Institutional Predictors of Volunteer Retention

Abstract

Institutional predictors of volunteer service retention in AmeriCorps*State and National are assessed using data from the quasiexperimental, longitudinal study of AmeriCorps. Multilevel logistic regression is used, with individual member characteristics at the first level and institutional facilitation at the second level (across 107 program sites). Members are more likely to complete their terms of service if the program sites involve members in planning their service activities, match the activities to their career interests, help them develop a relationship with a mentor, and facilitate their reflection on the service experience with others. This study identifies strategies that may promote retention as an intermediate yet fundamental outcome.

Keywords

national service policy volunteerism

Introduction

The Edward M. Kennedy Serve America Act (2009) calls for the expansion of the United States’s flagship national service program, AmeriCorps. AmeriCorps has survived contested political tides since it was legislated in 1993 and now has been swept up in a new era of public service and civic engagement ushered in by the Obama administration. By 2017, the ranks of AmeriCorps’ volunteers, or “members” as they are called, are intended to swell from 75,000 to 250,000 (Edward M. Kennedy Serve America Act, 2009). As a decentralized national service scheme, members complete their volunteer service through sponsoring public and nonprofit organizations throughout the United States. This policy design has been analogized to a Swiss Army knife, serving multiple functions to achieve many goals at once (Perry, Thomson, Tschirhart, Mesch, & Lee, 1999): The members increase knowledge, skills, and positive behaviors (Frumkin et al., 2009; Perry & Katula, 2001; Perry & Thomson, 2004); the organizational partners have person-power (Perry & Thomson, 2004); and communities and the nation have needs met (Perry & Thomson, 2004; Vogenbeck, 2007). But these outcomes are compromised if members do not complete their terms of service.

Volunteer retention is a pressing issue facing volunteer programs nationwide. National population data reveal that volunteers are serving for shorter periods of time, and approximately a third who serve in one year do not continue to volunteer the following year (Corporation for National and Community Service [CNCS], 2007c). Retention issues span all types of volunteer roles, including national volunteer service with attrition rates variously calculated at state, regional, and national levels over the years, such as 39% across seven states in 1997 (U.S. General Accounting Office [GAO], 1997) to 21% at the national level in 2005 (U.S. Office of Management and Budget [OMB], 2005). When the service term is not completed, the sponsoring organization must complete the recruitment and preparation process again, which has financial and service delivery implications. Moreover, as with paid employment positions, high-volunteer turnover may demoralize remaining staff and volunteers and adversely affect those service beneficiaries with whom volunteers work directly (Miller, Powell, & Seltzer, 1990). The intended outcomes of the program are also compromised, as reduced time commitments (variously measured) lead to fewer benefits for the volunteers and compromised impacts of their service activities (Perry & Thomson, 2004). However, given these costs, we know little about how to increase retention as a crucial intermediate outcome in national volunteer service (Mesch, Tschirhart, Perry, & Lee, 1998; Perry & Thomson, 2004). Volunteer retention is not a given. It must be programmed or facilitated (Cnaan & Cascio, 1999).

This study uses public access data from the quasiexperimental, longitudinal study of service in AmeriCorps to assess institutional predictors of AmeriCorps service retention (CNCS, 2007b). The present study builds on and extends the only other published AmeriCorps retention study (Mesch et al., 1998). The goal of the present study is to identify malleable institutional facilitators of service retention that can inform AmeriCorps policy implementation and maximize impact. The rest of the article reviews relevant research, posits a theory on institutional facilitation of volunteer retention, presents the methods and findings, and discusses the results with implications for national volunteer service policy and programs.

Background

Institutional Features of AmeriCorps

AmeriCorps and its three distinct programs are operated by the independent federal agency, the CNCS (the Corporation). The programs are AmeriCorps*State and National (general service in full- and part-time placements), AmeriCorps*VISTA (poverty-focused capacity building in year-long placements), and AmeriCorps*NCCC (team-based, residential, full-time placements for 18- to 24-year-olds). AmeriCorps national volunteer service is considered a social institution that structures human behavior in an intensive, time-limited service commitment. The concern of this study is with how national service policy is implemented and how that implementation affects individual member participation. Other works have examined how service has emerged as an institution (Perry & Thomson, 2004; Sherraden, 2001a). The policy goals and institutional features are designed to be consistent at a national level, whereas programmatically, AmeriCorps is implemented locally through sponsoring public and nonprofit organizations throughout the nation (Perry et al., 1999). The local organizational sponsors must apply to be selected and once selected follow required procedures, including providing members with orientation, training, and supervision.

