Psychology and Aging199g, Vol. 13, No. 4, 608-621

Copyright 1998 by the American Psychological Association, Inc.0882-7974/98/33.00

Motivation to Volunteer by Older Adults:A Test of Competing Measurement Models

Morris A. Okun and Alicia BarrArizona State University

A. Regula HerzogUniversity of Michigan

Four measurement models of the structure of motivation to volunteer were evaluated in 2 samplesof older (minimum age = 50 years), active volunteers. Motivation to volunteer was assessed withthe Volunteer Functions Inventory. Whereas no support was found for either unidimensional orbipartite models, qualified support was observed for both 6-factor and 2nd-order factor models. Thebest fit of the data was obtained with the 6-factor model of motivation to volunteer (career, enhance-ment, protective, social, understanding, and values). Contrary to the prediction derived from the 2nd-order factor model, the 6 volunteer motives were differentially related to demographic variables andnumber of hours spent volunteering for the organization during the past year. Implications forassessing motivation to volunteer among older adults and recruiting older adults as volunteers arediscussed.

During the past 25 years, there has been a substantial increase

in the proportion of older adults who engage in formal organiza-

tional volunteering. Chambre (1993) noted that, whereas in

1965 11% of people over the age of 65 volunteered, in 1990

41% of people over the age of 65 volunteered. Chambre sug-

gested that the increased rate of volunteer work over time among

older adults was due to several factors, including (a) improve-

ments across cohorts in educational attainment and health; (b)

favorable shifts in the public attitude toward aging and elders;

(c) an increased valuing of volunteer work; and (d) expanded

opportunities for older volunteers in the public and private

sectors.

What benefits does society accrue from volunteer work by

elders? With the exception of paid work and child care, age has

been shown to be unrelated to the percentage of adults who are

engaged in various productive activities such as housework,

home maintenance, grocery shopping, informal help, and volun-

teer work (Herzog, Kahn, Morgan, Jackson, & Antonucci,

1989). On the basis of data reported by Herzog and Morgan

(1992), it can be estimated that during 1986, people older than

54 years of age contributed more than 1.3 billion hours of volun-

teer work in organizations. It can also be estimated that for

1986, Americans over the age of 54 years contributed more than

7.8 billion dollars through their volunteer activities. Through

Morris A. Okun and Alicia Barr, Department of Psychology, ArizonaState University; A. Regula Herzog, Institute for Social Research andDepartment of Psychology, University of Michigan.

We wish to thank Martha Taylor, director of the Maricopa Retiredand Senior Volunteer Service Area Agency on Aging, Region One, andDoric Roepke, former director of volunteer services, Scottsdale Memo-rial Health Systems Incorporated, and the volunteers of these organiza-tions for their cooperation in the conduct of this research. Some workon this article was supported by an anonymous donor.

Correspondence concerning this article should be addressed to MorrisA. Okun, Department of Psychology, Arizona State University, Tempe,Arizona 82287-1104. Electronic mail maybe sent to dr.morris@asu.edu.

the efforts of older volunteers, churches, civic organizations,

and social welfare and health agencies are able to maintain

current services and to expand both the quantity and quality of

their services (Fischer, Mueller, & Cooper, 1991).

What benefits accrue to older people from volunteering? In

an experimental field study of older adults, Midlarsky (1989)

examined the effects of providing personalized information

about volunteer opportunities and encouragement to volunteer

on volunteer behavior and subjective well-being, including self-

esteem. The results of that study indicated that, compared with a

control group, the intervention group engaged in more volunteer

activity and had higher self-esteem scores. Moreover, the amounl

of volunteer activity was a significant predictor of self-esteem

scores. Productive activities such as volunteering provide an

opportunity for older people to validate the self-perception that

"I am competent" and to sustain their self-esteem (Herzog,

Franks, Markus, & Holmberg, 1998). Thus, volunteering repre-

sents one of the avenues to aging well both for older adults and

for society (Herzog & House, 1991).

In several studies of motives for volunteering (Anderson &

Moore, 1978; Gidron, 1978; Morrow-Howell & Mui, 1989;

Stone & Velmans, 1980-1981), older adults were asked the

reasons why they served as volunteers. According to the Marriott

Seniors Volunteerism Study (MSVS; Marriott Senior Living

Services, 1991), the motive endorsed most frequently by older

volunteers as a major reason for volunteering was "to help

others" (83%). Other motives for volunteering included "to

feel useful or productive" (65%) and "to fulfill a moral respon-

sibility" (51%).

The primary purpose of the present study was to test three

previously proposed models and one new model of seniors'

motivation to volunteer. To establish the generalizability of our

findings with respect to these models, we tested them using

confirmatory factor analyses (CFAs) in two separate samples.

Another purpose was to supplement our CFAs with correlational

analyses that examined the associations between motives for

volunteering on the one hand and demographic variables and

60S

VOLUNTEER MOTIVATION 609

frequency of volunteering on the other. The remainder of the

introduction is devoted to reviewing (a) weaknesses in research

on motivation to volunteer; (b) measurement models of motiva-

tion to volunteer; and (c) research on the correlates of motiva-

tion to volunteer.

Weaknesses in Measures of Motivation to Volunteer

Most research on motivation to volunteer has been flawed in

one or more ways (Clary & Snyder, 1991; Cnaan & Goldberg-

Glen, 1991). With respect to sampling, two common problems

have been the reliance on volunteers from a single site and the

use of small samples (Cnaan & Goldberg-Glen, 1991). With

respect to measurement, researchers have often used one of

three strategies: (a) analyze each individual item on a list of

motives; (b) group the items on a conceptual basis without

empirical validation; or (c) group the items on an empirical

basis without a conceptual framework (Clary & Snyder, 1991).

Cnaan and Goldberg-Glen (1991) reported that only 3 of the

27 studies they reviewed included an examination of the interre-

lationships among the motives that were assessed. Furthermore,

researchers have generally ignored the issue that demand charac-

teristics associated with completing items assessing motivation

to volunteer may influence the responses given. Thus, Clary and

Snyder (1991) observed that "the widespread use of measures

of unknown reliability and validity is troublesome" (p. 137).

Measurement Models of Motivation to Volunteer

In research aimed at understanding motivation to volunteer,

investigators have proposed one-factor (Cnaan & Goldberg-

Glen, 1991), two-factor (Latting, 1990), and six-factor (Clary,

Snyder, & Ridge, 1992) models. We empirically evaluated these

three measurement models of motivation to volunteer. In addi-

tion, we examined the viability of a second-order factor model

of seniors' motivation to volunteer. We now describe each of

these four models in detail.

Unidimensional Factor Model

The Unidimensional model of motivation to volunteer

(Cnaan & Goldberg-Glen, 1991) postulates that "volunteers act

not from a single motive or a category of motives but from a

combination of motives that can be described overall as 'a re-

warding experience' " (p. 281). Consistent with Tinker's

(1979) theory of commitment, the Unidimensional model pro-

poses that people are motivated to volunteer by a particularmeaningful whole that is relevant for them.

Support for this conceptualization is derived from a study of

258 committed volunteers from human service agencies and

104 nonvolunteers (mean age = 50.6 years). When Cnaan and

Goldberg (1991) conducted an exploratory factor analysis using

varimax rotation and an unspecified extraction technique on

responses to the 28 items culled from their literature review,

they found that a single-factor model provided a better fit to the

data than several multifactor models. Twenty-two items loaded

on this factor. Cnaan and Goldberg formed the Motivation to

Volunteer Scale by summing the responses to the 22 items (coef-

ficient a = .86).

One problem with the items on the Motivation to Volunteer

Scale is that they were not derived within the context of a

theoretical framework. When Luciani (1993) performed an ex-

ploratory factor analysis using varimax rotation and an unspeci-

fied extraction technique on the items on the Motivation to Vol-

unteer Scale (N = 432; minimum age = 55 years), she found

seven factors.

