝑡=2007 /𝑅𝑅𝐴𝑖,𝑡=2007 ),. (2). Change in RRAI of pure-risk assets is defined as follows:. 𝑝 𝑝 △𝑅𝑅𝐴𝐼𝑖𝑝 =𝑅𝑅𝐴𝐼𝑖,𝑡=2009 -𝑅𝑅𝐴𝐼𝑖,𝑡=2007. (3). where riskyp =pure-risk asset and i=ith household. Furthermore, MPR is the market price of risk and RRA is relative risk aversion. The other RRAI measure is based on the sum of mixed-risk assets (riskym) plus pure-risk assets to net worth.. 𝑝+𝑚 𝑅𝑅𝐴𝐼𝑖,𝑡=2007 𝑝 𝑚 =(1-(𝑟𝑖𝑠𝑘𝑦𝑖,𝑡=2007 +𝑟𝑖𝑠𝑘𝑦𝑖,𝑡=2007 )/𝑤𝑒𝑎𝑙𝑡ℎ𝑖,𝑡=2007)=(1-𝑀𝑃𝑅𝑖,𝑡=2007 /𝑅𝑅𝐴𝑖,𝑡=2007 ), (4). 8.

(13) And change in RRAI of pure risk and mixed risk assets is defined as follows:. 𝑝+𝑚 𝑝+𝑚 △𝑅𝑅𝐴𝐼𝑖𝑝+𝑚 =𝑅𝑅𝐴𝐼𝑖,𝑡=2009 -𝑅𝑅𝐴𝐼𝑖,𝑡=2007. (5). For retirement account, we only use the stock value as pure-risk asset in retirement account to define the following RRAI.. 𝑟𝑒𝑡𝑖𝑟𝑒 𝑅𝑅𝐴𝐼𝑖,𝑡=2007 =(1-𝑠𝑡𝑜𝑐𝑘 𝑣𝑎𝑙𝑢𝑒 𝑖𝑛 𝑟𝑒𝑡𝑖𝑟𝑒𝑚𝑒𝑛𝑡 𝑎𝑐𝑐𝑜𝑢𝑛𝑡𝑖,𝑡=2007/ 𝑟𝑒𝑡𝑖𝑟𝑒𝑚𝑒𝑛𝑡 𝑎𝑐𝑐𝑜𝑢𝑛𝑡 𝑣𝑎𝑙𝑢𝑒𝑖,𝑡=2007 ),. (6). 𝑟𝑒𝑡𝑖𝑟𝑒 𝑟𝑒𝑟𝑖𝑟𝑒 △ 𝑅𝑅𝐴𝐼𝑖𝑟𝑒𝑡𝑖𝑟𝑒 =𝑅𝑅𝐴𝐼𝑖,t=2009 - 𝑅𝑅𝐴𝐼𝑖,t=2007 ,. (7). For regular account, we only use the stock value as pure-risk asset in regular account to define the following RRAI.. 𝑟𝑒𝑔𝑢𝑙𝑎𝑟. 𝑅𝑅𝐴𝐼𝑖,𝑡=2007= (1-𝑠𝑡𝑜𝑐𝑘 𝑣𝑎𝑙𝑢𝑒 𝑖𝑛 𝑟𝑒𝑔𝑢𝑙𝑎𝑟 𝑎𝑐𝑐𝑜𝑢𝑛𝑡𝑖,𝑡=2007 /𝑟𝑒𝑔𝑢𝑙𝑎𝑟 𝑎𝑐𝑐𝑜𝑢𝑛𝑡 𝑣𝑎𝑙𝑢𝑒𝑖,𝑡=2007 ),. 𝑟𝑒𝑔𝑢𝑙𝑎𝑟. △ 𝑅𝑅𝐴𝐼𝑖. 𝑟𝑒𝑔𝑢𝑙𝑎𝑟. 𝑟𝑒𝑔𝑢𝑙𝑎𝑟. =𝑅𝑅𝐴𝐼𝑖,2009 - 𝑅𝑅𝐴𝐼𝑖,2007 ,. (8). (9). 9.

(14) An increase in RRAI means an increase in risk aversion. RRAI decreases with each age category until the 65-and-older (retirement) but it decreases with wealth increasing. 3.3 Models 3.3.1 Experience Effect 3.3.1.1 Snake-bit Effect SCF provides the information regarding family annual income and net gains or losses from the sales of stocks, bonds and real estate. Moreover, SCF also provide the information regarding household’s opinions and whether they have been the most important negative events for your finances over the past two years. We run the following model for testing snake-bit effect (Hypothesis 1.1). Finishing running the model, we use △ 𝑅𝑅𝐴𝐼𝑖𝑝+𝑚 instead of △ 𝑅𝑅𝐴𝐼𝑖𝑝 to run the following regression model (10) once more.. △ 𝑅𝑅𝐴𝐼𝑖𝑝 =α+𝛽1 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐿𝑜𝑠𝑠𝑖 +𝛽2 𝐵𝑎𝑑 𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑖 +𝛽3 𝐵𝑎𝑑 𝐻𝑒𝑎𝑙𝑡ℎ𝑖 + 𝛽4 𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡𝑐𝑦𝑖 +𝛽5 𝑁𝑜 𝑊𝑎𝑔𝑒𝑖 +𝛽6Agei+𝛽7Collegei+ 𝛽8 △High Schooli+𝛽9 △Net Worthi+𝛽10Incomei+𝛽11Dagei+𝛽12Dincomei+𝛽13Dwealthi+𝜀𝑖,. (10). Where △ 𝑅𝑅𝐴𝐼𝑖𝑃 =change in RRAI of pure risk assets, Capital Lossi=dummy variable for capital loss; if capital loss>0, Capital Lossi =1; else Capital Lossi =0, Bad Investmenti=dummy variable, when the negative event is decline in stock price, decline in real estate prices, decline in other asset values, lost money, or. bad. investment. Bad Investmenti =1, else Bad Investmenti =0, Bad Healthi=dummy variable, when the negative event is had bad health, or death in 10.

(15) family, Bad Healthi =1, else =0, Bankruptcyi=dummy variable, when the negative event is bankruptcy, Bankruptcyi =1, else Bankruptcyi =0, No Wagei=dummy variable, when the negative is “no income” or “lost/laid off from/could not find job”, No Wagei =1, else =0, Agei=households age in 2009, Collegei=dummy variable for education; if have college diploma, Collegei =1; else Collegei =0, High Schooli=dummy variable for education; if have high school diploma, High Schooli =1; else High Schooli =0, △Networthi= each household net worth in 2009-each household net worth in 2007, △Incomei=each household income in 2009- each household income in 2007, Dagei=dummy variable for age in 2009; if <65, Dagei =0 and if >=65, Dagei =1, Dincomei=dummy variable for income in 2009; if<$22050, Dincomei =0 and if>=$22050, Dincomei =1, Dwealthi=dummy variable for wealth in 2009; if<$5056000, Dwealthi =0 and if>=$5056000, Dwealthi =1.. 3.3.1.2 House Money Effect SCF provide the information regarding household’s opinions about the most important positive events for finances over the past two years. We run the following model for testing snake-bit effect (Hypothesis 1.2). We also use △ 𝑅𝑅𝐴𝐼𝑖𝑝+𝑚 instead of △ 𝑅𝑅𝐴𝐼𝑖𝑝 to run the regression model (11) again.. △ 𝑅𝑅𝐴𝐼𝑖𝑝 = α+𝛽1 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐺𝑎𝑖𝑛𝑖 +𝛽2 𝑊𝑎𝑔𝑒𝑖 +𝛽3 𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑖 +𝛽4 Age+𝛽5Collegei+𝛽6High Schooli+ 𝛽7 △Net Worthi+𝛽8 △Incomei+𝛽9Dagei+𝛽10Dincomei+𝛽11Dwealthi+𝜀𝑖 (11) 11.