In this study, institutional features such as orientation, training, and supervision are conceptualized as institutional facilitators (Farmer & Fedor, 2000; McBride & Sherraden, 2007; McBride, Sherraden, Benitez, & Johnson, 2004; Morrow-Howell, Hinterlong, Rozario, & Tang, 2003; Morrow-Howell, Hinterlong, Sherraden, et al., 2003; Morrow-Howell, Hinterlong, Sherraden, & Rozario, 2001; Tang, Morrow-Howell, & Hong, in press). Institutional facilitators are defined as features that leverage volunteer participation by providing volunteers with the support they need to perform. Institutional facilitators may have a range of outcomes, which can be viewed across the volunteer engagement process—from creating access to the volunteer role, to keeping volunteers in the role, to promoting positive service outcomes, including increased service effectiveness and direct volunteer benefits (Sherraden, 2001b). There are other examples of possible institutional facilitators in AmeriCorps: During their service, eligible members may receive health benefits, a child care subsidy through AmeriCorps CARE, and a modest annual living allowance or stipend. These institutional features in particular are explicitly offered to promote access and retention among disadvantaged or otherwise civically disengaged populations (McBride, 2007).

The degree to which an institutional feature acts as a facilitator is an empirical question. In this study, the concern is with retention, and how variation in the offering and perceived quality of select features may have consequences for volunteer participation. When the impact of these features is known for national service, they can be controlled and maximized to improve the desired result, which is especially important in a policy context such as AmeriCorps when several hundred thousand individuals may eventually participate.

Possible Predictors of Volunteer Retention

Research in the area of volunteer retention is sparse and somewhat inconclusive. Volunteer retention is most commonly discussed in the negative as volunteer dropout or turnover. This is likely due to the costs associated with losing a volunteer midservice. Most studies explore a variety of relationships among potential salient factors (Omoto & Snyder, 1995). The outcome is variously measured as service duration (hours or weeks served), completion of a term of service, or intention or commitment to serve another year. Another aspect of research on retention is that most of the examined volunteer roles vary in terms of expectations as well as activities performed and beneficiaries. The majority of research examines volunteer roles with no defined duration, meaning that duration is not predetermined and is to be maximized. As such, predictors from one study may not apply to another (Cnaan & Amrofell, 1994), which speaks to the salience of this study of retention in national volunteer service. The following discusses potential predictors at the individual and institutional levels.

Individual factors

Research on various types of volunteer roles identifies sociodemographic factors as well as personal factors that are associated with some aspect of retention. National population data reveal that 66% of volunteers who return a 2nd year are older (CNCS, 2007c). A consistent finding across studies is that the more education volunteers have, the more likely they are to remain in the volunteer role (Lammers, 1991; Mesch et al., 1998; Wilson & Musick, 1999). Volunteer motivation to take on a particular volunteer role is a prominent factor in volunteer retention as well (Clary, Snyder, & Ridge, 1992; Omoto & Synder, 1993). Multiple studies have found that those who undertake the role for altruistic reasons, as opposed to instrumental reasons such as career building, are found to volunteer longer (Cheung, Tang, & Yan, 2006; Lammers, 1991; Mesch et al., 1998). Consistent with the so-called “virtuous circle” of civic action, those with prior volunteering experience may be more likely to continue volunteering as well (Gidron, 1985), though this finding is not consistent across studies (Cnaan & Cascio, 1999).