Two-Factor Model

Several researchers contend that people have more than one

reason for volunteering (Unger, 1991; Van Til, 1988). The bipar-

tite model (Frisch & Gerrard, 1981; Latting, 1990) posits that

people are motivated to volunteer by concerns for others (altruis-

tic motives) and self (egoistic motives).

Frisch and Gerrard (1981) administered 11 items on motives

for volunteering to a sample of Red Cross volunteers (mean

age = 49 years). A principal-components analysis (PCA) of

the responses to these items supported the notion that altruistic

and egoistic motives were distinct dimensions of volunteer moti-

vation. It is not clear whether the researchers specified an orthog-

onal or nonorthogonal rotation of the factors that were extracted

in this analysis. Similar findings were reported by Latting

(1990). She administered nine items on motives for volunteering

to a sample of Big Brothers and Big Sisters (mean age = 32

years). Responses to these items were subjected to an explor-

atory factor analysis (factor extraction technique not reported)

using varimax rotation. The results supported the bipartite

model. The two factors that emerged were labeled Altruistic and

Egoistic.

Clary and Snyder (1991) observed that there were several

distinct egoistic motives for volunteering. Consequently, Clary et

al. (1992) developed a more complex model of the motivational

underpinnings of volunteering.

Multifactor Model

Clary et al. (1992) proposed a functional theory of motivation

to volunteer. According to the functional theory, acts of volun-

teerism can be analyzed in terms of differences in the motives

that are satisfied, the needs that are met, and the goals that

are reached. Clary and Snyder (1990) identified six primary

functions that are relevant to volunteering. These functions in-

clude (a) values (i.e., expressing deeply held beliefs about the

importance of helping others); (b) social (i.e., conforming to

the norms of significant others); (c) career (i.e., seeking ways

to get started or advance in the world of work); (d) understand-

ing (i.e., engaging in activities that promote learning); (e) en-

hancement (i.e., enhancing one's sense of self worth); and ( f )

protective (i.e., escaping negative feelings).

On the basis of this classification scheme, Clary et al. (1992)

developed the Volunteer Functions Inventory (VFI). They dem-

onstrated that the subscales of the VFI that assess the six motives

for volunteering have adequate internal consistency reliability

(lowest coefficient a = .80) in samples of adult volunteers

(mean age = 41 years) and undergraduates (60% had experi-

ence as volunteers). Test-retest reliability estimates for VFI

subscales over a 1-month period for a sample of college students

(62% had experience as volunteers) ranged from .64 (Protec-

610 OKUN, BARR, AND HERZOG

live) to .78 (Values). Furthermore, scores on the VFI subscales

are unrelated to social desirability (Clary et al., 1992).

Clary et al. (1998) performed two separate sets of factor

analyses on the VFI. They began by carrying out an exploratory

factor analysis, using the PCA extraction technique (factor rota-

tion not reported), of the responses of a sample of volunteers

(mean age = 41 years). As expected, six components emerged

with eigenvalues greater than one. Next, Clary et al. performed

a secondary exploratory factor analysis using the principal axis

factor extraction technique with oblique rotation, in which six

factors were specified. The results of this analysis supported

their functional motivation for volunteering model, with items

almost always loading on their intended factor.

Finally, CFA was applied to the VFI data and five-, six-, and

seven-factor oblique models were tested. The method of factor

extraction for these analyses was not reported. The best fit was

obtained for the six-factor model, x2( 120, N = 434) = 412.69,

goodness-of-fit index (GFI) = .91, nonnormed fit index (NFI)

= .90, and Root Mean Square residual = .057. Clary et al.

(1998) then cross-validated their factor-analytic results in a sep-

arate sample of undergraduates (60% had experience volunteer-

ing). The pattern of findings showed a high degree of generaliz-

ability. Again, the best fit was observed for the six-factor model,

X2(120, N = 535) = 630.37, GFI = .89, NFI = .88, and

RMSres = .065.

Second-Order Factor Model

Conceptually, one might expect negative correlations between

altruistic and egoistic motive scores. However, Clary and Miller

(1986) reported a positive correlation (r = . 37) between endors-

ing altruistic and egoistic motives for volunteering. In a study

of AIDS volunteers, Omoto and Snyder (1995) reported that

all 10 correlations among five motives for volunteering subscale

scores (i.e., Values, Community Concern, Understanding, Per-

sonal Development, and Esteem Enhancement) were positive

and significant (p < .05). These correlations ranged from .20

(Values with Personal Development) to .46 (Community Con-

cern with Personal Development) and had a median correlation

of .335. Similarly, Clary et al. (1998) reported mean correla-

tions among the six VFI motive scales of .34 for a sample of

adult volunteers and .41 for a sample of students. Therefore,

building on the first-order factor model, it can be hypothesized

that the six functions identified by Clary et al. (1992) are indica-

tors of a more global construct that can be labeled a general

motivation to volunteer,

The second-order factor model posits that there are several

interrelated dimensions of motivation to volunteer. Furthermore,

the second-order factor model assumes that the covariation

among the first-order dimensions is due to their common depen-

dence on a more abstract and overarching construct reflecting

one's overall motivation to volunteer.

Correlates of Volunteer Motives

The secondary purpose of the present study was to examine

the correlations between (a) VFI total scale scores and VFI

subscale scores and (b) the demographic variables and fre-

quency of volunteering. In several cross-sectional studies, age

has been found to be inversely related to the career, understand-

ing, and social motives for volunteering (Clary & Snyder, 1990;

Fischer & Schaffer, 1993; Francies, 1983; Frisch & Gerrard,

1981; Gidron, 1978; Wiehe & Isenhour, 1977) and positively

related to the protective motive for volunteering (Fischer et al.,

1991; Stone & Velmans, 1980-1981). Inconsistent results have

been found for sex differences in volunteer motives. For exam-

ple, whereas Sainer and Zander (1971) reported that men were

more motivated to volunteer than women by the protective mo-

tive, Morrow-Howell and Mui (1989) observed that women

were more motivated to volunteer than men by the protective

motive. As educational attainment increases, protective motives

for volunteering appear to decrease in importance (Anderson &

Moore, 1978).

Relative to volunteers who are retired, volunteers who are

working have been shown to be more motivated by values and

social motives (Marriott Senior Living Services, 1991). Ac-

cording to the MSVS (Marriott Senior Living Services, 1991),

single and divorced volunteers are more motivated by the protec-

tive motive than married volunteers.

Scant attention has been paid to the correlates of frequency

of senior volunteering. In a causal model, Herzog and Morgan

(1993) found that only two variables—extraversion and organi-

zational participation—had significant direct effects on the

number of hours spent volunteering. Okun (1994) also investi-

gated predictors of frequency of volunteering. He observed that

the background variables of educational attainment and affiliat-

ing with the Democratic Party were positively related to the

frequency of volunteering and that working full time was in-

versely related to frequency of volunteering.

Of particular interest to the present study, two specific motives

for volunteering predicted frequency of volunteer (Okun, 1994).

The motive of "volunteering to feel useful or productive" was

the strongest predictor of frequency of volunteering. The motive

of "volunteering to fulfill a moral obligation" was also a sig-

nificant predictor of frequency of volunteering. Furthermore,

controlling for the contribution of these motive variables, the

number of major reasons for volunteering did not add to the

prediction of frequency of volunteering. Okun (1994) con-

cluded that it is important to consider the specific motives of

older volunteers as opposed to their overall level of motivation.

Method

Participants

In the present study, models were tested on data collected from two

separate samples of volunteers. Participants from the first sample were

adults (minimum age = 50 years) who volunteered at health care facili-

ties located in Scottsdale, Arizona. The facilities were operated by

Scottsdale Memorial Health Systems Incorporated (SMHSI sample).

