indicators of personal financial debt using a multi-disciplinary behavioral model

14
Journal of Economic Psychology 27 (2006) 543–556 www.elsevier.com/locate/joep 0167-4870/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.joep.2005.11.002 Indicators of personal Wnancial debt using a multi-disciplinary behavioral model Brice Stone 1 , Rosalinda Vasquez Maury ¤ Metrica Inc., 2 8620 N New Braunfels, Suite 603, San Antonio, TX 78217-6363, United States Received 5 June 2004; received in revised form 29 October 2005; accepted 30 November 2005 Available online 7 February 2006 Abstract Financial indebtedness has risen quite signiWcantly during the 1990s and early 2000s. This research eVort developed a multi-disciplinary model of consumer indebtedness used for early identiW- cation of US Air Force enlisted members most at risk for developing indebtedness. A survey was designed and administered to Wrst term enlisted Air Force personnel. A multi-faceted behavioral model of indebtedness was speciWed including demographic characteristics, institutional characteris- tics of the military, economic and Wnancial information, situational information about the events occurring in an individuals life, and psychological personality traits with respect to money attitudes, beliefs and behavior. The product of the analysis was a model of consumer indebtedness using per- sonal unsecured debt versus no debt. The estimated logit model exhibited strong predictive capabili- ties, as well as high percentages for sensitivity and selectivity of prediction. In addition, likelihood ratio tests strongly supported the thesis that indebtedness was a multi-faceted behavior which is com- prised of demographic/institutional, Wnancial, economic, psychological, and situational aspects, each critically important to the overall explanation of indebtedness behavior. © 2006 Elsevier B.V. All rights reserved. JEL classiWcation: A12; H63 PsycINFO classiWcation: 3800; 3920; 3120 Paper presented at the Proceedings of the SABE/IAREP 2004 Conference: Cross Fertilization between Economics and Psychology, Philadelphia, USA (July 15–18). * Corresponding author. Tel.: +1 210 822 2310x208; fax: +1 210 804 0836. E-mail addresses: [email protected] (B. Stone), [email protected] (R.V. Maury). 1 Tel./fax: +1 210 545 1028. 2 URL: http://www.metricanet.com.

Upload: brice-stone

Post on 26-Jun-2016

229 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Indicators of personal financial debt using a multi-disciplinary behavioral model

Journal of Economic Psychology 27 (2006) 543–556

www.elsevier.com/locate/joep

Indicators of personal Wnancial debt usinga multi-disciplinary behavioral model �

Brice Stone 1, Rosalinda Vasquez Maury ¤

Metrica Inc.,2 8620 N New Braunfels, Suite 603, San Antonio, TX 78217-6363, United States

Received 5 June 2004; received in revised form 29 October 2005; accepted 30 November 2005Available online 7 February 2006

Abstract

Financial indebtedness has risen quite signiWcantly during the 1990s and early 2000s. Thisresearch eVort developed a multi-disciplinary model of consumer indebtedness used for early identiW-cation of US Air Force enlisted members most at risk for developing indebtedness. A survey wasdesigned and administered to Wrst term enlisted Air Force personnel. A multi-faceted behavioralmodel of indebtedness was speciWed including demographic characteristics, institutional characteris-tics of the military, economic and Wnancial information, situational information about the eventsoccurring in an individuals life, and psychological personality traits with respect to money attitudes,beliefs and behavior. The product of the analysis was a model of consumer indebtedness using per-sonal unsecured debt versus no debt. The estimated logit model exhibited strong predictive capabili-ties, as well as high percentages for sensitivity and selectivity of prediction. In addition, likelihoodratio tests strongly supported the thesis that indebtedness was a multi-faceted behavior which is com-prised of demographic/institutional, Wnancial, economic, psychological, and situational aspects, eachcritically important to the overall explanation of indebtedness behavior.© 2006 Elsevier B.V. All rights reserved.

JEL classiWcation: A12; H63

PsycINFO classiWcation: 3800; 3920; 3120

� Paper presented at the Proceedings of the SABE/IAREP 2004 Conference: Cross Fertilization betweenEconomics and Psychology, Philadelphia, USA (July 15–18).

* Corresponding author. Tel.: +1 210 822 2310x208; fax: +1 210 804 0836.E-mail addresses: [email protected] (B. Stone), [email protected] (R.V. Maury).

1 Tel./fax: +1 210 545 1028.2 URL: http://www.metricanet.com.

0167-4870/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.joep.2005.11.002

Page 2: Indicators of personal financial debt using a multi-disciplinary behavioral model

544 B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556

Keywords: Economic psychology; Consumer debt; Credit; Model; Attitudes towards money; Factors inXuencingWnancial mismanagement or indebtedness

1. Introduction

The amount of revolving credit in the US averaged about $692 billion in 2001 (FederalReserve Board, 2003) versus $475 billion in 1996, an increase of over 45% in 5 years.Revolving credit is a type of open-end credit. By the end of 2002, the amount of consumerrevolving credit had risen to over $716 billion dollars. During this same time period (1996–2001), the gross domestic product rose about 22%, disposable personal income rose byover 30%, and civilian unemployment rates Xuctuated from 5.4% in 1996 down to 4% in2000 and back up to 5.78% by 2002. Consumer indebtedness is an increasing national issue,especially with high civilian unemployment rates.