(16) Where △ 𝑅𝑅𝐴𝐼𝑖𝑝 = change in RRAI of pure risk assets, Capital Gaini= dummy variable for capital gain; if capital gain>0, Capital Gaini =1; else Capital Gaini =0, Wagei=dummy variable, when the positive event is had higher income from assets, had higher income, or got/kept job. Wagei =1; else Wagei =0, Investmenti= dummy variable, when the positive event is “purchased a home or other assets/ investments”, “sold a home or other assets”, “rearranged assets at a good time; was able to make favorable investments”, “assets increased in value”, or “market increased”. Investmenti =1; else Investmenti =0, Agei=household age in 2009, Collegei=dummy variable for education; if have college diploma, Collegei =1; else Collegei =0, High Schooli=dummy variable for education; if have high school diploma, High Schooli =1; else High Schooli =0, △Networthi= each household net worth in 2009- each household net worth in 2007, △Incomei=each household income in 2009- each household income in 2007, Dagei=dummy variable for age in 2009; if <65, Dagei =0 and if >=65, Dagei =1, Dincomei=dummy variable for income in 2009; if<$22050, Dincomei =0 and if>=$22050, Dincomei =1, Dwealthi=dummy variable for wealth in 2009; if<$5056000, Dwealthi =0 and if>=$5056000, Dwealthi =1.. 12.

(17) Finally, we pool all the snake-bit effect and house money effect variables and run the following model.. △ 𝑅𝑅𝐴𝐼𝑖𝑝 =α+𝛽1 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐿𝑜𝑠𝑠𝑖 +𝛽2 𝐵𝑎𝑑 𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑖 +𝛽3 𝐵𝑎𝑑 𝐻𝑒𝑎𝑙𝑡ℎ𝑖 + 𝛽4 𝐵𝑎𝑛𝑘𝑟𝑢𝑝𝑡𝑐𝑦𝑖 +𝛽5 𝑁𝑜 𝑊𝑎𝑔𝑒𝑖 +𝛽6 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐺𝑎𝑖𝑛𝑖 +𝛽7 𝑊𝑎𝑔𝑒𝑖 +𝛽8 𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑖 + 𝛽9Agei+𝛽10Collegei+ 𝛽11High Schooli +𝛽12 △NetWorthi+𝛽13 △Incomei+𝛽14Dagei+𝛽15Dincomei+𝛽16Dwealthi+𝜀𝑖. (12). All the variables are defined in model (10) and (11), and also we use △ 𝑅𝑅𝐴𝐼𝑖𝑝+𝑚 instead of △ 𝑅𝑅𝐴𝐼𝑖𝑝 to. run the regression above once more.. 3.3.2 Retirement and regular accounts: SCF provide detailed information of each household about their retirement accounts. We also could know every household invested how much money in different kinds of risk asset. We run the following two regressions to test our hypothesis 2.. 𝑅𝑅𝐴𝐼𝑖,𝑡 =α+𝛽1Retirementi,t+𝛽2Agei,t+𝛽3Collegei,t+𝛽4 High Schooli,t+ 𝛽5Net Worthi,t+𝛽6Incomei,t+𝛽7Dagei,t+𝛽8Dincomei,t+𝛽9Dwealthi,t+𝛽10Yeari,t+𝜀𝑖. (13). 13.

(18) Where 𝑟𝑒𝑔𝑢𝑙𝑎𝑟. 𝑅𝑅𝐴𝐼𝑖,𝑡 = Stack all 𝑅𝑅𝐴𝐼𝑖𝑟𝑒𝑡𝑖𝑟𝑒 and 𝑅𝑅𝐴𝐼𝑖. of 2007 and 2009, 𝑟𝑒𝑔𝑢𝑙𝑎𝑟. Retirementi,t:=when 𝑅𝑅𝐴𝐼𝑖 =𝑅𝑅𝐴𝐼𝑖𝑟𝑒𝑡𝑖𝑟𝑒 then Retirementi,t =1; 𝑅𝑅𝐴𝐼𝑖 =𝑅𝑅𝐴𝐼𝑖 Retirementi,t =0, Agei,t =household age,. Collegei,t =dummy variable for education; if have college diploma, Collegei,t =1; else Collegei,t =0, High Schooli,t =dummy variable for education; if have high school diploma, High Schooli,t =1; else High Schooli,t =0, Net Worthi,t = net wealth of each household, Incomei,t =income of each household, Dagei,t =dummy variable for age; if <65, Dagei,t =0 and if >=65, Dagei,t =1, Dincomei,t =dummy variable for income; if<$22050, Dincomei,t =0 and if>=$22050, Dincomei,t =1, Dwealthi,t =dummy variable for wealth; if<$5056000, Dwealthi,t =0 and if>=$5056000, Dwealthi,t =1, Yeari,t =dummy variable for year; if year=2007, Yeari,t =1 and if year=2009, Yeari,t =0.. 14.

(19) △ 𝑅𝑅𝐴𝐼𝑖 =α+𝛽1Retirementi+𝛽2 Agei +𝛽3Collegei+𝛽4 High Schooli+ 𝛽5 △Net Worthi+𝛽6 △Incomei+𝛽7Dagei+𝛽8Dincomei+𝛽9Dwealthi+𝜀𝑖. (14). Where 𝑟𝑒𝑔𝑢𝑙𝑎𝑟. △ 𝑅𝑅𝐴𝐼𝑖 = stack all △ 𝑅𝑅𝐴𝐼𝑖𝑟𝑒𝑡𝑖𝑟𝑒 and △ 𝑅𝑅𝐴𝐼𝑖. , 𝑟𝑒𝑔𝑢𝑙𝑎𝑟. Retirementi:=when 𝑅𝑅𝐴𝐼𝑖 =𝑅𝑅𝐴𝐼𝑖𝑟𝑒𝑡𝑖𝑟𝑒 then Retirementi =1; 𝑅𝑅𝐴𝐼𝑖 =𝑅𝑅𝐴𝐼𝑖. ,. Retirementi =0, Agei =household age in 2009, Collegei,t =dummy variable for education in 2009; if have college diploma, Collegei,t =1; else Collegei,t =0, High Schooli =dummy variable for education in 2009; if have high school diploma, High Schooli=1; else High Schooli=0, △Net Worthi= each household net worth in 2009- each household net worth in 2007, △Incomei=each household income in 2009- each household income in 2007, Dagei=dummy variable for age in 20009; if <65, Dagei =0 and if >=65, Dagei =1, Dincomei=dummy variable for income in 2009; if<$22050, Dincomei =0 and if>=$22050, Dincomei =1, Dwealthi=dummy variable for wealth in 2009; if<$5056000, Dwealthi =0 and if>=$5056000, Dwealthi =1.. 15.