Of the 808 potential respondents, 409 volunteers participated, resulting

in a response rate of 51%.

Participants from the second sample were adults (minimum age =

55 years) who were members of the Maricopa Chapter of the Retired

and Senior Volunteer Program (RSVP sample). Maricopa RSVP is spon-

sored by the Area Agency on Aging, Region One, and serves Scottsdale,

Phoenix, and the western part of Maricopa County. Maricopa RSVP

acts as a "broker," matching members with organizations that need

volunteers. Thus, RSVP members volunteer in heterogeneous settings

VOLUNTEER MOTIVATION 611

and carry out diverse assignments. Of the 677 potential RSVP respon-

dents, 372 volunteers participated, resulting in a response rate of 55%.

Comparison of the RSVP and SMHSI samples. The demographic

characteristics of the two samples are summarized in Table 1. The sam-

ples were compared on demographic variables using chi-square tests.

The two samples did not differ significantly (p > .05) with respect to

sex, ethnicity, and marital status. Significant (p < .05) differences be-

tween the RSVP and SMHSI samples were observed with respect to

age, X2(3, N = 781) = 14.76, education, *2(4, N = 781) = 20.14,

perceived health, x2(4, AT = 781) = 37.56, and work status, x2(3, N

= 781) = 21.92.

Relative to the RSVP sample, the SMHSI sample had a higher percent-

age of volunteers who were younger than 60 (12% vs. 4%) and a lower

percentage of volunteers who did not complete high school (2% vs.

10%). Relative to the SMHSI sample, the RSVP sample had a lower

percentage of volunteers who rated their health as very good or excellent

(43% vs. 68% ) and a higher percentage of volunteers who were working

(23% vs. 9%).

Comparison of participants and nanparticipants within each organi-

zation. From archival records, we were able to obtain data on the

population of volunteers for several variables that were assessed on our

questionnaire. For both organizations, we examined whether participants

and nonparticipants differed on sex, ethnicity, and hours volunteered for

the organization during the past year. In addition, for the SMHSI sample

we were able to compare participants and nonparticipants on self-rated

Table 1

Summary of Demographic Characteristics of Each Sample

Variable RSVP

SexFemaleMale

EthnicityWhite, non-HispanicOther

Age50-59 years60-69 years

70-79 years>79 years

EducationDid not complete high schoolCompleted high schoolComplete some collegeCompleted a college degreeCompleted a graduate degree

Perceived healthPoor or fairGoodVery goodExcellent

Marital statusMarried

WidowedOther

Work statusRetiredEmployed part timeEmployed full timeUnemployed

7030

982

4364713

102141209

14433013

513514

7411123

7525

982

12374110

224402311

6264523

612910

84276

health; for the RSVP sample we were able to compare participants and

nonparticipants on age. Because we obtained the frequency distributions

for the population of volunteers and generated the frequency distributions

for the volunteers who participated in our study, we were able to compute

the frequency distributions for the volunteers who chose not to partici-

pate in the study.

Results of chi-square tests indicated that within the SMHSI organiza-

tion, participants and nonparticipants did not differ significantly (p >

.05) with respect to ethnicity and self-rated health. Significant (p <

.05) differences between participants and nonparticipants were observed

with respect to sex, x2(l, N = 808) = 29.83, and hours volunteered

for SMHSI during the past year, x2( l , N = 808) = 10.94. Whereas

25% of the participants were men, only 14% of the nonparticipants were

men. Whereas 72% of the participants devoted more than 208 hr to

volunteering for SMHSI during the past year, only 63% of the nonpartici-

pants devoted more than 208 hr to volunteering for SMHSI during the

past year.

Results of chi-square tests indicated that within the RSVP organiza-

tion, participants and nonparticipants did not differ significantly (p >

.05) with respect to gender. Significant (p < .05) differences between

participants and nonparticipants were observed with respect to ethnicity,

X 2 ( l , N = 677) = 51.15, age, x2(l, N = 677) = 22.43, and hours

volunteered for RSVP during the past year, x J(l, N = 677) = 21.79.

Whereas 98% of the participants were White, non-Hispanic, only 82%

of the nonparticipants were White, non-Hispanic. Volunteers younger

than 70 made up 40% of the participants but only 23% of the nonpartici-

pants. Whereas 19% of the participants devoted more than 520 hr to

volunteering for SMHSI during the past year, only 7% of the nonpartici-

pants devoted more than 520 hr to volunteering for SMHSI during the

past year. These results suggest that participants were younger and de-

voted more time to volunteering relative to nonparticipants.

SMHSI Procedure

Note. RSVP = Retired and Senior Volunteer Program; SMHSIScottsdale Memorial Health Systems Incorporated.

The same procedures were used with both samples. Data were col-

lected from SMHSI volunteers during the winter of 1994 and from

RSVP volunteers during the spring of 1995. The packet of materials

sent to the volunteers contained (a) a letter from the director of volunteer

services and one of us inviting them to participate in a research study of

senior volunteer motivation; (b) a questionnaire; and (c) a preaddressed,

prestamped return envelope. The volunteers were guaranteed anonymity

to minimize social desirability bias in completing the VFI. Consequently,

it was not possible to follow up with nonrespondents.

Measures

The survey packet associated with each sample contained several

questionnaire measures. Survey packets for both samples assessed rea-

sons for volunteering, hours spent volunteering, and demographics using

the same questions. In addition, the survey packets for each sample

contained a unique section that addressed different issues that had been

identified by each director of volunteer services.

VFI. On the VFI (Clary & Snyder, 1991) respondents indicate on a

7-point scale how important and accurate each reason is for why they

volunteer. Scale endpoints are not at all important/accurate for you (1)

and extremely important/accurate for you (7). When completing the

VFI, respondents were instructed to remember the particular organiza-

tion that they volunteered for (SMHSI or RSVP). The VFI consists of

30 items, 5 items per motive. Tb minimize order effects, we mixed items

tapping different motives together and changed the order of the VFI

items from the first to the second sample.

Hours spent volunteering. The hours spent volunteering were con-

structed from responses to two questions. The first question asked re-

spondents how many weeks during the past year they had volunteered

612 OKUN, BARR, AND HERZOG

at this placement The second question asked respondents how many

hours a week they volunteered at this placement during a typical week

when they volunteered. The number of hours spent volunteering was

computed by multiplying the number of weeks devoted to volunteering

at this placement during the past year by the number of hours spent

volunteering at this placement during a typical week.

Demographics. Participants were asked questions regarding their

age, sex, ethnicity, marital status, work status, education, and health

status and the number of weeks during the past year that they resided

in Phoenix.

Results

CFAs

A series of four CFAs was conducted on the 30 observed

indicators listed in Table 2. These analyses were conducted with

the LISREL VIII statistical package (Joreskog & Sorbom, 1993)

and used maximum-likelihood estimation. By conducting the

analyses separately for each sample, we addressed the issue of

the generalizability of our findings with respect to the fit of the

models to the data.

Several fit indices were used to evaluate each model. Two

absolute GFIs were used: the chi-square statistic and the x2'-df

ratio. Models are deemed to provide acceptable fits when the

chi-square statistic is not significant. A perfect model fit is indi-

cated by a x2'-df ratio that is equal to 1.00. Because the chi-

square statistic and the \*'.df ratio are both absolute indices of

goodness of fit, they are potentially biased by large sample

sizes and degrees of freedom. Therefore, we also evaluated the

measurement models using several indices that correct for these

potential biases: the root-mean-square error of approximation

(RMSEA), the comparative fit index (CFI), and the incremental

fit index (IFI). A value that is not statistically greater than .05

on the RMSEA suggests an adequate fit, whereas a value of .90

or higher on the CFI and IFI indicates an adequate fit. Note that

Table 2

Observed Indicators Used in Each Measurement Model

Variable Item description

! Values (value expressive)

X,X2

X3

X,

x,2 Social

X6

X,X,

3 Protective

X,,

X12

x,3X14

4 Understanding

X,6X,,

X,,

5 Career

X2I

X22

X33

X24

X2,

6 Enhancement (esteem)

I am concerned about those less fortunate than myself.