The focus of this research was to create an indebtedness model to be used for early iden-tiWcation of US Air Force members most at risk for developing personal Wnancial manage-ment problems. Recent data shows that active duty Air Force E-3s, E-4s and E-5s (whichare identiWers assigned to personnel that is tied to pay ranges within the military salarystructure) are experiencing the greatest Wnancial diYculty and our exhibiting high levels ofunsecured debt and Wnancial mismanagement behaviors (i.e. bounced checks, late pay-ments to creditors) (Air Force Financial Status Survey, 2001; Defense Manpower DataCenter (DMDC) Survey of Active Duty Personnel, 1999). Much like their civilian peers,and members of other military services, 20–32 year old Air Force personnel appear to usean extensive amount of personal unsecured long-term debt and have a high prevalence ofWnancial mismanagement behaviors.

The magnitude of the impact of these indebtedness behaviors on personal preparednessis unknown, but exploratory research with civilian subjects has identiWed associationsbetween decreased worker productivity and Wnancial mismanagement behaviors (Garman,Kim, Kratzer, Brunson, & Joo, 1999; Garman, Leech, & Grable, 1996; Joo & Garman,1998; Luther, Garman, Leech, GriYtt, & Gilroy, 1997; Williams, Haldeman, & Cramer,1996).

2. Literature review of Wnancial indebtedness or mismanagement

As an initial step in developing an indebtedness model, a review of the existing literatureconcerning the modeling of Wnancial indebtedness was performed. The Wnancial indebted-ness or mismanagement behavior literature crosses several disciplines such as psychology,economics, sociology, and political science. For developing and specifying an indebtednessmodel, the literature from psychology, economics, sociology and combinations thereofprovided the most useful research base. One essential characteristic of an indebtednessmodel would be its ability to accurately identify individuals who are at risk for developingpersonal Wnancial management problems. Proper speciWcation of such a model is the Wrststep to a healthy predictive model. Obviously, the deWnition of what represents Wnancialindebtedness has numerous options which have been analyzed and reviewed in the litera-ture, such as total debt, total debt excluding home mortgage, debt to income ratio, etc. Inaddition, the list of explanatory variables is rather daunting such as gender, ethnicity,

Page 3: Indicators of personal financial debt using a multi-disciplinary behavioral model

B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556 545

education, family history, income, number of credit cards, use of credit cards, use of debt,and numerous psychological indices such as locus of control, self-esteem, sensation seek-ing, money attitudes, self-eYcacy and decision making under risk (e.g., Livingstone & Lunt,1992; Tokunaga, 1993).

Financial indebtedness has risen quite signiWcantly during the 1990s and early 2000s. Asindicated earlier, the amount of revolving credit in the US averaged about $692 billion in2001 (Federal Reserve Board, 2003) the result of a more than 30% increase in just over5 years. Credit card use, auto loans, home mortgages, student loans, equity loans, etc. havebecome common components of the personal budget in the United States, though othercountries such as the United Kingdom, are also exhibiting similar trends. Researchers con-tribute these spending and borrowing trends to numerous demographic, economic psycho-logical, and situational factors.

Though indebtedness may have become a primary medium of exchange for consumerbehavior, recent concern and research has been directed at topics such as shoplifting, com-pulsive gambling and compulsive buying as these types of aberrant consumer behaviorhave become increasingly evident in society (Faber & O’Guinn, 1992; Friese & Koenig,1993; Hanley & Wilhelm, 1992). Most analyses in these areas have been directed at psycho-logical attributes or personality traits such as self-esteem, sensation seeking, risk loving,money attitudes, etc. Thus, indebtedness of one form or another may be the norm for themajority of young enlisted personnel (people who enter the armed forces voluntarily) butexcessive indebtedness can become an impediment to dependable productivity and missionreadiness. For an indebtedness model to be a good predictor of individuals with excessiveindebtedness or Wnancial mismanagement, the model must be well speciWed with thosedemographic, economic, psychological, and situational factors which are most closelyassociated with indebtedness or Wnancial mismanagement. The model building literaturefor indebtedness or Wnancial mismanagement is relatively recent (30 years) and in mostcases focuses on a few or a subset of demographic, economic, psychological, and situa-tional factors, thus, limiting its predictive capability, as well as its generalizability.

Livingstone and Lunt (1992) state that “While many factors inXuencing personal debthave been proposed, no clear conceptual model which integrates these has yet emerged” (p.114). One of the primary reasons for this conclusion is the lack of availability of data thatpossesses suYcient demographic, economic, psychological, or situational factors to fullyspecify an explanatory model. The authors collected information that included but not lim-ited to demographic variables, economic variables, enduring psychological variables, eco-nomic attributions and behaviors, and satisfaction in life. The data was collected in 1989resulting in a sample of 219 respondents.

Livingstone and Lunt (1992) used discriminant and regression analysis, both resultingin highly statistically signiWcant functions. The estimated discriminant function correctlyclassiWed respondents as either in debt or not in debt about 95% of the time, in-sample.Table 1 in the Livingstone and Lunt paper presents those explanatory variables that corre-lated above a cutoV value of 0.1 with the discriminant function. Two regression analyseswere performed. The Wrst regression analysis, which was performed on respondents withpositive debt, explained 66% of the total variance in the amount of total debt. The secondregression analysis was performed on respondents with regular debt payments, explaining60% of the total variance in the amount of regular debt payments. In each instance, thestatistically signiWcant explanatory variables were a combination of demographic (veryfew), economic and psychological variables (see Table 1). The important aspect of the

Page 4: Indicators of personal financial debt using a multi-disciplinary behavioral model

546 B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556

Livingstone and Lunt (1992) study is the attempt to develop a fully speciWed model ofindebtedness encompassing demographic, economic and psychological variables. As aresult of this multi-disciplinary speciWcation, the in-sample predictive accuracy of the esti-mated functions was moderate to high.