(20) 4. Result Table 1 presents means, median, 25th percentile, and 75th percentile of all kinds of assets. These assets are defined and classified based on the definitions of Friend and Blume’s paper (1975). From Table 2, the mean value of pure risk of RRAI in 2007 is 0.5149, the mean value of pure risk of RRAI in 2009 is 0.7464, the mean value of pure and mixed risk of RRAI in 2007 is 0.5045, and the mean value of pure and mixed risk of RRAI in 2009 is 0.7318. All of them are between 0 and 1. The value of RRAI in 2009 are greater than in 2007, so on average investors in 2009 are more risk aversion than those in 2007. From Table 3, in average, the median value for income in 2007 is $50,000, and the median value for income in 2009 is $49,000. The median value for financial assets in 2007 is $30,000, and the median value for financial assets in 2009 is $29,500. These values of them are similar with the paper of Bricker, Bucks, Kennickell, Mach, and Moore (2011). The percentage of households who suffered capital loss is 3.8% in 2007, but it is 14.5% in 2009. The percentage of households who suffered capital gain is 19.6% in 2007, but it decreases to 11.76% in 2009. Similarly, in 2007 the percentage of the other behavioral variables is 0%, and all of the percentage increases in 2009. The regression results of snake-bit effect and house money effect are given in Table 4 panel A. In model (1), we focused on snake-bit effect and change in RRAI of pure risk assets, the signs for the capital loss, bad investment, bad health, and bankruptcy variables are positive. The t-statistic for capital loss variable is significant at the 0.01 level, the t-statistic for bad investment variable is significant at the 0.05 16.

(21) level, and the t-statistic for bad health variable is significant at the 0.1 level. They all mean when you had any bad experience in the period from 2007 to 2009, bad investment results in the period of financial crisis, and your health got worse off than you are before, made the change in RRAI greater what means householders became more risk aversion. The results match with our hypothesis. The sign for age variable is positive and significant at the 0.01 level. It indicated that the more in age, the more risk aversion. In this financial crisis period, most people are relative risk aversion. The signs for the college and high school variables are negative, and high school variable is significant at the 0.01 level. It indicates that the higher level of education, the more risk seeking. The sign for change in net worth from 2007 to 2009 variable is positive and significant at the 0.01 level. It indicates that if you increased in net worth, you are a risk aversion investor. In other words, in the period of financial crisis maybe if you are a risk aversive investor, you will be able to increase your net worth. The sign for Dwealth variable is positive and significant at the 0.01 level. People with wealth reached the 90th percentile, are more risk aversion. Based on 2008 and 2010 world wealth report information, it showed that the high net worth individuals’ allocation of financial assets in fixed income changed from 27% to 31%, and in equities changed from 33% to 29%. That both represented the high net worth individuals are more risk aversion. In model (2), we run the regression once more with change in RRAI of pure and mixed risk assets instead of change in RRAI of pure risk assets. The results are consistent with the results we mentioned. In model (3) and (4), we focus on testing house money effect. First, we use change in RRAI of pure risk assets to run. The signs for the wage and investment variables are negative, and the t-statistics of wage variable is significant at the 0.05 level. If the households got more money in wage against the others, they are more risk seeking. The result is matched with our 17.

(22) hypothesis 1.2. We use change in RRAI of pure and mixed risk assets instead of change in RRAI of pure risk assets running the regression once more. The results are consistent with the results mentioned. Especially, the sign for the investment variable is negative and its t-statistics is significant at the 0.05 level. If the households got a good experience in investment, they are more risk seeking. However, the sign for the capital gain is positive and its t-statistics is significant at the 0.05 level. The result is beyond our imagination. According to Burman and Ricoy’s paper-Capital gains and the people who realize them (1997), they said “people who have higher incomes realize a large fraction of the taxable capital gains. In fact, most of the tax returns that report capital gains are filed by people whose incomes are under $50,000 a year. But their capital gains are very small compared with those of taxpayers who have high incomes.” So, the possible explanation is that realizing capital gain are the people who had higher incomes, and higher incomes people are more risk aversion in this period, based on world wealth report information. In the model (5) and (6), we pool all the variable mentioned, and run the regression. The results of the both regressions are consistent. We separate households into three groups by net worth, and use the highest one-third group as my target group. To run the regression model (10) to (12) again, the results are showed in Table 4 panel B. The signs for the capital loss, bad investment, bad health, and no wage variables are positive. The t-statistic for capital loss variable is significant at the 0.01 level, the t-statistic for bad investment variable is significant at the 0.05 level, the t-statistic for bad health variable is significant at the 0.05 level, and the t-statistic for no wage variable is significant at the 0.01 level. They all mean when you had any bad experience in the period from 2007 to 2009, bad investment results in the period of financial crisis, and your health got worse off than 18.

(23) you were before, made the change in RRAI greater what means householders became more risk aversion. The sign for the investment variable is negative, and the t-statistics of investment variable is significant at the 0.01 level. If the households got good experience in investing against the others, they are more risk seeking. These results in Table 4 panel B are consistence with the results in Table 4 panel A. Table 5 displays the results of hypothesis 2. The model (1), fixed-effect model, is designed to test whether RRAI of regular account is bigger than RRAI of retirement account. It’s a cross-sectional data. The sign for retirement variable is negative and its t-statistics is significant at the 0.01 level. The RRAI of retirement account is less than the RRAI of regular account. That is, the retirement account (relative long-term account) is more risk seeking than the regular account (relative short-term account). Risk aversion decreases with age and education increasing. The more income you got, and you are more risk seeking. If the households whose income is higher than property level or net worth is above 90th percentile are more risk seeking. RRAI of 2009 is bigger than RRAI of 2007. It indicated that, in general, people in 2009 are more risk aversion than in 2007. From Table 5, the model (2) is designed to test whether the change in RRAI of retirement is less or not. The results match with our hypothesis 2. The sign of retirement variable is negative and its t-statistic is significant at the 0.01 level. The change in RRAI of retirement account is less than change in RRAI of regular account. That is, the relative long-term account is more risk seeking than the relative short-term account. On the other hand, we find the change in RRAI increased with age, education, and change in income. But, the sign of Dincome variable is negative and its t-statistics is significant at the 0.01 level. Change in RRAI of the households whose income is higher than property level is less than the others. That is, those households whose income is higher are more risk seeking. 19.

(24) 5. Robustness analysis Because change in RRAI includes not only change of individual investing behavior but also change of market value, we use SRT (Subjective Risk Taking) as a new dependent variable to test hypothesis 1.1 and 1.2 once more. In 2007 to 2009 SCF survey, they asked households that which of the following statements comes closest to describing the amount to financial risk that you are willing to take when you save or make investments? 1: Take substantial financial risks expecting to earn substantial returns. 2: Take above average financial risks expecting to earn above average returns. 3. Take average financial risks expecting to earn average returns. 4. Not willing to take any financial risks. We use the change of attitude as my new dependent variable-SRT. From 2007 to 2009, first, if household changed their status from smaller number to bigger number, for example from 1 to 4 or 2 to 3, means they become more risk aversion. Secondly, if they didn’t change their status means they are no change. Thirdly, if they changed their status from bigger number to smaller number, for example from 3 to 2 or 4 to 1, means they become more risk seeking. The dependent variable-SRT is a dummy variable. When the household belongs to condition 1, the SRT equals 1. When the household belongs to condition 2, the SRT equals 2. When the household belongs to condition 3, the SRT equals 3. We use SRT instead of change in RRAI to run multinomial logistic regression of the model (10) to (12). The result of these multinomial regressions is showed in Table 6. The results of snake-bit effect are consistent with the results we mentioned. If the household have 20.

(25) ever encountered negative events such as capital losses, lose their jobs or bad health, they become more risk aversion. However, the results of house money effect are not consistent. The probable reason is the difference between household’s determination of subjective risk taking and their financial investing behavior.. 6. Conclusion The relative risk aversion research of William Riley and Victor Chow presented the RRAI decreases with age, education and wealth. Following the research of them, we associated the RRAI measures with behavioral finance concepts to test the relations of RRAI and household behavioral finance factors. This study uses the survey data from the Survey of Consumer Finances and examines the effect of snake-bit effect, house money effect, and mental accounting on the measure of relative risk aversion. We have several findings, first, we find that households who encountered bad experience in 2008 became more risk aversion when they made investment decision in 2009, where bad experience includes bad investment performance, getting worse health condition, or ever suffering capital losses in the financial crisis in 2008. On the other hand, when households have good experience, or have so-called "house money" by encountering positive events, they become more risk seeking in their 2009 decision-making. Second, we find that households have less risk aversion in their retirement accounts than in regular accounts. The results are consistent with mental accounting concept.. 21.