Institutional factors

In a sample of 1,753 nonprofit organizations, Hager and Brudney (2004) find that four out of five organizations rely on volunteers but do not manage them effectively. The structure put around the volunteer experience matters. The very presence of paid volunteer administrators is a good first step toward promoting volunteer retention (Cnaan & Cascio, 1999; Hager & Brudney, 2004). However, the nuances of volunteers’ connections to and impressions of their supervisors and other volunteers are perhaps more important. While most studies use differing measures, a consistent finding is that satisfaction with supervision is predictive of retention (Cheung et al., 2006). However, it is worthy of note that Hager and Brudney find an inverse relationship between regular supervision and retention. They speculate that this counterintuitive finding may be related to the troubled nature of the organizations’ volunteer management overall and not this particular strategy. Volunteering with other volunteers may be important for retention, although for varying reasons such as social acceptance, friendship, or other peer effects (Cnaan & Amrofell, 1994; Gidron, 1983; Haski-Leventhal & Cnaan, 2009).

The nature of the volunteer work assignment may matter as well. Although some studies have found that direct service work with individuals keeps volunteers serving (Gidron, 1985), others have found that professional or administrative assignments are associated with continued service (CNCS, 2007c). Regardless, others have found that the work should be challenging (Jamison, 2003), and the volunteer should feel like the task is being achieved (Gidron, 1985). Accordingly, Morrow-Howell and Mui (1989) and Tschirhart (1998) find that older adult volunteers drop out when they do not believe they are effective. Hager and Brudney (2004) find that nonprofit organizations have higher rates of retention when they offer training and professional development opportunities, formally screen and match volunteers to tasks, and recognize volunteers. Orientation and training is a prominent predictor of retention (Gidron, 1985), with studies suggesting that preservice and in-service training are important (Jamison, 2003; Persson, 2004). A study on stipend receipt by older adults in national service found that those who receive stipends serve longer and are more likely to complete their term of service than those who do not receive a stipend (McBride, Gonzales, Morrow-Howell, McCrary, & Lee, in press). Stipend receipt may both incentivize and facilitate completion, especially for those with limited financial resources.

Research Hypothesis

We hypothesize that institutional features facilitate retention in AmeriCorps, beyond the impact of individual features. Categorical completion of the term of service is used as the dependent measure of retention. A term of service in AmeriCorps may be part-time or full-time in intensity and from 9 to 12 months in duration. Members may opt to serve for one additional term at a maximum. This time commitment differs from other volunteer programs, which may be less intensive and not time limited. The nature and facilitators of retention may differ as well; the time frame of the role is compressed, thus, so is the curve for effectively matching the individual to the institution and retaining the volunteer. At the institutional level, we examine the impact of orientation, training, member involvement in planning, volunteer activity and interest matching, mentorship, supervision, and reflection. At the individual level, controlling for other factors, we examine the specific impact that higher education and prior work and volunteer experience have on retention. Through multilevel modeling, we are able to compare the relative impact that variables at the two levels have on retention (Yanay & Yanay, 2008). As is described in more detail below, the AmeriCorps longitudinal study supports this analytical test of our hypothesis because it was conducted across more than 100 program sites and the data about facilitators at each site can be aggregated from individual members’ responses.

Method

Data Source

This study uses data from the quasiexperimental, longitudinal study of service in AmeriCorps, which was sponsored by the Corporation and implemented by Abt Associates between 1999 and 2002, the data for which are available from the Corporation. Surveys were administrated to the member and comparison groups in four waves at baseline (1999-2000), postprogram (2000-2001), postprogram follow-up (2003), and longitudinally (2007). The initial study sample included a nationally representative group of approximately 2,200 full-time, 1st-year members from more than 100 AmeriCorps*State and National programs and 3 AmeriCorps*NCCC programs. The AmeriCorps*VISTA program was not included. About 2,000 nonmembers who demonstrated interest but did not participate in AmeriCorps comprised the comparison group, the use of which is not relevant to the topic of the present study. A program director’s survey was administered to directors at 100 AmeriCorps program sites, but that survey was not used for this study due to the lack of relevant variables and limited variation on those that were relevant. Systematic missing data on the dependent variable obviated use of survey responses from the AmeriCorps*NCCC program. The final longitudinal survey was also not used because the outcome variable of interest in this study is an intermediate outcome, measured immediately following program participation through the postprogram survey.