Castellani and DeVaney (2001) and Kim and DeVaney (2001) used a data base spon-sored by the Federal Reserve Board of Governors the Survey of Consumer Finances(SCF). Castellani and DeVaney (2001) used the 1995 SCF and Kim and DeVaney (2001)used the 1998 SCF. Both attempting to model varying forms of credit card use. Castellaniand DeVaney used a binary dependent variable representing whether the respondentthought it was right to borrow money to cover living expenses when income is cut andlogistic regression. Kim and DeVaney used a binary variable representing whether therespondent had an outstanding credit card balance and probit, as well as a continuous var-iable representing the amount of outstanding credit card balances and regression analysis.In both, the models exhibited statistical signiWcance for only a few of the explanatory vari-ables (see Table 2).

Table 1Statistically signiWcant (95% or better) variable list

General category SpeciWc variables

Demographic variables Social class (positive debt, only)Partner’s social class (positive debt, only)

Economic variables Disposable income (both equations)Number of debts (positive debt, only)Total amount of debt (amount of regular debt repayment, only)

Enduring psychological variables People get respect they deserve (disagree) (positive debt, only)Believe credit useful but complicated (positive debt, only)General coping – less cool and calm (positive debt, only)Value achievement over social concern (amount ofregular debt repayment, only)Attitudes pro-credit rather than anti-debt (amount of regulardebt repayment, only)

Economic attribution Important to keep up with the Jones’s (disagree)(positive debt, only)Blame external disasters rather than hedonism(amount of regular debt repayment, only)

Satisfaction in life SatisWed if no change in own standard of living(amount of regular debt repayment, only)

Economic behavior Reward self with purchase (disagree) (positive debt, only)Think about money (both equations)Willing to use credit (positive debt, only)Number of bank accounts (positive debt, only)Enjoy shopping for clothes (disagree) (positive debt, only)Shop in favorite shops (disagree) (positive debt, only)Pay total oV credit card each monthEnjoy shopping with family (disagree)(amount of regular debt repayment, only)

Page 5: Indicators of personal financial debt using a multi-disciplinary behavioral model

B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556 547

The adjusted R-square for the regression analysis was 0.3013, which implies that onlyabout 30% of the variation in outstanding credit card balances was explained by the modelspeciWcation. This is signiWcantly lower than the Livingstone and Lunt (1992) results, how-ever Livingstone and Lunt had only 107 respondents in their sample whereas Kim andDeVaney (2001) had a sample of 1500. In addition, Livingstone and Lunt had a broaderarray of demographic, economic and psychological variables though Kim and DeVaneyincluded a larger number of economic factors in their model speciWcation.

Tokunaga (1993) studied the use and abuse of consumer credit that focused on behav-ioral, psychological, and situational factors as diVerentiators between individuals who hadexperienced serious problems in the use of credit cards or those who had not. Three pri-mary sets of explanatory variables were used: background characteristics, psychologicalvariables, and adverse life events (presented in Table 3).

Though many of the explanatory variables were statistically signiWcant, the predictivecapability of the estimated discriminant function was limited. Of the 61% of the respon-dents (120 total respondents) which were correctly classiWed, only 26% of the individualswho had experienced serious problems in the use of credit cards were predicted correctly(16 of 62) while 98% of the control group (no serious problems) were predicted correctly(57 of 58). Though the psychological variables were generally statistically signiWcant, theydid not provide enough of a diVerentiation to do a good job of predicting individuals whohad experienced serious problems in the use of credit cards. This suggests that family

Table 2Statistical results for Castellani and DeVaney (2001) and Kim and DeVaney (2001)

Question SpeciWc variables

Is it right to borrow money to cover livingexpenses when income is cut?(1995 SCF, observations D 4299)

Age – less than 35, 35–45 and 45–54WhiteHousehold income – less than $10,000, $10,000–$19,999Payment pattern – late payment

Credit card users (1998 SCF, observations D 3376) Age and age squaredEducationNatural logarithm of incomeNatural logarithm of liquid assetsNatural logarithm of investment assetsNatural logarithm of real assetsNatural logarithm of debtNumber of credit cardsCredit limitAttitude toward credit – positive and ambivalentAttitude toward the use of credit for vacationTime horizon – 5–10 years and above 10 years

Outstanding credit card balances (1998 SCF,observations D 1500)

EducationNatural logarithm of incomeNatural logarithm of real assetsNumber of credit cardsCredit card limitCredit card interest rateAttitude toward credit – positive and ambivalentPayment habit – behind schedule or miss payment

Page 6: Indicators of personal financial debt using a multi-disciplinary behavioral model

548 B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556

background and psychological variables alone are not suYcient to develop a reliable pre-dictive model, i.e., the absence of economic variables makes a non-trivial diVerence. AsTokunaga (1993) indicates, beginning to understand one’s ability, motivation and desire tospend money is a critical step in gaining control over one’s Wnancial situation. Thus, it is theachievement of control that is the foundation of successful personal Wnancial management.