(26) 7. Reference Ackert, Lucy F., and Richard Deaves, 2009, Behavioral finance: psychology, decision-making, and markets, South-western Cengage Learning (USA). Burman, Leonard E. and Peter D. Ricoy, 1997, Capital gains and the people who realize them, National Tax Journal 50, 427-451. Friend, Irwin, and Marshall E. Blume, 1975, The demand for risky assets, The American Economic Review 65, 900-922. Jesse Bricker, Brian Bucks, Arthur Kennickell, Traci Mach, and Kevin Moore, 2011, “Surveying the Aftermath of the Storm: Changes in Family Finances from 2007 to 2009.” Morin, Roger-A. and A. Fernandez Suarez, 1983, Risk aversion revisited, The Journal of Finance 38, 1201-1216. Richard H. Thaler and Eric J. Johnson, 1990, Gambling with the house money and trying to break even: the effects of prior outcomes on risky choice, Management Science 36, 643-660. Thaler, R. H., 1999, Mental accounting matters, Journal of Behavioral Decision Making 12, 183-206. Thomas, D. L., and E. Diener, 1990, Memory accuracy in the recall of emotions, Journal of Personality and Social Psychophysics 16, 355-357 William B. Riley Jr. and K. Victor Chow, 1992, Asset allocation and individual risk aversion, Financial Analysts Journal 48, 32-37.. 22.

(27) Table 1. Summary Statistics of Assets. This information is from Survey of Consumer Finances, and Uninc. is unincorporated. Panel A year=2007 Mean Risk-Free Assets. Media. 25th percentile. 75th percentile. Checking Accounts. 77,234. n 3,000. 800. 15,000. Saving Accounts. 202,079. 300. 0. 13,000. Call accounts. 77,468. 0. 0. 0. Certificate of deposit. 58,727. 0. 0. 0. Saving Bonds. 3,223. 0. 0. 0. Life Insurance. 64,980. 0. 0. 2,400. 363,883. 22,000. 0. 208,000. 571,322. 0. 0. 0. Common and Preferred Stock. 1,781,132. 0. 0. 12,000. Mutual Funds. 1,192,556. 0. 0. 0. 340,279. 0. 0. 0. 2,823,088. 18900. 0. 341,000. 110,276. 0. 0. 0. Media. 25th percentile. 75th percentile. Retirement Accounts m. Mixed-risk Assets (risky ) State and Local Bonds p. Pure-risk Assets (risky ). Other Managed Assets Equity in Uninc. Business Other Financial asset Number of households Panel B. 3577. year=2009 Mean. Risk-Free Assets Checking Accounts. 47,590. n 3,000. 630. 11,000. Saving Accounts. 198,005. 1,000. 0. 22,000. Call accounts. 222,056. 0. 0. 0. Certificate of deposit. 89,595. 0. 0. 0. Saving Bonds. 5,544. 0. 0. 0. Life Insurance. 83,943. 0. 0. 2,500. 294,270. 22,000. 0. 187,500. 626,011. 0. 0. 0. 1,262,428. 0. 0. 8,100. Mutual Funds. 599,774. 0. 0. 0. Other Managed Assets. 443,796. 0. 0. 0. 1,983,525. 16,750. 0. 225,000. 139,198. 0. 0. 0. Retirement Accounts m. Mixed-risk Assets (risky ) State and Local Bonds p. Pure-risk Assets (risky ) Common and Preferred Stock. Equity in Uninc. Business Other Financial asset Number of households. 3472 23.

(28) Table 2. Statistics of Relative risk Aversion Index (RRAI). Relative risk Aversion Index (RRAI) is defined as follows. 𝑝. 𝑝. 𝑅𝑅𝐴𝐼𝑖,𝑡=2007 =(1-𝑟𝑖𝑠𝑘𝑦𝑖,𝑡=2007 /𝑤𝑒𝑎𝑙𝑡ℎ𝑖,𝑡=2007 )=(1-𝑀𝑃𝑅𝑖,𝑡=2007 /𝑅𝑅𝐴𝑖,𝑡=2007 ), 𝑝+𝑚. 𝑝. 𝑚 𝑅𝑅𝐴𝐼𝑖,𝑡=2007 =(1-(𝑟𝑖𝑠𝑘𝑦𝑖,𝑡=2007+𝑟𝑖𝑠𝑘𝑦𝑖,𝑡=2007 )/𝑤𝑒𝑎𝑙𝑡ℎ𝑖,𝑡=2007 )=(1-𝑀𝑃𝑅𝑖,𝑡=2007/𝑅𝑅𝐴𝑖,𝑡=2007) 𝑟𝑒𝑡𝑖𝑟𝑒 𝑅𝑅𝐴𝐼𝑖,𝑡=2007 =(1-𝑠𝑡𝑜𝑐𝑘 𝑣𝑎𝑙𝑢𝑒 𝑖𝑛 𝑟𝑒𝑡𝑖𝑟𝑒𝑚𝑒𝑛𝑡 𝑎𝑐𝑐𝑜𝑢𝑛𝑡𝑖,𝑡=2007 /𝑟𝑒𝑡𝑖𝑟𝑒𝑚𝑒𝑛𝑡 𝑎𝑐𝑐𝑜𝑢𝑛𝑡 𝑣𝑎𝑙𝑢𝑒𝑖,𝑡=2007 ), 𝑟𝑒𝑔𝑢𝑙𝑎𝑟. and 𝑅𝑅𝐴𝐼𝑖,𝑡=2007 =(1-𝑠𝑡 �𝑜𝑐𝑘 𝑣𝑎𝑙𝑢𝑒 𝑖𝑛 𝑟𝑒𝑔𝑢𝑙𝑎𝑟 𝑎𝑐𝑐𝑜𝑢𝑛𝑡𝑖,𝑡=2007 /𝑟𝑒𝑔𝑢𝑙𝑎𝑟 𝑎𝑐𝑐𝑜𝑢𝑛𝑡 𝑣𝑎𝑙𝑢𝑒𝑖,𝑡=2007 ).. Panel A RRAI in 2007 and 2009. 25th percentile. 75th percentile. Mean. Median. RRAI 𝑝 in 2007. 0.5149. 0.8998. 0.6029. 1.0000. RRAI 𝑝 in 2009. 0.7464. 0.9068. 0.6175. 1.0000. 𝑝+𝑚. in 2007. 0.5045. 0.8952. 0.5809. 1.0000. RRAI 𝑝+𝑚 in 2009. 0.7318. 0.9007. 0.5889. 1.0000. RRAI. Panel B RRAI of Retirement Accounts and RRAI of Regular Accounts in 2007 and 2009. RRAI 𝑟𝑒𝑡𝑖𝑟𝑒 in 2007 RRAI. 𝑟𝑒𝑡𝑖𝑟𝑒. in 2009. RRAI 𝑟𝑒𝑔𝑢𝑙𝑎𝑟 in 2007 RRAI. 𝑟𝑒𝑔𝑢𝑙𝑎𝑟. in 2009. Mean. Median. 25th percentile. 75th percentile. 0.8095. 1.0000. 0.6000. 1.0000. 0.8097. 1.0000. 0.6000. 1.0000. 0.8747. 1.0000. 0.8814. 1.0000. 0.8855. 1.0000. 0.9114. 1.0000. 24.