I am genuinely concerned about the particular group I am

serving.

I feel compassion toward people in need.

1 can do something for a cause that is important to me.

I feel it is important to help others.

My friends volunteer.

People I am close to want me to volunteer.

Others to whom I am close place a high value on community

service.

Volunteering is an important activity to the people 1 know best.

People I know share an interest in community service.

No matter how bad I've been feeling, volunteering helps me to

forget about it.

By volunteering I feel less lonely.

Volunteering helps me work through my own personal problems.

Volunteering is a good escape from my own troubles.

Doing volunteer work relieves me of some of the guilt over being

more fortunate than others.

I can learn more about the cause for which I am working.

Volunteering allows me to gain a new perspective on things.

Volunteering lets me learn through direct hands-on experience.

1 can learn how to deal with a variety of people.

I can explore my own strengths.

Volunteering can help me get my foot in the door at a place

where I would like to work.

I can make new contacts that might help my business or career.

Volunteering allows me to explore different career options.

Volunteering will help me succeed in my chosen profession.

Volunteering experience will look good on my resume.

Volunteering makes me feel important.

Volunteering increases my self-esteem.

Volunteering makes me feel needed.

Volunteering is a way to make new friends.

Volunteering makes me feel better about myself.

VOLUNTEER MOTIVATION 613

these criterion values are to be used as guides, not as absolute

measures of model fit.

Unidimensional model. The first CFA tested the theory that

the combination of motives reported by older volunteers is the

result of an underlying motivation for a rewarding experience.

In the single-factor model, each of the 30 indicators in Table 2

was specified to load on the single Reward factor.

The GFIs associated with the single-factor model revealed

that this measurement specification did not provide an adequate

fit to the data. As shown in Table 3, none of the absolute or

relative fit indices approached the cutoff value associated with

an acceptable measurement model.

Two-factor model. The second CFA tested the theory that

older volunteers are motivated by either altruistic or egoistic

motives. For this model, the first five indicators in Table 2 (i.e.,

the value-expressive items) were specified to load on the Altruis-

tic Motive factor, and the remaining 25 indicators in Table 2 were

specified to load on the Egoistic Motive factor. Furthermore, the

two factors were allowed to correlate with each other. Inspection

of Table 3 reveals that the absolute and relative GFIs were

unacceptable, suggesting that a two-factor model of motivation

for volunteering did not adequately fit the data for either sample.

Multlfactor model. The third measurement model tested the

theory that older adults are motivated by multiple, distinct mo-

tives. Figure 1 depicts the multifactor model that was empirically

evaluated. The following six motivation factors were specified:

Values (value expressive), Social, Protective, Understanding,

Career, and Enhancement (esteem). In terms of the specification

of the model's parameters, (a) the lambda-x matrix of factor

loadings was specified such that each of the 30 indicators in

Table 1 had a nonzero loading on one factor, whereas the load-

ings on the other five factors were constrained to be zero; (b)

all off-diagonal elements of the theta-delta matrix of errors of

measurement were fixed to zero, consistent with the expectation

of uncorrelated error terms; and (c) because the dimensions of

motivation to volunteer were predicted to be correlated, the

covariances among the six latent constructs were specified as

free to be estimated in phi, the latent factor covariance matrix.

In addition, to identify the model, the variance of each latent

construct was fixed to be a value of 1.0.

As can be seen in Table 3, the six-factor model of motivation

to volunteer in the SMHSI sample came close to meeting several

of the cutoffs associated with the GFIs. Although the chi-square

values were large in both samples, and the xl'-df ratios were

above 2.4 in both samples, in the SMHSI sample the CFI and

IFI values were .90 and the RMSEA was .06. By comparison,

the RMSEA was somewhat higher and the CFI and IFI were

somewhat lower in the RSVP sample.

Another way to examine the utility of the multifactor model

is to inspect the factor loadings. In general, loadings of .40 or

higher are acceptable. As can be seen in Figure 1, only 2 of the

30 factor loadings for each sample were below .40. The loading

of the indicator "My friends volunteer" on the social motive in

the SMHSI sample was .39, as was the loading of the indicator

"Doing volunteer work relieves me of some of the guilt over

being more fortunate than others'' on the protective motive in

the RSVP sample. Moreover, in each sample, all factor loadings

were significant at the .01 level.

We had anticipated that the motive factors would be substan-

tially correlated with each other. With the exception of the career

motive, the results were consistent with our expectations. As

can be seen in Figure 1, excluding the career motive, the median

factor correlation was .665 in the SMHSI sample and .530 in

the RSVP sample.

Second-order factor model. The final CFA tested the theory

that the six functional motives are indicators of a more general

motive to volunteer among older adults. Figure 2 depicts the

second-order factor model that was empirically evaluated.

Again, six latent factors were specified—Values (value expres-

sive), Social, Protective, Understanding, Career, and Enhance-

ment (esteem)—as dimensions of older adults' responses to

the VFI items. The second-order factor, called General Motiva-

tion to Volunteer, was hypothesized to ' 'cause'' each of the first-

order factors. The specification of the parameters for the second-

order model factor model was similar to the specification used

for the multifactor model, except that the loadings of the first-

Table 3

Fit Indices for the One-, Two-, Multi-, and Second-Order Factor Models

Model df RMSEA CFI IFI

One factorSMHSI sampleRSVP sample

Two factorSMHSI sampleRSVP sample

MultifactorSMHSI sampleRSVP sample

Second orderSMHSI sampleRSVP sample

2,496.322,852.70

2,236.532,470.65

940.621,114.74

992.671,160.64

405405

404404

390390

399399

6.167.04

5.54

6.12

2.412.86

2.492.91

.11

.14

.11

.13

.06

.08

.062

.081

.62

.48

.67

.56

.90

.84

.89

.84

.63

.48

.67

.56

.90

.85

.89

.84

Note. All chi-square measures and tests of RMSEA close fit were significant at the .001 level. RMSEA= root-mean-square error of approximation; CFI = cumulative fit index; IFI = incremental fit index; SMHSI= Scottsdale Memorial Health Systems Incorporated (N = 409); RSVP = Retired and Senior VolunteerProgram (N = 372).

614 OKUN, BARR, AND HERZOG

.27 (.38)

.21 (.19)

.12 (.14)

.15 (.16)

.55 (.72)

.72 (.56)

.58 (.42)

.16 (.09)

.12 (.12)

.11 (.22)

.32 (.44)

.26 (.16)

.27 (.23)

.18 (.25)

.26 (.26)

.23 (.35)

.30 (.12)

.17 (.13)

.20 (.04)

.19 (.15)

.46 (.53)

.18 (.13)

.12 (.05)

.28 (.55)

.20 (.05)

VOLUNTEER MOTIVATION 615

order factors on the second-order factor were estimated and the

correlations among the first-order factors were set to zero (off-

diagonal elements of PSI matrix = 0).

The fit of the second-order factor model was highly similar

to that of the multifactor model. The chi-square for each model

was significant, and the x2-df ratio was above 2.40 in both

samples. Whereas the values for the relative fit indices ap-

proached the cutoffs in the SMHSI sample, the RMSEA index

was somewhat higher and the CFIs and IFIs were somewhat

lower in the RSVP sample (see Table 3).