College students’ debt has recently received increased visibility from researchers.Davies and Lea (1995), for example, looked at US students studying at the University ofExeter in the United Kingdom. They compared debt levels of Wrst-year, second-year, andthird-year students and found that the tolerance of debt (more positive attitude towardindebtedness) increased with increasing time at university. Thus indebtedness can changeattitudes over time as well as impact health. Roberts et al. (2000) examined the relation-ship between the physical, social and psychological health of students and their Wnancialcircumstances. They found that Wnancial circumstances of students had an adverse impacton their health.

Several psychological indices were identiWed in the literature review, which were associ-ated with Wnancial indebtedness or mismanagement, each comprised questions concerningbehavior, attitude or expectations. Tokunaga (1993) studied some of the key psychologicalvariables/scales cited and used in the literature: locus of control (Levenson, 1981); self-eYcacy (Sherer et al., 1982); self-esteem (Rosenberg, 1989); attitude towards money(Yamauchi & Templer, 1982); decision making (Hershey & Schoemaker, 1980); and sensa-tion seeking (Zuckerman, 1979).

The money beliefs and behavior scale (MBBS) developed by Furnham (1984) encom-passes many aspects of the psychological scales previously mentioned. MBBS is comprisedby six factors: Obsession (reXects preoccupation with superiority, the use of money forcontrol or comparison, and the idea that money serves as a major solution for any prob-lem); Power-Spending (represents the intensity of the need to spend money in ways whichreXect status); Retention (reXects an extremely cautious use of money and insecurity inspending); Security-Conservative (reXects a very traditional, old-fashioned approach to theuse of money); Inadequacy (reXects the feeling that one does not have enough money, espe-cially as compared to friends); and EVort/Ability (measures the extent to which one feels

Table 3Variable types

General description SpeciWc variables

Background characteristics Parents’ use of forms of creditParents’ view of creditFamily Wnancial situation (insigniWcant)

Number of adverse life events in the last 12 months Such as medical bills, marital problems, and work-related problems on Wnancial diYculties

Psychological variables Attitude towards money – retentionDecision making – gain situationsAttitude towards money – powerSensation seeking – total scoreAttitude towards money – anxietySelf-eYcacyLocus of control – internal controlDecision making – loss situations (insigniWcant)

Page 7: Indicators of personal financial debt using a multi-disciplinary behavioral model

B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556 549

that money earned reXects eVort and ability). The MBBS originally consisted of 60 itemsscored on a 7-point agree/disagree scale. Factor analysis conducted by Furnham (1984)resulted in 47 items loading on the six distinct factors presented above. Furnham con-cluded that people’s attitudes toward, and habits of, money usage are complex and multi-faceted and that there were a number of variables such as demographics and social beliefvariables which discriminate between people’s beliefs and behaviors.

Other researchers have studied the validity and generalizability of the items and factors.One such study was performed by Bailey and Gustafson (1991). They found that 27 of the60 items loaded signiWcantly on three factors instead of six factors (the three factors wereObsession, Inadequacy, and Retention). With these 27 items, Bailey and Gustafson found arelationship between an individual’s personality and their attitude towards money. Hayhoeand Leach (1997) found that 30 of the items loaded signiWcantly on Wve factors instead ofsix factors (the Wve factors were Obsession, Retention, EVort/Ability, Security, and Inade-quacy). Hayhoe, Leach, and Turner (1999) later used these 30 items and found that moneyattitudes can be used as a discriminator between the use/non-use of credit cards. Wilhelmand Varcoe (1991) used a modiWed version of Furnham’s scale to study how money atti-tudes inXuence individuals’ perceptions of their economic well being. They found themoney attitudes of inadequacy and eVort/ability were signiWcant predictors of Wnancialsatisfaction for men and women.

The studies mentioned are only a few of the works in this area of the literature but pro-vided a baseline against which the survey instrument for the data was designed and theestimated indebtedness model was speciWed, estimated and compared. Few studies existwhich have attempted to account for a demographic, economic, psychological, and situa-tional factors in explaining the variation in total debt or classiWcation of individuals withrespect to their credit card habits (Livingstone & Lunt, 1992). A database that includes aextensive list of demographic, economic, psychological, and situational factors, as well assuYcient information to deWne dependent variables representing Wnancial indebtedness,does not exist in the literature. The literature review does provide a basis for the deWnitionof alternative dependent variables and an array of independent variables that are catego-rized as demographic, economic, psychological, and situational factors.

3. Method – survey instrument design and data collection

The literature review was used as the basis for designing a survey instrument for collectinginformation on individual demographic, economic, sociological, psychological, and situa-tional factors. The questions was created from a combination of diVerent surveys (more spe-ciWcally Air Force Community Needs Assessment Survey, 2000; Air Force Financial StatusSurvey, 2001; Survey of Consumer Finances, 1995, 1998; Survey of Active Duty Personnel,1999). The survey also included the Money Beliefs and Behavior Scale (MBBS) (Furnham,1984). Respondents to the survey were provided by the Technical Training Wing andFirst Term Airman Center (FTAC) at Lackland Air Force Base (AFB), San Antonio, TX.Data collection occurred from March through July of 2003. The survey data was collectedthrough MTISurv, a software package developed by Metrica, Inc. (2001) that was designedfor data collection through the Internet. Respondents who did not complete the entire surveywere not used in the analysis. A total of 501 enlisted personnel completed the survey duringthe early stages of their Wrst term of enlistment (act of entering the armed forces voluntarilyand typically last for about 4years, but there is a 6 year Wrst term enlistment program).