The resulting sample includes AmeriCorps members who participated in the AmeriCorps*State and National program. Among the 1,752 AmeriCorps*State and National members who participated in the baseline survey, 1,385 members completed the postprogram survey, which included items of interest to this study. We excluded 191 members who were still enrolled in the program at the time of the postprogram survey because their eventual completion status could not be confirmed. Due to missing data across other study variables, the final sample size is limited to 1,087 AmeriCorps*State and National members.¹

Measures

Dependent variable

The dependent variable in this study is a dichotomous measure of AmeriCorps completion obtained from the postprogram survey. Members self-reported as either completing their term of service (1) or not (0).

Covariates

Both individual and institutional variables are utilized as covariates in this study. Individual-level variables include sociodemographic characteristics (age, gender, race/ethnicity, marital status, having children, disability status, education, and income) and prior experience (prior employment experience and prior volunteering experience). These individual-level variables are drawn from the baseline survey.

The race/ethnicity variable is recoded from six categories into four, including Hispanic, African American, White, and Others (American Indian, Asian, or Native Hawaiian or other Pacific Islander). The marital status variable is treated dichotomously (1 = married, 0 = unmarried). The unmarried category includes AmeriCorps members who are never married, widowed, divorced, or separated. Presence of children below age 18 living in the household is treated dichotomously (1 = yes, 0 = no), as is disability status. Educational attainment is a 10-category measure ranging from 1 = 8th grade or less to 10 = PhD, MD or other professional degree, which is treated continuously. Personal income is a 14-category measure ranging from 1 = less than US$5,000 to 14 = US$100,000 or more, which is treated continuously as well. Prior employment and prior volunteering are treated dichotomously (1 = yes, 0 = no).

Institutional facilitation is measured by eight items from the postprogram survey and one item from the postprogram supplement survey. Measures are chosen that are consistent with a conceptualization of facilitation as institutional support that leverages volunteer participation. There are two items measuring orientation, attendance at orientation (1 = yes, 0 = no), and days of attendance at orientation (empirical range = 0-30). Frequency of member training is measured across seven categories that are numerically ordered and treated continuously from 1 to 7 (never, one to two times per year, three to four times per year, monthly, every other week, weekly, and daily). Frequency of members’ involvement in planning their service activities is measured using a 5-point Likert-type scale ranging from 1 (never) to 5 (always).

The degree of relationship between service activities and long-term career interests is measured using a three-level scale ranging from “not at all” to “a lot,” which is reverse coded so that a higher number is positive. Two items assessing the relationship between members and their mentors and members and their supervisors are included as institutional predictors. Development of a meaningful relationship with a mentor is a dichotomous measure (1 = yes, 0 = no). Members’ satisfaction with their supervisory relationship is measured by a 5-point Likert-type scale ranging from “very satisfied” to “very dissatisfied,” which is reverse coded so that a higher number represents more satisfaction. Reflection on the service experience is measured by two items including solitary reflection through journaling or writing about the experience and talking about it with other volunteers, friends, or relatives. These two items are treated continuously and measured using a three-level scale ranging from “never” to “frequently.”

Other relevant measures

Two additional items are relevant for this study and are reported descriptively. They include an item that asked the main reason a member did not complete the entire term of service and another that asked those with disabilities to report how well their needs related to their disability were accommodated.

Analytical Methods

Descriptive statistics and corresponding tests of difference by group (completers vs. leavers) are computed for relevant variables. Multilevel logistic regression is used to examine predictors of AmeriCorps service completion, with individual member characteristics at the first level and institutional facilitation at the second level. The data used in this study have a hierarchical structure as the members are clustered at the program site-level. In these clustered data, members within the same program are more likely to be correlated, and a failure to incorporate these within-cluster correlations into the analytic model often leads to underestimation of standard errors (Raudenbush & Bryk, 2002; Snijders & Bosker, 1999). Multilevel analysis corrects this problem and allows for assessment of the relative importance of the higher level context in relation to the outcome. Logit link is used in the analytic model because the outcome variable is a binary measure of AmeriCorps completion.

In this study, we analyze three nested models. In Model 1, we estimate an empty (or unconditional) model with no predictors; the probability of AmeriCorps completion is only a function of random intercepts at the program site-level. The empty model is important because it provides the basic partition of the variance in the outcome variable between two levels. If the magnitude of program-level variance is substantial in the empty model, the use of multilevel analysis is justified. In Model 2, we include individual-level predictors. In Model 3, we add the institutional, program site–level variables to assess the relative effect of each institutional feature on AmeriCorps completion. Three models used in this study are presented below. For all three models, the random intercept (u₀ _j ) that has a mean of zero and variance of τ₀₀ is allowed to vary across the program sites to correct for the clustering effect.