As can be seen in Figure 2, the first-order factor loadings

ranged from .39 to .82 in the SMHSI sample and from .37 to

.90 in the RSVP sample. Note that these loadings are virtually

identical to the loadings obtained from estimating the first-order,

six-factor model. Figure 2 also shows the second-order factor

loadings, which indicate how strongly the first-order factors

were related to the higher order motivation to volunteer dimen-

sion. In the SMHSI sample, the second-order factor loadings

were highest for the enhancement (esteem; .95) and understand-

ing (.94) motives and lowest for the career motive (.29). Simi-

larly, in the RSVP sample, the second-order factor loadings were

highest for the understanding (.87) and enhancement (esteem;

.83) motives and lowest for the career motive (.27). Except for

the career motive, the second-order factor loadings were within

acceptable limits.

Comparison of Jit among models. Because all four models

were nested, we performed chi-square difference tests (James,

Mulaik, & Brett, 1983) to compare their goodness of fit to the

data. As can be seen in Table 4, specification of a one-factor

model resulted in a significant degradation in fit relative to the

other three models. Similarly, the fit of the two-factor model to

the data was not as good as the fit of either the multifactor

model or the second-order factor model. Finally, the fit of the

second-order factor model to the data was not as good as the

fit of the six-factor model.

Sources of Misfit for the Six-Factor Model

To determine the sources of misfit, we inspected the modifi-

cation indices and covariances between latent variables and indi-

cators for the six-factor model in both samples. Of the 15 en-

hancement, understanding, and protective items, 6 had modifi-

cation indices above 10 for cross-loadings on the Enhancement,

Understanding, and Protective factors in one of the two samples.

For example, in the RSVP sample, the modification indices for

loading the enhancement item "Volunteering is a way to make

new friends" on the Understanding and Protective factors were

66.94 and 26.73, respectively. In the RSVP sample, the covari-

ance of this enhancement item was 1.04 with the Enhancement

factor and 0.75 and 0.66 with the Understanding and Protective

factors. In contrast, the covariances of this enhancement item

with the other three factors was .55 (Values) .47 (Social), and

.16 (Career). The magnitude of the cross-loadings of some of

the enhancement, protective, and understanding items on the

Enhancement, Protective, and Understanding factors suggested

that these items may not be "pure" markers of only one dimen-

sion of motivation to volunteer among older adults.

Correlates of Motivation to Volunteer

Although the original six-factor model provided a better fit

to the data than the second-order factor model, the chi-square

change was relatively small (see Table 4). Therefore, one can

argue that the second-order factor model is superior to the six-

factor model because the former provides a more parsimonious

explanation of motivation to volunteer by older adults than the

latter. Practically speaking, if the relations between the various

motive subscales and other variables are uniform, then research-

ers and practitioners would be justified in collapsing the six

first-order dimensions into a single summary score denoting an

overall motivation to volunteer.

To address this issue, we examined whether the VFI scales

were related to demographic variables, self-reports of the num-

ber of hours spent volunteering for the organization during the

past 12 months, and sample. Using unit weighting, for each of

the six VFI subscales, we averaged the responses to five items.

For the VFI total scale score, we averaged the responses to the

30 items. Thus, scores on each of the six VFI subscales and the

VFI scale score could range from 1 to 7. Table 5 shows estimates

of internal consistency reliability, means, and standard devia-

tions for the VFI scales.

Table 6 shows the correlations of the VFI scales with the

demographic variables and with the number of hours spent vol-

unteering during the past month. To eliminate an extraneous

source of variance from these analyses, we excluded seasonal

residents. Obviously, seasonal residents would have fewer op-

portunities to volunteer at these sites relative to year-round resi-

dents. We decided to report these correlations separately for

each sample because the SMHSI and RSVP samples differed

significantly on several demographic variables.

Within each sample, except for education, each demographic

variable was correlated significantly with only one or two of

the motives. For example, age was correlated positively with the

social motive in both samples and age was correlated negatively

with the career motive in the RSVP sample. Furthermore, in the

SMHSI and RSVP samples alike, at least one VFI subscale

exhibited a stronger relation with each of the demographic vari-

ables than the VFI total scale score. For example, the correlation

between being widowed and Protective scale scores was .18 in

the RSVP sample and .19 in the SMHSI sample, whereas the

Figure 1. A multifactor model of motivation to volunteer LISREL-derived completely standardized coeffi-

cients are presented first for the Retired and Senior Volunteer Program sample, followed by those for

Scottsdale Memorial Health Systems Incorporated sample in parentheses. Curved arrows represent relation-

ships between latent factors (phi coefficients), straight arrows from latent factors to observed indicators

represent factor loadings (lambda coefficients), and the remaining straight arrows to observed indicators

represent error variance (theta coefficients).

616 OKUN, BARR, AND HERZOG

.29 (.21)

.42 (.38)

.10 (.07)

.34 (.34)

.20 (.16)

.45 (.52)

.18 (.12)

.12 (.05)

.29 (.55)

.21 (.06)

VOLUNTEER MOTIVATION 617

Table 4

Pairwise Comparison of Fit Among the Four Measurement Models

Models compared SMHSI RSVP Outcome

One versus two factorOne versus multifactorOne versus second orderTwo versus multifactorTwo versus second orderMultiple versus second order

259.791,555.701,503.651,295.911,243.86

52.05

382.051,737.961,692.061,355.911,310.01

45.90

1156

1459

Two-factor model superior (p < .001)Multiple-factor model superior (p < .001)Second-order model superior (p < .001)Multiple-factor model superior (p < .001)Second-order model superior (p < .001)Multiple-factor model superior (p < .001)

Note. SMHSI = Scottsdale Memorial Health Systems Incorporated; RSVP = Retired and Senior Volunteer Program.

correlation between being widowed and the VFI total scale score

was .08 in both samples.

In the SMHSI sample, the number of hours spent volunteering

for the organization during the past year was significantly corre-

lated with the Enhancement scale, r(300) = .13, p < .05, and

the VFI total scale, K300) = .12, p < .05. For the RSVP

sample, the number of hours spent volunteering for the organiza-

tion during the past year was significantly correlated with the

Enhancement scale, r(246) = .21, p < .01, the Protective scale,

r(246) = .14, p < .05, the Understanding scale, K246) = .14,

p < .05, and the VFI total scale, /<246) = .18, p < .01.

The volunteers in the two samples did not differ significantly

(p > .05) with respect to the Enhancement and Career scales.

In contrast, RSVP volunteers had significantly (p < .05) higher

scores than SMHSI volunteers on the Social (Ms = 3.32 vs.

2.92), Values (Ms = 5.67 vs. 4.55), Understanding (Ms = 4.61

vs. 4.35), and Protective (Ms = 3.53 vs. 3.13) scales as well

as the VFI total scale (Ms = 3.88 vs. 3.62).

Discussion

Our discussion is divided into five sections: (a) evaluation of

the measurement models; (b) comparison with other studies

using the VFI; (c) limitations of the study; (d) implications for

research; and (e) implications for practice.

Evaluation of the Measurement Models

In 1991, Fischer et al. concluded that "the truth of the matter

is that there is little understanding of why people volunteer"

(p. 186). Cnaan and Goldberg-Glen (1991) noted that an im-

portant gap in knowledge pertains to the structure of motivation

to volunteer. In the present study, we empirically tested four

measurement models of motivation in two samples of older

adults.

The earliest measurement models of motivation to volunteer

posited simple conceptual structures. The one-factor model

(Cnaan & Goldberg, 1991) proposes that motivation to volun-

teer represents a unitary composition of motives. The two-factor

model (Frisch & Gerrard, 1981) hypothesizes that motivation

to volunteer consists of two distinct motives: altruistic (concern

for others) and egoistic (concern for self). The x*'-dfs ratios

and the GFIs provide no support for the notions that motivation

to volunteer by older adults is a unidimensional or bipartite

construct. Thus, more complex conceptualizations of older

adults' motivation to volunteer are warranted.