Page 8: Indicators of personal financial debt using a multi-disciplinary behavioral model

550 B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556

4. Results

4.1. Demographics

Of the 501 respondents, 379 (75.7%) were male, 309 (61.7%) were Caucasian, 94 (18.8%)were African American, 61 (12.2%) were Hispanic (e.g., Spanish, Mexican, Puerto Rican,etc.), 312 (62.3%) were high school graduates, 68 (13.6%) had some college but less than1 year, 96 (19.2%) had one or more years of college but no degree, 51 (10.7%) were marriedand 44 (8.8%) had children. Of the 501 respondents, 459 (91.6%) were in initial training, 457(91.2%) were serving on active duty while the remaining 44 (8.8%) were members of thereserve components; 415 (82.8%) lived in Government quarters and 25 (5.0%) owned,rented or leased civilian housing; and 487 (97.2%) were in their Wrst year of service.

4.2. Income

The key focus of the data collection eVort was to collect detailed information about theindividual’s Wnancial status with respect to their income and debt. A series of questionswere directed at debt and income. One characteristic of the sample population that wasexpected to aVect the responses to the debt/income questions was the young age of thesample population. The average age for the sample population was slightly less than21 years of age (standard deviation of less than 2.67) with 95% of the respondents less than25 years old. The youthfulness of the respondents aVects the level of debt that they are ableto accumulate in such an abbreviated work history. Respondents that indicated that theywere very clueless about their estimate of their total debt were excluded from the analysis(22 respondents). Only active duty personnel were included in the following analysis (44reservists were excluded from the analyst because they were not within the scope of thisproject). Thus, the sample used in the analysis is comprised of 438 respondents.

The average gross (before-tax) household income is approximately $1936 per month(standard deviation of $2025) or about $23,232 annually. Averages for income are interpola-tions because all money valued questions were provided in ranges (in $500 increments withexcept of the Wrst and last). For example, total income was divided into 16 possible monthlyranges: $1000 or less, $1001–$1500, $1501–$2000, ƒ, $7001–$7500, $7501–$8000, and $8000and above. Thus, a mean of 4.25 for income is presented as $2650 to provide the reader witha value which is easier to interpret. Of the 49 respondents who reported a second job withpositive gross monthly earnings, 82% (40) report less than $1500 per month from the secondjob. Of the 49 respondents who reported a second job with positive gross monthly earnings,30 (61%) worked less than 20 h per week. Of the 438 respondents, only 31 reported spouses(or “signiWcant others”) with positive gross monthly incomes (only 50 respondents reportedbeing married or having a signiWcant other). Twenty-six (84%) of those reporting spouseswith positive gross monthly incomes reported incomes less $2000 per month.

4.3. Indebtedness

Numerous questions were directed at the debt issue from the perspective of the level ofdebt and the perspective of how the debt was accumulated (e.g., credit cards, store cards,auto and home loans). These series of debt related questions provide for a number of alter-native measures of debt. For example, debt can be measured as a continuous variable, an

Page 9: Indicators of personal financial debt using a multi-disciplinary behavioral model

B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556 551

ordinal variable or a dichotomous variable. The focus of this research eVort is to develop amodel that can identify individuals who exhibit a high propensity for excessive debt levels,regardless of the deWnition of debt or excessive debt.

The survey provided several questions that elicited a series of estimations concerningthe level of various types of debt. One such question was concerned with the amount ofpersonal unsecured debt: (including: cards (credit, store, gasoline, government); loans (stu-dent/personal); etc., and excluding: home mortgage and car loans). The average personalunsecured debt was $910.96 (standard deviation of $2245.05) with 295 (67.4%) respondentsreporting no unsecured personal debt. Means for all debt values are provided based onmidpoints assigned for each range of debt. For example, the range $1–$1000 is assigned avalue of $500 for each respondent that selected the $1–$1000 range. Of the 438 respon-dents, 347 (79.2%) reported $1000 or less unsecured personal debt while over 92% reported$5000 or less unsecured debt. Only 16 respondents (3.7%) reported personal unsecureddebt of more than $10,000. Thus, debt excluding home mortgage and car loans is not anobvious issue with most of the respondents in the sample.

Another component of the accumulation of debt in our modern society is the creditcard. Once again, the survey elicited information concerning credit card possession anduse, as well as other types of cards such as debit cards, store cards, gasoline cards, etc. Withthe exception of the debit card, the majority of all the respondents did not own or possess acredit, store, gasoline or government card. The credit card was the most often possessedcard, with 165 (37.7%) of the respondents indicating that they owned one or more creditcards. The absence of credit, store, gasoline and Government cards may be the result of theyouth of the sample population, i.e., this young population has not had suYcient time toaccumulate money cards. Of the 438 active duty respondents, 263 (60.1%) possessed one ormore types of debit, credit, store, gasoline or government cards. Only 6 (1.4%) of the sam-ple possess credit, store, and gasoline cards; 8 (1.83%) possess credit and gasoline cards;and 68 (15.5%) possess credit and store cards.

4.4. Indebtedness model analysis

The primary objective of the data collection eVort was to assemble demographic, psy-chological, situational, and Wnancial information on a population of enlisted personnelthat would allow a thorough analysis of indebtedness and the factors aVecting the level ofindebtedness among junior enlisted personnel. The survey was carefully designed to ensurethat all types of information were collected from each respondent.