(29) Table 3. Summary Statistic of main variables. This information is from Survey of Consumer Finances. Capital Lossi=dummy variable for capital loss; if capital loss>0, Capital Lossi =1; else Capital Lossi =0, Bad Investmenti=dummy variable, when the negative event is decline in stock price, decline in real estate prices, decline in other asset values, lost money, or bad investment. Bad Investment i =1, else Bad Investmenti =0, Bad Healthi=dummy variable, when the negative event is had bad health, or death in family, Bad Healthi =1, else =0, Bankruptcyi=dummy variable, when the negative event is bankruptcy, Bankruptcyi =1, else Bankruptcyi =0, No Wagei=dummy variable, when the negative is “no income” or “lost/laid off from/could not find job”, No Wagei =1, else =0, Capital Gaini= dummy variable for capital gain; if capital gain>0, Capital Gaini =1; else Capital Gaini =0, Wagei=dummy variable, when the positive event is had higher income from assets, had higher income, or got/kept job. Wage i =1; else Wagei =0, Investmenti= dummy variable, when the positive event is “purchased a home or other assets/ investments”, “sold a home or other assets”, “rearranged assets at a good time; was able to make favorable investments”, “assets increased in value”, or “market increased”. 35, Investmenti =1; else Investmenti =0, Agei=households age, Collegei=dummy variable for education; if have college diploma, Collegei =1; else Collegei =0, High Schooli=dummy variable for education; if have high school diploma, High Schooli =1; else High Schooli =0, Networthi= each household net worth, Incomei=each household income, Dagei=dummy variable for age; if <65, Dagei =0 and if >=65, Dagei =1, Dincomei=dummy variable for income; if<$22050, Dincome i =0 and if>=$22050, Dincome i =1, Dwealthi=dummy variable for wealth in 2009; if<$5056000, Dwealth i =0 and if>=$5056000, Dwealth i =1. 2007 Mean. 2009 Median. Mean. Median. Behavioral variables 0.0380. 0.1450. Bad Investment. 0. 0.3093. Bad Health. 0. 0.0409. Bankruptcy. 0. 0.0053. No wage. 0. 0.1117. 0.1956. 0.1176. Wage. 0. 0.2233. Investment. 0. 0.0983. Capital Loss. Capital Gain. Characteristic and financial variables Age. 52. Dage. 0.2091. 0.2609. College. 0.494509. 0.5093. High School. 0.263966. 0.2472. Income. 763,827. Dincome. 0.9561. Net worth. 6,850,342. Dwealth. 0.6216. Financial Assets. 218,290. Number of households. 52. 50,000. 55. 55. 428,966. 49,000. 0.9466 352,600. 6,571,501. 335,810. 0.6203 30,000 3577. 193,665. 29,500 3472. 25.

(30) Table 4. Regression Results: Snake-bit Effect and House Money Effect. All dummy variables are in 2009. Capital Lossi=dummy variable for capital loss; if capital loss>0, Capital Lossi =1; else Capital Lossi =0, Bad Investmenti=dummy variable, when the negative event is decline in stock price, decline in real estate prices, decline in other asset values, lost money, or bad investment. Bad Investmenti =1, else Bad Investmenti =0, Bad Healthi=dummy variable, when the negative event is had bad health, or death in family, Bad Healthi =1, else =0, Bankruptcyi=dummy variable, when the negative event is bankruptcy, Bankruptcyi =1, else Bankruptcyi =0, No Wagei=dummy variable, when the negative is “no income” or “lost/laid off from/could not find job”, No Wagei =1, else =0, Capital Gaini= dummy variable for capital gain; if capital gain>0, Capital Gaini =1; else Capital Gaini =0, Wagei=dummy variable, when the positive event is had higher income from assets, had higher income, or got/kept job. Wagei =1; else Wagei =0, Investmenti= dummy variable, when the positive event is “purchased a home or other assets/ investments”, “sold a home or other assets”, “rearranged assets at a good time; was able to make favorable investments”, “assets increased in value”, or “market increased”. 35, Investmenti =1; else Investmenti =0, Agei=households age in 2009, Collegei=dummy variable for education; if have college diploma, College i =1; else Collegei =0, High Schooli=dummy variable for education; if have high school diploma, High School i =1; else High Schooli =0, △Networthi= each household net worth in 2009-each household net worth in 2007, △Incomei=each household income in 2009- each household income in 2007, Dagei=dummy variable for age in 20009; if <65, Dagei =0 and if >=65, Dagei =1, Dincomei=dummy variable for income in 2009; if<$22050, Dincomei =0 and if>=$22050, Dincomei =1, Dwealthi=dummy variable for wealth in 2009; if<$5056000, Dwealthi =0 and if>=$5056000, Dwealthi =1.. 26.

(31) Panel A Full Sample Snake-bit effect Variables. (1)Pure risk. House money effect. (2)Pure+mixed risk. coefficient. t-value. -0.1056***. -4.46. -0.1044***. -4.42. Capital Loss. 0.0686***. 5.88. 0.0521***. Bad Investment. 0.0220**. 2.39. Bad Health. 0.0375*. Bankruptcy. Intercept. No Wage. coefficient. t-value. (3)Pure risk coefficient -0.0960***. t-value -4.08. Full model. (4)Pure+mixed risk coefficient. t-value. -0.0947***. -4.04. (5)Pure risk coefficient. (6)Pure+mixed risk. t-value. coefficient. t-value. -0.0956***. -3.99. -0.0936***. -3.92. 4.48. 0.0826***. 6.77. 0.0689***. 5.66. 0.0188**. 2.05. 0.0190**. 2.05. 0.0153*. 1.66. 1.85. 0.0357*. 1.77. 0.0367*. 1.81. 0.0349*. 1.73. 0.0578. 1.02. 0.0582. 1.03. 0.0539. 0.95. 0.0535. 0.95. -0.0104. -0.81. -0.0123. -0.96. -0.0091. -0.71. -0.0108. -0.84. 0.0275**. 2.23. 0.0413***. 3.37. Wage. -0.0242**. -2.48. -0.0252***. -2.60. -0.0223**. -2.30. -0.0236**. -2.43. Investment. -0.0212. -1.59. -0.0283**. -2.13. -0.0274**. -2.05. -0.0333**. -2.50. Capital Gain. Age-2009. 0.1647***. 4.03. 0.1620***. 3.97. 0.1774***. 4.36. 0.1709***. 4.21. 0.0510***. 0.1541***. 3.97. 3.76. 0.0609***. 0.1504***. 4.75. 3.68. College-2009. -0.0151. -1.47. -0.0145. -1.40. -0.0048. -0.47. -0.0071. -0.70. -0.0185*. -1.77. -0.0184*. -1.78. High School-2009. -0.0294***. -2.62. -0.0277**. -2.47. -0.0313***. -2.78. -0.0291***. -2.59. -0.0287**. -2.56. -0.0269**. -2.40. △Net Worth. 0.0648***. 2.60. 0.0457*. 1.84. 0.0697***. 2.79. 0.0519**. 2.08. 0.0706***. 2.83. 0.0525**. 2.11. △Income. 0.0261. 0.23. -0.0918. -0.80. -0.0271. -0.23. -0.1434. -1.25. 0.0077. 0.07. -0.1147. -1.00. Dage-2009. 0.0117. 0.87. 0.0097. 0.72. 0.0084. 0.62. 0.0058. 0.43. 0.0074. 0.54. 0.0048. 0.36. Dincome-2009. 0.0035. 0.30. 0.0054. 0.47. 0.0016. 0.13. 0.0036. 0.31. 0.0039. 0.33. 0.0057. 0.48. Dwealth-2009. 0.0391***. 3.43. 0.0212*. 1.87. 0.0554***. 4.93. 0.0321***. 2.87. 0.0307***. 2.62. 0.0116. 0.99. Number of households. 3,298. Adjusted R-square. 0.01. 0.01. 0.01. 0.01. 0.01. 0.01. 27.