We examined a six-factor model derived from a functional

analysis of motivation to volunteer (Clary et al., 1992). Fur-

thermore, because the first-order factors were observed to be

correlated in younger samples (Clary et al., 1998; Omoto &

Snyder, 1995), we tested a second-order factor model. Concep-

tually, the difference between these two models rests on the

assumption made in the second-order factor model that the

correlations among the first-order factors are attributable to

their common dependence on a more abstract higher order

construct reflecting one's overall motivation to volunteer. The

X2'-df ratios and the relative GFIs provided qualified support

for both of these measurement models. Also, the IFIs listed in

Table 2 were similar for the six-factor and second-order factor

models. For instance, the difference in the NFI between the

six-factor model and the second-order factor model was only

.01 in both the SMHSI and RSVP samples. Nevertheless, the

fit of the multifactor model to the data was superior to the fit

of the second-order factor model.

Krause (1993) suggested that two other criteria could be used

to evaluate the relative utility of first- and second-order factor

models. First, the factor loadings can be compared. Because the

differences between the two models in the estimates of the first-

order factor loadings were negligible, this criterion did not per-

mit us to distinguish between the utility of the first- and second-

order factor models.

Figure 2. A second-order factor model of motivation to volunteer. LISREL-derived completely standardizedcoefficients are presented first for the Retired and Senior Volunteer Program, followed by those for theScottsdale Memorial Health Systems Incorporated sample in parentheses. Straight arrows from £, to rjsrepresent the loadings of first-order latent factors on the second-order factor (gamma coefficients), straightarrows from latent factors to observed indicators represent factor loadings (lambda coefficients), and theremaining straight arrows to observed indicators represent error variance (theta coefficients).

618 OKUN, BARR, AND HERZOG

Table 5

Descriptive Statistics and Coefficient Alphas for Volunteer Motivation Scales

Scale

CareerEnhancementProtectiveSocialUnderstandingValues

Total

a

.84

.83

.83

.80

.83

.81

.93

SMHSI

M

1.364.553.132.924.355.443.62

SD

0.79

1.481.561.441.551.321.03

a

.88

.83

.79

.83

.82

.84

.92

RSVP

M

1.48

4.663.533.324.615.673.88

SD

1.111.641.611.051.651.391.08

Note. SMHSI = Scottsdale Memorial Health Systems Incorporated; RSVP =Program.

Retired and Senior Volunteer

The second criteria suggested by Krause (1993) is to assess

whether the dimensions of the construct are differentially related

to other variables. If the effects of the first-order factors are

uniform, then researchers and practitioners are justified in using

the global construct. However, if the first-order factors exert

differential effects, then researchers and practitioners need to

use the various dimensions of the construct.

In the present study, we focused on whether scores on the

VFI scales would be related to demographic variables, self-

reported hours spent volunteering for the organization during

the past year (see Table 6), and sample. Clearly, the bivariate

relations between the six motivation to volunteer scales and the

demographic variables, frequency of volunteering, and sample

(RSVP vs. SMHSI) were not uniform. Furthermore, within the

SMHSI and RSVP samples, the VFI total score did not exhibit

the largest association with any of the demographic variables

or with frequency of volunteering. Thus, the six-factor model

appears to be more viable than the second-order factor model

because the scale scores varied in terms of their relationships

to the demographic variables and the frequency of volunteering,

and the six-factor model provides a slightly better fit to the data

than the second-order factor model.

Table 6

Correlates of Volunteer Motivation

Motive scale

CareerRSVPSMHSI

EnhancementRSVPSMHSI

ProtectiveRSVPSMHSI

SocialRSVPSMHSI

UnderstandingRSVPSMHSI

ValuesRSVPSMHSITotal

RSVPSMHSI

Frequency ofvolunteering

.10.05

.21**

.13*

.14*

.10

.12.07

.14**

.11

.04

.06

.18**

.12*

Gender"

.10

.09

.05

.03

.19**

.15**

-.01-.01

.12

.15**

-.01-.06

.10

.10

Age

-.15*-.03

.11

.03

.11

.09

.16**

.20***

-.02-.04

.05-.01

.08

.06

Education

-.05-.11*

-.19**-.16**

_.24**«

-.14*

-.17**-.02

-.03-.16**

-.14*-.03

-.20***-.14*

Physical health

-.12*-.01

-.03-.08

-.09-.15**

-.04-.08

.14*

.02

.07

.08

-.01-.05

Marital status'

.06

.08

.02

.03

.18**

.19"**

.04

.03

.04

.01

-.02.03

.08

.08

Employment status'

.23***.02

.04

-.03

.03-.07

-.02.02

-.01.03

.01

.04

.05

.00

Note. Correlations for the SMHSI sample appear in boldface type, and the remaining correlations represent the RSVP sample. For correlationsbetween volunteer motives and frequency of volunteering, the sample sizes were 248 and 302 in the RSVP and SMHSI samples, respectively. Forcorrelations between volunteer motives and demographic variables, the sample size were 264 and 339 in the RSVP and SMHSI samples, respectively.* Women = 1; men = 0. b Widowed individuals = 1; married, divorced, and single individuals = 0. c Unemployed individuals = 1; employedand retired individuals = 0.*p < .05. **p < .01. ***p < .001.

VOLUNTEER MOTIVATION 619

Comparison With Other Studies Using the VFI

Clary et al. (1998) used CFA to analyze the structure of

motivation to volunteer in college student and middle-aged sam-

ples. In the present study, we performed similar analyses on two

samples of older volunteers. It is interesting to compare the

results with respect to the factors that emerge, the pattern of

factor correlations, and indices of goodness of fit. Across all the

samples, the same six dimensions appear to underlie motivation

to volunteer. In addition, the pattern of factor correlations across

samples of different ages suggests that the dimensions of motiva-

tion to volunteer are interrelated. The one exception is that the

Career factor is correlated less with the other factors in the older

adult sample. This discrepancy probably is attributable to the

fact that older adults rarely endorse career items as reasons

for volunteering. Given that volunteering in later life may be

associated with a transition from full- to part-time work (Her-

zog & Morgan, 1993), it is not surprising that older people

rarely are motivated to volunteer by career concerns.

On some indices (e.g., the GFI, NFI, and RMSres) Clary et

al. (1998) achieved an acceptable fit in their middle-aged sam-

ple and a marginally acceptable fit in their college student sam-

ple. However, the x2: df 'ratios for the six-factor model were high

in both the middle-aged (3.44:1) and college student (5.25:1)

samples. In our six-factor model, the fit was marginal for the

SMHSI sample and substandard for the RSVP sample. The

X2:d/ratios for the six-factor model were 2.41:1 and 2.86:1 in

the SMHSI and RSVP samples, respectively.

Inspection of the modification indices and covariances re-

vealed that some of the protective, enhancement, and understand-

ing items were strongly related to other factors (e.g., an enhance-

ment indicator with the Protective factor). Understanding the

relationships among the Enhancement, Protective, and Under-

standing factors and their indicators may help explain some of

the misfit in the six-factor model. One possible explanation is

that the protective and understanding items tap into different

components of esteem enhancement. Recently, in a study of

older adnlts, Ranzijn, Keeves, Luszcz, and leather (1998) found

support for a hierarchial factor model of self-esteem in which

Global Self-Esteem was a second-order factor and Positive Self-

Regard and Usefulness/Competence were first-order factors. On

the one hand, protective and enhancement items may both assess

volunteering to enhance one's positive self-regard. For example,

the protective item, "Volunteering helps me work through my

own personal problems," and the enhancement item, "Volun-

teering makes me feel better about myself," may be indicators

of positive self-regard motivation for volunteering. On the other

hand, understanding and enhancement items may both assess

volunteering to enhance one's sense of usefulness and compe-

tence. For example, the understanding item, "I can explore my

own strengths," and the enhancement item, "Volunteering

makes me feel needed," may be indicators of usefulness and

competence motivation for volunteering.