4.5. Logit model

We begin by specifying a logit model (zero/one dichotomous dependent variable) usingunsecured personal debt as the dependent variable (143 of 438 respondents had unsecureddebt, 32.65% of the sample). Unsecured personal debt includes debt from cards (credit, store,gasoline, government); loans (student/personal); etc., and excluding: home mortgage and carloans. Two types of data screening were performed in order to: a Wrst order analysis toremove those variables (from the initial list of 76 variables) that have no Wrst order relation-ship with the dependent variable and data condensation of the MBBS items using Baileyand Gustafson’s (1991) factor loadings. Table 4 provides a list of the remaining 19 indepen-dent variables used in the model speciWcation. Most of these independent variables can be

Page 10: Indicators of personal financial debt using a multi-disciplinary behavioral model

552 B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556

found in the literature, many in more than one study. Bailey and Gustafson’s (1991) factorloadings were used because the loadings did a better job of discriminating between debtorsand non-debtors versus the six factor loadings of Furnham (1984) and Hayhoe et al. (1999).

The estimated logit model displayed a pseudo R-square of 0.3329, i.e., explaining over33% of the variation in the dependent variable and predicted 87 of the 143 of the respon-dents exhibiting unsecured personal debt, 60.84% sensitivity. In addition, the model pre-dicts 266 of the 295 respondents who do not exhibit unsecured personal debt (90.17%speciWcity) for a total predictive accuracy (in-sample) of 80.59%. This is a model speciWca-tion which satisWes the ability to predict well in sample, does not predict many respondentsto be included in the group with unsecured personal debt who are actually not in the group(56 of 143, 39.16%) and misses less than 25% of the respondents who are not predicted tobe in the group with unsecured personal debt who are actually are in the group (29 of 116,25.00%). Of the 19 explanatory variables speciWed in the model, Wve were statistically sig-niWcant at the 0.01 signiWcance level and Wve variables were statistically signiWcant at the0.05 signiWcance level (these results are proved in Table 5).

Table 4Statistically signiWcant stepwise regression – unsecured personal debt

Variable description Odds ratio z-Statistic P > �z�Demographic/economicAge 1.1362 2.10 0.036Parents’ attitude toward credit card use 1.4099 2.56 0.010Number of credit and store cards 1.5858 4.65 0.000Rank (pay grade) 3.1794 2.47 0.014Rank (pay grade) 1.6688 1.27 0.204Perceived Wnancial condition 1.6255 2.81 0.005Own a vehicle 1.4954 1.38 0.168

Money beliefs and behavior scaleBailey average of factor 1:

obsession (number of items 10)0.7388 ¡1.76 0.079

Bailey average of factor 2:inadequcy (number of items 12)

1.5940 2.40 0.017

Bailey average of factor 3:retention (number of items 5)

0.7693 ¡1.95 0.051

Life altering eventsBirth of a child(ren)/adoption 0.2012 ¡2.65 0.008Relocation/PCS 3.8409 1.84 0.066

Financial events of respondentReached the maximum limit on a credit card 4.0519 2.32 0.020None of the above 0.4185 ¡2.91 0.004

Financial events of parentsStressed over Wnancial diYculty 0.6128 ¡1.58 0.115Was unable to aVord needed medical care 3.1523 1.79 0.074

Relieve stressGambling 3.0581 1.83 0.067Push/shove/punch/throw something 0.4460 ¡1.77 0.076Spend time alone/pray/meditate 1.7642 2.03 0.042

Page 11: Indicators of personal financial debt using a multi-disciplinary behavioral model

B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556 553

To better understand the contribution of each group of explanatory variables (demo-graphic/economic, MBBS, Wnancial and life altering events, etc.), an analysis was per-formed which investigated the groups of variables presented in Table 5 and a series oflikelihood ratio tests (LR Test) were calculated (Lea, Webley, & Levine, 1993; Lea, Webley,& Walker, 1995). One group included demographic/economic explanatory variables whichproduced a pseudo R-square of 0.2043, predicting 60 of the 143 respondents exhibitingunsecured personal debt, 41.96% sensitivity, and 270 of the 295 respondents exhibiting nounsecured personal debt, 91.53% selectivity. Regardless, the LR Test provided a value of71.16 which is statistically diVerent from zero at the 0.0000 level, indicating that the contri-bution of the demographic/economic explanatory variables to the full model speciWcationwas statistically signiWcant. Similar LR Tests were calculated for three groups of variables:non-demographic/economic variables, Bailey and Gustafson’s (1991) MBBS loadings, andnon-demographic/economic model excluding Bailey and Gustafson’s (1991) MBBS load-ings. The results are presented in Table 5. In each case the groups of independent variablesexhibited lower pseudo R-squares and weaker prediction but made a statistically signiW-cant contribution to the explanatory power of the fully speciWed model. Thus, each groupof variables of the fully speciWed model exhibited diYculty discriminating between the twodebtor groups, debt and no debt, but together they formed a strong predictive model whichdoes discriminate well between the two groups. The full speciWed model of indebtednesswas the best Wt to the data, pseudo R-square and prediction, conWrming Furnham’s (1984)conclusion that the nature of people’s attitudes toward, and habits of, money usage werecomplex and multifaceted, requiring an indebtedness model capturing demographic, eco-nomic and psychological aspects of the individual.