(32) Panel B. Subsample: Based on Household Net Worth ranked at the top one-third Snake-bit effect. Variables. Intercept. (1)Pure risk. House money effect. (2)Pure+mixed risk. coefficient. t-value. -0.0891**. -2.57. coefficient -0.0702**. t-value -2.09. (3)Pure risk coefficient -0.0455. t-value -1.32. Full model. (4)Pure+mixed risk coefficient. t-value. -0.0402. -1.21. (5)Pure risk coefficient. (6)Pure+mixed risk. t-value. -0.0982***. -2.81. coefficient -0.0805**. t-value -2.38. Capital Loss. 0.0489***. 5.64. 0.0287***. 3.42. 0.0595***. 6.28. 0.0450***. 4.90. Bad Investment. 0.0181**. 2.24. 0.0114. 1.45. 0.0168**. 2.07. 0.0095. 1.21. Bad Health. 0.0797***. 2.65. 0.0668**. 2.29. 0.0755**. 2.51. 0.0629**. 2.16. Bankruptcy. -0.1617***. -1.66. 0.0682***. 2.94. No Wage. -0.1549. -1.64. 0.0655***. -0.1689*. 2.91. -1.74. -0.1669*. -1.78. 0.0691***. 2.98. 0.0666***. 2.97. Capital Gain. 0.0009. 0.1. 0.0226**. 2.48. 0.0289***. 2.81. 0.0441***. 4.42. Wage. 0.0271**. 2.37. 0.0241**. 2.19. 0.0252**. 2.22. 0.0227**. 2.07. -0.0380***. -3.59. -0.0500***. -4.89. -0.0421***. -3.99. -0.0531***. -5.19. 0.2871***. 5.37. 0.2433***. 4.71. 0.2944***. 5.52. 0.2517***. 4.88. Investment Age-2009. 0.2892***. 5.43. 0.2480***. 4.80. College-2009. -0.0159. -1.28. -0.0157. -1.30. -0.0146. -1.18. -0.0154. -1.29. -0.0171. -1.38. -0.0170. -1.41. High School-2009. -0.0101. -0.57. -0.0007. -0.04. -0.0115. -0.66. 0.0005. 0.03. -0.0071. -0.40. 0.0031. 0.18. 0.0565***. 3.77. 0.0760***. 4.92. 0.0573***. 3.83. -0.1221*. -1.75. 0.0088. 0.12. -0.1193*. -1.72. △Net Worth. 0.0727***. 4.70. △Income. 0.0322. 0.45. Dage-2009. -0.0076. Dincome-2009. -0.0414***. Dwealth-2009. 0.0034. Number of households. 1,236. Adjusted R-square. 0.02. 0.0524***. 3.49. 0.0752***. 4.84. -0.0895. -1.29. 0.0039. 0.05. -0.56. -0.0065. -0.50. -0.0014. -0.1. -0.0032. -0.24. -0.0083. -0.62. -0.0086. -0.66. -2.62. -0.0318**. -2.07. -0.0601***. -3.86. -0.0452***. -3.00. -0.0416***. -2.63. -0.0319**. -2.08. 0.42. -0.0132*. -1.66. -0.0089. -1.12. 0.64. -0.0121. -1.51. 0.02. 0.0104. 1.27. 0.02. 0.02. 0.0053. 0.03. 0.02. 28.

(33) Table 5. Regression Results: Mental Accounting 𝑟𝑒𝑔𝑢𝑙𝑎𝑟. 𝑅𝑅𝐴𝐼𝑖 =Stack all 𝑅𝑅𝐴𝐼𝑖𝑟𝑒𝑡𝑖𝑟𝑒 and 𝑅𝑅𝐴𝐼𝑖 of 2007 and 2009, △ 𝑅𝑅𝐴𝐼𝑖 = stack all △ 𝑅𝑅𝐴𝐼𝑖𝑟𝑒𝑡𝑖𝑟𝑒 𝑟𝑒𝑔𝑢𝑙𝑎𝑟 and △ 𝑅𝑅𝐴𝐼𝑖 , Year=dummy variable for year; if year=2007, Year=1 and if year=2009, Year=0, 𝑟𝑒𝑔𝑢𝑙𝑎𝑟 Retirement=when 𝑅𝑅𝐴𝐼𝑖 =𝑅𝑅𝐴𝐼𝑖𝑟𝑒𝑡𝑖𝑟𝑒 then Retirement=1; 𝑅𝑅𝐴𝐼𝑖 =𝑅𝑅𝐴𝐼𝑖 Retirement=0. (2)△ 𝐑𝐑𝐀𝐈𝑖. (1) RRAIi Variable. coefficient. t-value. coefficient. t-value. Intercept. 1.0569***. 162.84. Intercept. -0.0222**. -1.97. Retirement. -0.0739***. -35.10. Retirement. -0.0185***. -5.27. Age. -0.1387***. -12.90. Age-2009. 0.0684***. 3.67. College. -0.0762***. -28.20. College-2009. 0.0135***. 2.97. High School. 0.0098***. 3.18. High School-2009. 0.0120**. 2.27. Net Worth. 0.0083. 1.30. △Net Worth. -0.0057. -0.54. Income. -0.2129***. -5.91. △Income. 0.0873*. 1.89. Dage. -0.0117***. -3.23. Dage-2009. 0.0057. 0.97. Dincome. -0.0379***. -10.60. Dincome-2009. -0.0173***. -2.94. Dwealth. -0.0758***. -22.65. Dwealth-2009. 0.0053. 0.12. Year. -0.0019. -0.91. Number of Households. 6,577. Number of Households. 3,360. Adjusted R-square. 0.10. Adjusted R-square. 0.01. 29.

(34) Table 6. Regression Results: Multinomial Logistic Regression. All dummy variables are in 2009. Capital Lossi=dummy variable for capital loss; if capital loss>0, Capital Lossi =1; else Capital Lossi =0, Bad Investmenti=dummy variable, when the negative event is decline in stock price, decline in real estate prices, decline in other asset values, lost money, or bad investment. Bad Investmenti =1, else Bad Investmenti =0, Bad Healthi=dummy variable, when the negative event is had bad health, or death in family, Bad Healthi =1, else =0, Bankruptcyi=dummy variable, when the negative event is bankruptcy, Bankruptcyi =1, else Bankruptcyi =0, No Wagei=dummy variable, when the negative is “no income” or “lost/laid off from/could not find job”, No Wagei =1, else =0, Capital Gaini= dummy variable for capital gain; if capital gain>0, Capital Gaini =1; else Capital Gaini =0, Wagei=dummy variable, when the positive event is had higher income from assets, had higher income, or got/kept job. Wagei =1; else Wagei =0, Investmenti= dummy variable, when the positive event is “purchased a home or other assets/ investments”, “sold a home or other assets”, “rearranged assets at a good time; was able to make favorable investments”, “assets increased in value”, or “market increased”. 35, Investmenti =1; else Investmenti =0, Agei=households age in 2009, Collegei=dummy variable for education; if have college diploma, Collegei =1; else Collegei =0, High Schooli=dummy variable for education; if have high school diploma, High School i =1; else High Schooli =0, △Networthi= each household net worth in 2009-each household net worth in 2007, △Incomei=each household income in 2009- each household income in 2007, Dagei=dummy variable for age in 20009; if <65, Dagei =0 and if >=65, Dagei =1, Dincomei=dummy variable for income in 2009; if<$22050, Dincomei =0 and if>=$22050, Dincomei =1, Dwealthi=dummy variable for wealth in 2009; if<$5056000, Dwealthi =0 and if>=$5056000, Dwealthi =1.. 30.