Demographic Correlates of Volunteer Motives

In contrast to past research, we did not find age differences

in the understanding (Gidron, 1978) and protective (Stone &

Velrnans, 1980-1981) motives. The previously reported inverse

relation between age and the importance of the career motive

(Frisch & Gerrard, 1981) was observed only in the RSVP sam-

ple. Whereas earlier studies have shown that the social motive

was less important among older than younger volunteers (Fran-

cies, 1983), our results indicate that the social motive was more

important among older than younger volunteers. In part, these

discrepancies may be attributable to the truncated age range of

the participants in our study (minimum age = 50 years).

In both studies, female volunteers had higher protective mo-

tive scores than did male volunteers, a finding that is consistent

with the sex differences in volunteer motivation reported by

Morrow-Howell and Mui (1989). Also consistent with previous

studies of the protective motive, education was inversely related

to volunteering to escape negative feelings (Anderson & Moore,

1978).

In the MSVS (Marriott Senior Living Services, 1991), retired

volunteers had lower scores than employed volunteers on the

social and value dimensions of motivation to volunteer and mar-

ried volunteers had lower scores than single and divorced volun-

teers on the protective dimension of motivation to volunteer. In

the present study, we replicated the finding that married older

volunteers report being less motivated to volunteer to escape

negative feelings than single and divorced older volunteers.'

However, we did not find any evidence that older volunteers

who were working differed from older volunteers who were

retired on the social and value scales (see Footnote 1).

Limitations of the Study

Our study has several limitations. First, our response rates

were only 51% and 55% for the SMHSI and RSVP samples,

respectively. In addition, participants and nonparticipants in

each sample differed on some characteristics. For example, in

the RSVP sample, younger volunteers were proportionately

more likely to participate in the study than older volunteers; in

the SMHSI sample, male volunteers were proportionately more

likely to participate in the study than female volunteers.

Furthermore, when we tested the configural invariance of the

multifactor model, constraining model specifications to be the

same for each group while allowing coefficient values to vary,

' To directly compare the findings of the Marriott Seniors VolunteerismStudy (Marriott Senior Living Services, 1991) with the results of thepresent study, we conducted three contrasts. For each contrast, the analy-ses were done on the combined Retired and Senior Volunteer Programand Scottsdale Memorial Health Systems Incorporated samples. The firstcontrast compared single and divorced or separated volunteers with

married volunteers on their protective motive scores. The results indicatethat married volunteers (M = 3.01) scored significantly lower on protec-tive motive scores than single and divorced or separated volunteers (M= 3.61),F(1, 563) = 10.11, p < .003. The second and third contrastscompared retired volunteers with full- and part-time employed volun-teers on the Social and Values scale scores. Neither contrast was signifi-cant. For the Social scale scores, the means for retired volunteers andfor employed volunteers were 3.04 and 3.10, respectively, F(l, 550) =0.27, p > .59. For the Value scale scores, the means for retired volunteersand for employed volunteers were 5.51 and 5.69, respectively, F(l, 550)= 1.12, p > .29.

620 OKUN, BARR, AND HERZOG

the results were mixed.2 Some fit indices were within acceptable

ranges (i.e., the RMSEA was .049, ns), whereas others were

borderline (i.e., the CFI and the IFI were .88). As were the tests

of the multifactor model in each sample, the test of configural

invariance across samples was a strong test. As the number of

degrees of freedom in a model increases, the strength of the

model test increases, as do the "opportunities" for model misfit.

Thus, although the fit indices from the multifactor model with

390 dfa and the configural invariance model with 780 dfs varied

from borderline to acceptable, the general conclusion is that the

multifactor model is a moderately well-fitting model that has

been shown to generalize across different samples. In fact, in

the two separate samples used in the present study as well as

in samples of college students and middle-aged adults used in

another study (Clary et al., 1998), the six-factor, first-order

model of motivation to volunteer has been shown to provide an

adequate fit to the data.

A second limitation of our study pertains to our decision to

focus on active volunteers. It is an empirical question as to

whether the results would have been different had we focused

on older adults who were being recruited to volunteer for an

organization. This issue can be addressed in future research by

comparing the volunteer motives of active volunteers and people

potentially interested in volunteering or by conducting a longitu-

dinal study of the volunteer motives of people potentially inter-

ested in volunteering.

A third limitation is that we asked volunteers to rate how

important or accurate each of 30 reasons was for why they

engaged in volunteer work. This approach assumes that volun-

teers know the real reasons for their actions. Relatedly, reasons

may represent justifications for volunteer work that are gener-

ated on a post hoc basis as opposed to motivations that are

springboards for volunteer work. One way to validate the VFI

in future research would be to have potential volunteers com-

plete the VFI and then watch persuasive appeals for volunteering

that emphasize different motives (e.g., social vs. career). Indi-

viduals high in the career motive should exhibit greater arousal

as measured by skin conductance responses (Dawson, Schell, &

Filion, 1990) when they are exposed to the career-oriented as

opposed to the social-oriented persuasive appeal. Similarly, indi-

viduals high in the social motive should exhibit greater arousal

when they were exposed to the social-oriented as opposed to

the career-oriented persuasive appeal.

A fourth limitation is that we chose to focus on site-specific

assessment of motivation to volunteer and frequency of volun-

teering. It is possible that a global approach to assessment would

yield a different factor structure for motivation to volunteer or a

different pattern of correlations with frequency of volunteering.

Finally, we used items from a single scale as indicators of

volunteer motivation. Because all the items were assessed using

the same format, method variance may have contributed to the

pattern of findings observed from the factor analyses. In addi-

tion, the question arises as to whether support for the six-factor

model is contingent on using items from the VFI as indicators

of volunteer motivation. For example, would a factor analysis

of a measure with an equal number of altruistic and egoistic

items be more likely to provide support for the bipartite model

of volunteer motivation? This issue can be addressed in future

research by using multiple formats and by using indicators fromvarious motivation to volunteer measures.

Implications for Research

Compared with young and middle-aged adults, older adults

are much less motivated to volunteer by career concerns (Clary

et al., 1998). However, other work-related rewards may serve

as especially strong motives for volunteering by older adults.

Specifically, volunteering, as a type of unpaid work, can provide

opportunities to engage in activities that promote a sense of

usefulness (Fischer & Schaffer, 1993).

Midlarsky (1991) postulated that individuals are motivated

to help others, partly because of outcomes related to a sense of

competence. Because older adults in the United States have been

depicted as occupying a roleless role (Riley, Kahn, & Foner,

1994; Rosow, 1985), they may be more likely than younger

adults to volunteer to fulfill needs related to competence (Law-

ton, 1985). Therefore, it would be interesting to develop a mea-

sure that explicitly focuses on competence and usefulness as a

motive for volunteering. Researchers could use this measure to

determine whether (a) there are age differences in competence

and usefulness motive scores and (b) competence and usefulness

motive scores are stronger predictors of frequency of volunteer-

ing as age increases.

Implications for Practice

Motivation to volunteer among older adults is a multidimen-

sional construct composed of distinct but related dimensions.

Clary and his colleagues (Clary et al., 1998; Clary, Snyder,

Ridge, Miene, & Haugen, 1994) have demonstrated that college

students are more likely to intend to volunteer when they are

exposed to persuasive messages that are tailored to then- most

salient motive for volunteering. Volunteer coordinators seeking

to recruit older adult volunteers can use the VFI to ascertain

the importance of the various motives for volunteering in their

target population. For example, if our findings based on active

older volunteers are generalizable to older adults who are being

recruited to volunteer, then older adults are motivated to volun-

teer mostly by the values, understanding, and enhancement di-

mensions. This knowledge can be used to develop appeals to

older adults to volunteer that emphasize the opportunities volun-

teering provides for (a) acting on beliefs about the importance

of helping others; (b) learning about oneself and the world in

which one lives; and (c) feeling useful and good about oneself.