5. Summary and conclusions

As noted earlier, one of the key conclusions stated by Furnham (1984) was that thenature of people’s attitudes toward, and habits of, money usage was complex and multifac-eted. Furnham (1984), as well as others in the literature (Castellani & DeVaney, 2001;Davies & Lea, 1995; Livingstone & Lunt, 1992; Roberts et al., 2000), deduced that demo-graphics, as well as other factors, could be discriminators among people’s beliefs andbehaviors. The results of the model estimation of this research eVort deWnitely supports thethesis that demographics, as well as Wnancial, institutional, psychological, and situationalfactors, are discriminators among people’s attitudes toward, and habits of, money usage.

Table 5Full and restricted model speciWcations – unsecured personal debt

Model deWnition R-Square Percent LR Test

Sensitivity SpeciWcity Overall

Full model 0.3329 60.84 90.17 84.70Demographic/economic 0.2383 47.55 90.51 76.26 128.17Non-demographic/economic 0.2043 41.96 91.53 81.28 76.33MBBS 0.0406 11.89 97.63 70.78 214.79Events, stress, parents and

respondents Wnancial events0.1822 34.27 95.59 78.08 131.38

Debtors (nD 143) 32.65Non-debtors (n D 295) 67.35

Page 12: Indicators of personal financial debt using a multi-disciplinary behavioral model

554 B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556

In addition, many of the statistically signiWcant factors in the estimated model have beenidentiWed in the literature as important to explaining the variations of indebtedness such asmajor life altering events, MBBS items, income, age, etc. However, given the homogeneityof the demographics of the respondent sample with respect to age, education and occupa-tion (self-selected US Air Force enlisted personnel) some of the factors which were or werenot statistically related to the constructed measure of indebtedness may become more orless important for older and more educated groups.

The existing literature was not aVorded the opportunity to analyze and study such adiversiWed and detailed set of relevant factors assembled in a single data base. The modelderived to identify the factors which discriminate between respondents with and withoutunsecured personal debt represent demographic (such as age, sex, race/ethnicity), psycho-logical (MBBS), sociological (such as parents attitudes toward credit card use, size of city,home of record and credit card use), economic aspects of the individual’s proWle (totalincome, Wnancial activities, Wnancial events), and situational aspects (life altering events inthe last 12 months). Analysis indicates that no single set of characteristics (economic, psy-chological, situational, institutional, etc.) is suYcient to explain the indebtedness of thesample, but once combined into a single explanatory equation, representing the complexand multifaceted nature of people’s attitudes toward, and habits of, money usage, the abil-ity to predict outcomes increases dramatically.

Acknowledgements

The authors would like to thank the United States Air Force for supporting this study. Avery special recognition goes to Major Carl Miller, Chief of Air Force Family Research, forhis involvement and support throughout the study. We also would like to express outappreciation to Major Jim Whitworth, who originally drafted the statement of work andconducted the Financial Status Survey which helped inform the current eVorts; Dr. PaulDiTullio (HQ USAF/DPFPT) for his support and enthusiasm and Mrs. Tina Strickland(AFPOA) for her continued support; Shontelle Rivers; Ms. Barbra Murray, executive direc-tor of the CAIB who authorized the study, and the CAIB Chair; Major General Wherle;and Linda Smith who ensured the continued funding of the project last summer whenthings were almost bottoming out. The authors would also like to extend their gratitude fora job well done to the 37th Training Group (AETC) at Lackland AFB, San Antonio, Texas.A special appreciation goes to Colonel Peter Micale, the Group Commander, and CynthiaWhite, the Chief of Faculty Development Flight, for their support and participation in thedata collection portion of the study. The conclusions and opinions expressed in this docu-ment are those of the authors. These conclusions and opinions do not reXect the oYcialposition of the US Government, Department of Defense, or the United States Air Force.

References

Air Force Community Needs Assessment Survey (2000). Collects information on the Air Force community regard-ing many family and community issues [Data Wle]. Fairfax, VA: Caliber Associate.

Air Force Financial Status Survey (2001). Review the Wnancial status of active duty Air Force members [Data Wle].Arlington, VA: Air Force Financial Hardship Integrated Process Team.

Bailey, W. C., & Gustafson, A. W. (1991). An examination of the relationship between personality factors and atti-tudes to money. In R. Frantz, H. Singh, & J. Gerber (Eds.), Handbook of behavioral economics (pp. 271–285).Greenwich, CT: JAI Press.

Page 13: Indicators of personal financial debt using a multi-disciplinary behavioral model

B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556 555

Castellani, G., & DeVaney, S. A. (2001). Using credit to cover living expenses: a proWle of a potentially riskybehavior. Family Economics and Nutrition Review, 13, 12–20.

Davies, E., & Lea, S. E. G. (1995). Student attitudes to student debt. Journal of Economic Psychology, 16, 663–679.Faber, R. J., & O’Guinn, T. C. (1992). A clinical screener for compulsive buying. Journal of Consumer Research,

19, 459–469.Federal Reserve Board (2003). Statistical release G.19: consumer credit. Federal Reserve Bulletin(Retrieved Sep-

tember 8) Available from http://www.federalreserve.gov/releases/G19/Current.Friese, S., & Koenig, H. (1993). Shopping for trouble. Advancing the Consumer Interest, 5, 24–29.Furnham, A. (1984). Many sides of the coin: the psychology of money usage. Personality and Individual DiVer-

ences, 5, 501–509.Garman, E. T., Kim, J., Kratzer, C. Y., Brunson, B. H., & Joo, S. (1999). Workplace Wnancial education improves

personal Wnancial wellness. Financial Counseling and Planning, 10(1), 79–88.Garman, E. T., Leech, I. E., & Grable, J. E. (1996). The negative impact of employee poor personal Wnancial

behaviors on employers. Financial Counseling and Planning, 7, 157–168.Hanley, A., & Wilhelm, M. S. (1992). Compulsive buying: an exploration into self-esteem and money attitudes.