(35) Snake-bit effect Variable. Full model. risk averse compared. no change compared. risk averse compared. no change compared to. risk averse compared. no change compared. to risk seeking. to risk seeking. to risk seeking. risk seeking. to risk seeking. to risk seeking. coefficient Intercept. House money effect. -0.0167. p-value. coefficient. 0.91. 0.4020***. p-value <0.01. coefficient. p-value. 0.0215. 0.88. coefficient. p-value. 0.4482***. <0.01. coefficient -0.0700. p-value. coefficient. 0.64. 0.3922***. p-value <0.01. Capital Loss. 0.1711**. 0.03. 0.0619. 0.38. 0.1840**. 0.02. 0.0990. 0.18. Bad Investment. 0.0553. 0.36. 0.0617. 0.26. 0.0431. 0.48. 0.0442. 0.42. Bad Health. 0.3000**. 0.03. 0.2293*. 0.08. 0.3140**. 0.03. 0.2398*. 0.06. Bankruptcy No Wage. -0.1063. 0.74. 0.2290***. 0.01. -0.3319 0.1533**. 0.26. -0.0940. 0.04. 0.2287*** -0.00458. Capital Gain. 0.77. -0.3276. 0.27. <0.01. 0.1565**. 0.04. 0.96. 0.1179. 0.11. 0.0554. 0.52. 0.1487*. 0.06. Wage. 0.1251**. 0.04. 0.0221. 0.69. 0.1298**. 0.03. 0.0252. 0.65. Investment. 0.2195***. 0.01. 0.1739**. 0.03. 0.2085**. 0.02. 0.1666**. 0.04. Age-2009. 0.8167***. <0.01. 1.7813***. <0.01. 0.8530***. <0.01. 1.7830***. <0.01. 0.8633***. <0.01. 1.7850***. <0.01. College-2009. 0.3092***. <0.01. 0.3471***. <0.01. 0.3051***. <0.01. 0.3353***. <0.01. 0.2953***. <0.01. 0.3286***. <0.01. High School-2009. -0.0919. 0.19. 0.1161**. 0.06. -0.0979. 0.17. 0.1133*. 0.07. -0.0928. 0.19. 0.1164*. 0.06. △Net Worth. -0.0294. 0.85. 0.1980. 0.17. -0.0470. 0.76. 0.1940. 0.18. -0.0416. 0.79. 0.1980. 0.17. △Income. 1.7385***. 0.01. 1.8642***. Dage-2009. 0.0438. 0.63. 0.1178. Dincome-2009. -0.1172. 0.14. Dwealth-2009. 0.0162. 0.83. Number of households. 3,298. <0.01. 1.6844***. <0.01. 0.14. 0.0533. 0.55. -0.3148***. <0.01. -0.1523*. 0.06. -0.2975***. <0.01. 0.0301. 0.67. 1.8181*** 0.1136. <0.01. 1.7232***. <0.01. 1.8288*** 0.1132. <0.01. 0.16. 0.0532. 0.55. 0.16. -0.3373***. <0.01. -0.1382*. 0.09. -0.3303***. <0.01. -0.3217***. <0.01. -0.00405. 0.96. -0.3400***. <0.01. 31.

(36) Appendix Appendix 1 X5711 P5711. X5712 P5712. QUESTION TEXT SAME AS 2009 VERSION Did you (or anyone else) have income or losses from net gains or losses from the sale of stocks, bonds, or real estate? IRS FORM 1040 LINE NUMBER: 13, 14 WE WANT TOTAL INCOME FOR THE YEAR, NOT MONTHLY INCOME. 1. *YES 5. *NO QUESTION TEXT SAME AS 2009 VERSION In total, what was your (family's) annual income from net gains or losses from mutual funds or from the sale of stocks, bonds, or real estate in 2008 [in 2007: 2006], before deductions for taxes and anything else? IRS FORM 1040 LINE NUMBER: 13, 14 Code amount -1. Nothing 0. Inap. (no capital gains or losses: X/P5711^=1). Appendix 2 P091401 P091402 P091403 P091404 P091405 P091406 P091407 P091408 P091409 P091410. Finally, I would like to ask about a few of your opinions. Over the past two years what have been the most important positive events for your (family's) finances? CODE ALL THAT APPLY. 1. 2. 3. 4. 5. 6. 9. 10. 11. 13. 14. 15. 16. 17. 20. 21. 22. 25. 26. 27. 28. 33. 34.. Got raise/promotion at work; got combat deployment pay Got increase in benefits at work Business is good Received inheritance/gift/scholarship/lottery winnings Received insurance/legal settlement Received or received increase in disability/unemployment/welfare payments Received or received increase in Social Security or pension payments Received income/support/help from others New family member received income Had higher income from assets Had higher income, n.f.s Income steady or little changed Had enough money to maintain lifestyle Had enough money to get by, n.f.s. Got/kept job Worked more New businesses/higher demand for workers in the area Purchased a home or other assets/investments Sold a home or other assets Rearranged assets at a good time; is able to make favorable investments Is able to save Assets retained some value Assets did not lose value 32.

(37) P091411 P091412 P091413 P091414 P091415 P091416 P091417 P091418 P091419 P091420. 35. Assets increased in value 36. "Market" increased 37. "Market" stabilized 38. "Market" decreased 39. Still have some assets, n.f.s. 44. Declared bankruptcy or obtained debt forgiveness 45. Able to get credit/refinance 46. Had no debt, paid off debt 47. Is able to pay down debt 48. Avoided going (further) into debt 49. Is able to keep or avoid foreclosure/repossession of an asset (house. car, business, etc.) 55. Little/no deterioration in finances 60. Traveled/made special purchases/expenditures other than assets 61. Had low expenses; expenses decreased n.e.c. 62. Prices (in general or in particular) did not rise, or went down 65. Economy did not collapse 66. Tax credit/rebate, "stimulus program," other one-time government program 67. Other fiscal policy; taxes or government spending n.f.s 68. Interest rates for borrowing low 69. Monetary policy or Federal Reserve n.f.s. 75. Present political situation 76. Past political situation 77. International economic situation; exchange rate 78. International political situation 85. Had good health; "still alive" 86. Had health insurance 87. Favorable legal decision 88. Got good advice/asset management 89. Completed education/training 90. Children independent/finished school 91. Other family events 92. Other personal events -1. Nothing, none 0. Inap. (/no further responses) Over the past two years what have been the most important negative events for your (family's) finances? CODE ALL THAT APPLY. 1. 2. 3. 4. 5. 8. 9. 10. 12. 13. 15. 16. 17. 20. 21. 22.. Salary/pay cut Decrease in benefits at work Business is bad, lost customers Lower or no government transfers/benefits No cost of living adjustment Lower income from/interest rates on assets Lower income, n.f.s Did not received child support/alimony/other payments owed to PEU Had to go on welfare Had to ask family/others for help Did not have enough to maintain lifestyle Did not have enough to get by, n.f.s. "No income" Lost/laid off from/could not find job Worked fewer hours, no overtime available Business closures/lower demand for workers in the area 33.