2 To meet the request of a reviewer, we tested the measurement equiva-

lence of our multifactor model across the Scottsdale Memorial Health

Systems Incorporated and Retired and Senior Volunteer Program sam-

ples. Note that researchers have been advised to perform such tests only

when they have developed a strong theoretical argument about how the

groups are expected to differ (Widaman & Reise, 1997).

References

Anderson, J. C., & Moore, L. (1978). The motivation to volunteer.

Journal of Voluntary Action Research, 7, 120-129.

Chambre, S. M. (1993). Volunteerism by elders: Past trends and future

prospects. The Geronlologist, 33, 221-228.

VOLUNTEER MOTIVATION 621

Clary, E. G., & Miller, J. (1986). Socialization and situational influences

on sustained altruism. Child Development, 57, 1358-1369.

Clary, E. G., & Snyder, M. (1990, March). A functional analysis of

volunteers'motivations. Paper presented at the meeting of the Spring

Research Forum, Boston.

Clary, E. G., & Snyder, M. (1991). A functional analysis of altruism

and prosocial behavior: The case of volunteerism. Review of Personal-

ity and Social Psychology, 12, 119-148.

Clary, E. G., Snyder, M., & Ridge, R. D. (1992). Volunteers' motiva-

tions: Strategies for the recruitment, placement, and retention of volun-

teers. Nonprofit Management and Leadership, 2, 330-350.

Clary. E. G., Snyder, M., Ridge, R. D., Copeland, J., Stukas, A. A.,

Haugen, J., & Miene, P. (1998). Understanding and assessing the

motivations of volunteers: A functional approach. Journal of Person-

ality and Social Psychology, 74, 1516-1530.

Clary, E. G., Snyder, M., Ridge, R. D., Miene, P. K., & Haugen, J. A.

(1994). Matching messages to motives in persuasion: A functional

approach to promoting volunteerism. Journal of Applied Social Psy-

chology, 24, 1129-1149.

Cnaan, R., & Goldberg-Glen, R. S. (1991). Measuring motivation to

volunteer in human services. Journal of Applied Behavioral Science,

27, 269-284.

Dawson, M. E., Schell, A. M., & Filion, D. L. (1990). Theelectroderrnal

system. In J. T. Cacioppo & L. G. Tassinary (Eds.), Principles of

psychophysiology: Physical, social and inferential elements (pp. 295-

324). Cambridge, England: Cambridge University Press.

Fischer, L. R., Mueller, D. P., & Cooper, P. W. (1991). Older volunteers:

A discussion of the Minnesota senior study. The Gerontologist, 31,

183-193.

Fischer, L. R., & Schaffer, K. B. (1993). Older volunteers: A guide to

research and practice. Newbury Park, CA: Sage.

Francies, G. R. (1983). The volunteer needs profile: A tool for reducing

turnover. Volunteer Administration, 16, 17-33.

Frisch, M. B., & Gerrard, M. (1981). Natural helping systems: Red

Cross volunteers. American Journal of Community Psychology, 9,

567-579.

Gidron, B. (1978). Volunteer work and its rewards. Volunteer Adminis-

tration, 11, 18-32.

Herzog, A. R., Franks, M. M., Markus, H. R., & Holmberg, D. (1998).

Activities and well-being in older age: Effects of self-concept and

educational attainment. Psychology and Aging, 13, 179-185.

Herzog, A. R., & House, J. S. (1991). Productive activities and aging

well. Generations, 15, 49-54.

Herzog, A. R., Kahn, R. L., Morgan, J. N., Jackson, J. S., & Antonucci,

T. C. (1989). Age differences in productive activities. Journal of Ger-

ontology: Social Sciences, 44, S129-S138.

Herzog, A. R., & Morgan, J. N. (1992). Age and gender differences in

the value of productive activities. Research on Aging, 14, 169-198.

Herzog, A. R., & Morgan, J. N. (1993). Formal volunteer work among

older Americans. In S. A. Bass, F. G. Caro, & Y.-P. Chen (Eds.),

Achieving a productive aging society (pp. 119-141). Westport, CT.

Auburn House.

James, L. R., Mulaik, S. A., & Brett, J. M. (1983). Causal analysis:

Assumptions, models, and data. Beverly Hills, CA: Sage.

Joreskog, K. G., & Sorbom, D. (1993). USREL (Version 8.12) [com-

puter software]. Chicago: Scientific Software.

Krause, N. (1993). Measuring religiosity in later life. Research on

Aging. 15, 170-197.

Latting, J. K. (1990). Motivational differences between Black and White

volunteers. Nonprofit and Voluntary Sector Quarterly, 19, 121-135.

Lawton, M. P. (1985). Activities and leisure. Annual Review of Geron-

tology and Geriatrics, 5, 127-164.

Luciani, J. (1993, November). Motivational determinants of volunteer

behavior: A logistic regression analysis using between-group and

within-group triangulation techniques. Paper presented at the annual

meeting of the Gerontological Society of America, New Orleans, LA.

Marriott Senior Living Services. (1991). Marriott Senior Volunteerism

Study. Washington, DC: Author.

Midlaisky, E. (1989, November). Helping and volunteering by the el-

derly. Paper presented at the annual meeting of the Gerontological

Society of America, Minneapolis, MN.

Midlarsky, E. (1991). Helping as coping. Review of Personality and

Social Psychology, 12, 238-264.

Morrow-Howell, N., & Mui, A. (1989). Elderly volunteers: Reasons

for initiating and terminating service. Journal of Gerontological So-

cial Work, 13, 21-34.

Okun, M. A. (1994). Relation between motives for organizational vol-

unteering and frequency of volunteering by elders. Journal of Applied

Gerontology, 13, 115-126.

Omoto, A. M., & Snyder, M. (1995). Sustained helping without obliga-

tion: Motivation, longevity, of service, and perceived attitude change

among AIDS volunteers. Journal of Personality and Social Psychol-

ogy, 68, 671-686.

Pinker, R. (1979). The idea of welfare. London: Heinemann.

Ranzijn, R., Keeves, J., Luszcz, M., & Feather, N. T. (1998). The role

of self-perceived usefulness and competence in the self-esteem of

elderly adults: Confirmatory factor analyses of the Bachman revision

of Rosenberg's Self-Esteem Scale. Journal of Gerontology: Psycho-

logical Sciences, 53B, P96-P104.

Riley, M. W., Kahn, R. L., & Foner, A. (Eds.). (1994). Age and struc-

tural lag: Society's failure to provide meaningful opportunities in

work, family, and leisure. New M>rk: Wiley.

Rosow. I. (1985). Status and role change through the life cycle. In R. H.

Binstock & E. Shanas (Eds.), Handbook of aging and the social

sciences (2nd ed., pp. 62-93). New TSjrk: Van Nostrand Reinhold.

Sainer, J. S., & Zander, M. L. (1971). SERVE: Older volunteers in

community service. New York: Community Service Society of New

York.

Stone, J., & Velmans, E. ( 1980-1981). Retirees as volunteers: Evalua-

tion of their attitudes and outlook. Volunteer Administration, 13, 4-

8.

Unger, L. S. (1991). Altruism as a motivation to volunteer. Journal of

Economic Psychology, 12, 71-100.

\an Til, J. (1988). Mapping the third sector: Voluntarism in a changing

social economy. New "York: Foundation Center.

Widaman, K. E, & Reise, S. P. (1997). Exploring the measurement

invariance of psychological instruments: Applications in the substance

abuse domain. In K. J. Bryant, M. Windle, & S. G. West (Eds.), The

science of prevention: Methodological advances from alcohol and

substance abuse research (pp. 281-324). Washington, DC: American

Psychological Association.

Wiehe, V. R., & Isenhour, L. (1977). Motivation of volunteers. Journal

of Social Welfare, 4. 73-79.

Received September 29, 1997

Revision received April 14, 1998

Accepted April 20, 1998 •