Journal of Economic Psychology, 13, 5–18.Hayhoe, C. R., Leach, L., & Turner, P. R. (1999). Discriminating the number of credit cards held by college stu-

dents using credit and money attitude. Journal of Economic Psychology, 20, 643–656.Hayhoe, C. R., & Leach, L. J. (1997). An exploration of college students’ credit use. A poster session at the annual

meeting of the American Association of Family and Consumer Sciences, Washington, DC, 1997.Hershey, J. C., & Schoemaker, P. J. (1980). Prospect theory’s reXection hypothesis: a critical examination. Organi-

zational Behavior and Human Performance, 25, 395–418.Joo, S., & Garman, E. T. (1998). The potential eVects of workplace Wnancial education based on the relationship

between personal Wnancial wellness and worker job productivity. In E. T. Garman, S. Joo, I. E. Leech, & D. C.Bagwell (Eds.), Personal Wnances and worker productivity, Proceedings of the personal Wnance employee educa-tion best practices and collaborations conference, Roanoke, VA (Vol. 2 (2), pp. 163–174). Blacksburg, VA:PFEE, Virginia Tech.

Kim, H., & DeVaney, S. A. (2001). The determinants of outstanding balances among credit card revolvers. Finan-cial Counseling and Planning, 12(1), 67–77.

Lea, S. E. G., Webley, P., & Levine, R. M. (1993). The economic psychology of consumer debt. Journal of Eco-nomic Psychology, 14, 85–119.

Lea, S. E. G., Webley, P., & Walker, C. M. (1995). Psychological factors in consumer debt: money management,economic socialization, and credit use. Journal of Economic Psychology, 16, 681–701.

Levenson, H. (1981). DiVerentiating among internally, powerful others and chance. In H. M. Lefcourt (Ed.),Research with the locus of control construct (Vol. 1, pp. 15–63). New York: Academic Press.

Livingstone, S. M., & Lunt, P. K. (1992). Predicting personal debt and debt repayment: psychological, social, andeconomic determinants. Journal of Economic Psychology, 13, 111–134.

Luther, R. K., Garman, E. T., Leech, I. E., GriYtt, L., & Gilroy, T. (1997). Scope and impact of personal Wnancialmanagement diYculties of service members on the Department of the Navy. Scranton, PA: Military FamilyInstitute, Marywood University.

Metrica, Inc. (2001). MTISurv: Metric Technological Incorporated Survey System (Version 1.2) [Computer Soft-ware]. San Antonio, TX: Metrica, Inc.

Roberts, R., Golding, J., Towell, T., Reid, S., Woodford, S., Vetere, A., et al. (2000). Mental and physical health instudents: the role of economic circumstances. British Journal of Health Psychology, 5(3), 289–297.

Rosenberg, M. (1989). Society and the adolescent self-image (rev. ed.). Princeton, NJ: Princeton University Press.Sherer, M., Maddux, J. E., Mercandante, B., Prentice-Dunn, S., Jacobs, B., & Rogers, R. W. (1982). The self-

eYcacy scale: construction and validation. Psychological Reports, 51, 663–671.Survey of Active Duty Personnel (1999). Provides information about active duty members and their spouses [Data

Wle]. Arlington, VA: Defense Manpower Data Center.Survey of Consumer Finances (1995). A triennial survey collects information concerning household Wnancial charac-

teristics and behavior to provide guidance to policy makers [Data File]. Available from the Federal ReserveBoard Web site, http://www.federalreserve.gov/pubs/oss/oss2/scWndex.html.

Survey of Consumer Finances (1998). A triennial survey collects information concerning household Wnancial charac-teristics and behavior to provide guidance to policy makers [Data Wle]. Available from the Federal ReserveBoard Web site, http://www.federalreserve.gov/pubs/oss/oss2/scWndex.html.

Page 14: Indicators of personal financial debt using a multi-disciplinary behavioral model

556 B. Stone, R.V. Maury / Journal of Economic Psychology 27 (2006) 543–556

Tokunaga, H. (1993). The use and abuse of consumer credit: application of psychological theory and research.Journal of Economic Psychology, 14(2), 285–316.

Wilhelm, M. S., & Varcoe, K. (1991). Assessment of Wnancial well-being: impact of objective economic indicatorsand money attitudes on Wnancial satisfaction and Wnancial progress. In S. M. Danes (Ed.), Proceedings of thefourth annual conference of the association of Wnancial counseling and planning education (pp. 184–201). KansasCity, MO.

Williams, F. L., Haldeman, V., & Cramer, S. (1996). EVect of Wnancial concerns upon workplace behavior andproductivity. Financial Counseling and Planning, 7, 147–155.

Yamauchi, K. T., & Templer, D. I. (1982). The development of a money attitude scale. Journal of PersonalityAssessment, 46, 522–528.

Zuckerman, M. (1979). Sensation seeking: Beyond the optimal level of arousal. Hillsdale, NJ: Erlbaum.