(38) 25. 26. 27. 30. 31. 35. 36. 37. 38. 39. 45. 46. 47. 48. 49. 50. 51. 60. 61. 62. 65. 66. 67.. Could not sell house/other assets Had to sell/borrow against assets Lost money on asset sale Foreclosure/repossession of an asset Bankruptcy Decline in stock prices Decline in real estate prices Decline in other asset values Lost money n.f.s. Bad investment n.f.s. Limited credit availability; unable to borrow Borrowed too much Higher interest expenses Mortgage reset to higher payments Prices (in general or in particular) rose; inflation Expenses increased n.f.s. "Bills" n.f.s. Unexpected repairs Had to make particular purchases Spent too much Financial/banking crisis "The economy," n.f.s. Tax credit/rebate, "stimulus program," other one-time government program. 68. 69. 70. 75. 76. 77. 78. 85. 86. 87. 88. 89. 90. 91. 92. 93.. Other fiscal policy; taxes or government spending n.f.s Monetary policy or Federal Reserve n.f.s. "The banks" n.f.s. Present political situation Past political situation International economic situation; exchange rate International political situation Had bad health Health care costs, lost health insurance Unfavorable legal decision Bad advice/asset management Had accident Natural disaster Robbery, identity theft, other crime Had to move Children or other family members returned home or required support; had to pay child support Children in college Divorce Death in family Other family events Other personal events Nothing, none Inap. (/no further responses). 94. 95. 96. 97. 98. -1. 0.. 34.

(39) Appendix 3 X3601 P3601. X091002. P091002. X091003. P091003. X091004. P091004. QUESTION TEXT SAME AS 2009 VERSION (SHOW CARD 7 [in 2007: CARD 10]) As we continue through the interview, I will be asking you about several types of retirement assets you may have, such as IRAs, annuities, and pensions and retirement accounts you may have through a current or past job. Here I would like to ask just about IRAs and Keogh accounts. These may include accounts that you "rolled over" into an IRA after leaving a previous job as well as Roth IRAs, or any other type of IRA or Keogh account that is not part of a retirement plan on a current or past job. Do you (or anyone in your family living here) have any Keoghs or IRAs? IF YES, SAY: Please do not include IRA-SEP or IRA-SIMPLE accounts, which we treat as job pensions. "EDUCATION IRAs" ARE SAVINGS ACCOUNTS. 1. *YES 5. *NO GENERATED FROM 2007 INDIVIDUAL VALUES: X3603/X3613/X3623 How many such accounts do you (or your family living here) have? Code number of accounts 0. Inap. (no IRA/Keogh accounts: X/P3601^=1) ********************************************************* ORIGINALLY ALLOWED VALUES: [1,...,99] IF < 1 OUT OF RANGE: ILLEGAL VALUE ERROR MESSAGE ********************************************************* FOR THE PUBLIC DATA SET, TOP-CODED AT 5 ********************************************************* GENERATED FROM 2007 INDIVIDUAL VALUES: X6551/X6559/X6567/X6552/X6560/X6568/X6553/X6561/X6569/ X6554/X6562/X6570 What is the (total) amount in (this account/all these accounts)? Code amount 0. Inap. (no IRA/Keogh accounts: X/P3601^=1) ********************************************************* ORIGINALLY ALLOWED VALUES: [1,...,999999999] IF < 1, OUT OF RANGE: ILLEGAL VALUE ERROR MESSAGE ********************************************************* GENERATED FROM 2007 INDIVIDUAL VALUES: X6555/X6563/X6571 How is the money in (this/these) account(s) invested? Is it all in stocks, all in interest-earning assets, is it split between these, or something else? IF R SAYS "MUTUAL FUND", PROBE FOR WHETHER IT IS A STOCK FUND, A BOND FUND, OR ONE SPLIT OVER BOTH TYPES 1. *ALL IN STOCKS 2. *ALL IN INTEREST EARNING ASSETS/BONDS 3. *SPLIT 4. Real estate 5. Hedge fund 6. Annuities 8. Mineral rights 9. GIC/guaranteed income contract 12. Business investment n.e.c. 35.

(40) X091005 P091005. 15. Life insurance 25. Non publicly traded business or other such investment 30. *MUTUAL FUND (NOT A PREFERRED RESPONSE) -7. *OTHER 0. Inap. (no IRA/Keogh accounts: X/P3601^=1) ********************************************************* FOR THE PUBLIC DATA SET, CODES 4, 5, 6, 8, 12, 15, AND 25 ARE COMBINED WITH CODE -7; CODE 9 IS COMBINED WITH CODE 2 ********************************************************* GENERATED FROM 2007 INDIVIDUAL VALUES: X7075/X7078/X7081/X7084/X7087/X7090 About what percent is in stocks? Code percent * 100 -1. Nothing 0. Inap. (no IRA/Keogh accounts: X/P3601^=1; holdings not split or in mutual funds: X/P091004^=3 or 30) ********************************************************* ORIGINALLY ALLOWED VALUES: [1,...,100] IF OUT OF RANGE: ILLEGAL VALUE ERROR MESSAGE ********************************************************* FOR THE PUBLIC DATA SET, ROUNDED TO NEAREST 500 IF BETWEEN 250 AND 9750 INCLUSIVE, OTHERWISE IF > 0 ROUNDED TO NEAREST 100 WITH A TOP-CODE AT 9900 AND A BOTTOM-CODE AT 100 FOR THOSE VALUES ROUNDED. 36.

(41) Appendix 4 X5729 P5729. QUESTION TEXT SAME AS 2009 VERSION What would be the correct total? How much is the total income you (and your family living here) received in 2008 [in 2007: 2006] from all sources, before taxes and other deductions are made? IF R SAYS TOTAL IS ZERO, ASK WHETHER THERE IS A LOSS OR IF THE AMOUNT IS ACTUALLY ZERO. Code amount -1. Nothing -9. Negative (public data set only) 0. Inap. (already have correct amount: X/P7361=1) NOTE: if X/P7361=1, the computed total is held in X/P5729. NOTE: in the public version of the data set, the X/P7361 may be YES, but the value of the income in X/P5729 may not be equal to X/P5702+X/P5704+X/P5706+X/P5708+X/P5710+X/P5712+ X/P5714+X/P5716+X/P5718+X/P5720+X/P5722+X/P5724 because of various operations applied to the data for disclosure avoidance. The details of this operation cannot be revealed, but a general description of the process is available in the papers cited in the introduction to this codebook. ********************************************************* ORIGINALLY ALLOWED VALUES: [-99999999,...,999999999] ********************************************************* FOR THE PUBLIC DATA SET, NEGATIVE VALUES ARE SET TO -9. Appendix 5 X3014 P3014. QUESTION TEXT SAME AS 2009 VERSION IN PERSON VERSION: (SHOW CARD 5 [in 2007: CARD 6]) Which of the statements on this page comes closest to the amount of financial risk that you (and your [husband/wife/partner]) are willing to take when you save or make investments? READ RESPONSES IF NECESSARY. TELEPHONE VERSION: Which of the following statements comes closest to describing the amount of financial risk that you (and your [husband/wife/partner]) are willing to take when you save or make investments? INTERVIEWER:. IF MORE THAN ONE RESPONSE IS GIVEN USE THE. FIRST CATEGORY THAT APPLIES. 1.. *Take substantial financial risks expecting to earn substantial returns. 2.. *Take above average financial risks expecting to earn above average returns. 3.. *Take average financial risks expecting to earn 37.

(42) average returns 4.. *Not willing to take any financial risks. NOTE: CARD 6 contains the following text in a vertical column: "Take substantial financial risks expecting to earn substantial returns," "Take above average financial risks expecting to earn above average returns," "Take average financial risks expecting to earn average returns," "Not willing to take any financial risks.". 38.

(43)