Model 1 (unconditional or empty model):

\begin{array}{l} Level 1 (individual level) : logit (π_{i j}) = β_{0 j} . \\ level 2 (program site-level) : β_{0 j} = γ_{00} + u_{0 j} . \\ \log i t (π_{i j}) = γ_{00} + u_{o j}, u_{o j} ~ N (0, τ_{00}) . \end{array}

Model 2 (including only individual-level predictors):

\begin{array}{l} Level 1: logit (π_{i j}) = β_{o j} + \sum β_{k} X_{k i j} . \\ Level 2 : β_{0 j} = γ_{00} + u_{0 j} . \\ logit (π_{i j}) = γ_{00} + \sum β_{k} X_{k i j} + u_{0 j}, u_{0 j} ~ N (0, τ_{00}) . \end{array}

Model 3 (including both individual-level and program site–level predictors):

\begin{array}{l} Level 1:logit (π_{i j}) = β_{0 j} + \sum β_{k} X_{k i j} . \\ Level 2 : β_{0 j} = γ_{00} + \sum γ_{0 m} Z_{m j} + u_{0 j} . \\ logit (π_{i j}) = γ_{00} + \sum β_{k} X_{k i j} + \sum γ_{0 m} Z_{m j} + u_{0 j}, u_{0 j} ~ N (0, τ_{00}), \end{array}

where X_kij is kth individual-level predictors for individual i in program site j, Z_mj is the mth institutional predictors for program site j, β _k ₀ is a coefficient for kth individual-level variable, γ₀ _m is a coefficient for mth institutional variables, γ₀₀ is an average log odds of service completion across program sites, u₀ _j is random effects (random intercepts) at Level 2 (program site-level), and τ₀₀ is variance of random effects at a Level 2 (program site-level).

To demonstrate the importance of program site–level effects, we decompose the total variance into within and between program sites and calculate the intraclass correlation (ICC) for each model. The ICC represents the proportion of program site–level variance calculated as dividing program site–level variance by total variance for a continuous outcome variable. However, in multilevel logistic regression, we should use alternative calculations of the ICC because the individual-level variance and program site–level variance are not directly comparable as the former is on the probability scale, whereas the latter is on the logistic scale (Goldstein, Browne, & Rasbash, 2002; Larsen & Merlo, 2005). Considering these technical difficulties, we chose the alternative threshold method suggested by Snijders and Bosker (1999). According to them, the ICC for multilevel logistic regression could be calculated as follows:

ρ = τ_{00} / (τ_{00} + π^{2} / 3),

where τ₀₀ is the second-level variance.

An advantageous alternative for the ICC in the multilevel logistic regression is the median odds ratio (MOR), which measures program site–level variance in the odds ratio (OR) scale (Larsen, 2006; Larsen & Merlo, 2005). The MOR is calculated as follows: $MOR = \exp (z_{0.75} \times \sqrt{2 σ^{2}}),$ , where z_0.75 is the value of standard normal distribution at 75th percentile, and σ² represents program site–level variance. If the MOR is equal to 1, there would be no difference between the program sites in explaining the probability of AmeriCorps completion (no program site–level variance), but the higher the MOR value, the more important the program site–level effects. We calculated both the ICC and MOR in the model to assess the relative importance of institutional, program site–level effects.

The interval odds ratio (IOR) also introduced by Larsen and Merlo (2005) can be used to interpret the influence of each program site–level predictor. The IOR allows us to consider program site–level residual variations when interpreting the program site–level predictors. The IOR is calculated as follows: $I O R_{U p p e r} = \exp (β + z_{0.90} \times \sqrt{2 σ^{2}})$ and $I O R_{L o w e r} = \exp (β + z_{0.90} \times \sqrt{2 σ^{2}})$ where β is a regression coefficient for each program site–level predictor, σ² represents program site–level variance, and z_0.90 and z_0.10 are the values for standard normal distribution at 90th and 10th percentile, respectively. If the IOR for a specific program site–level predictor does contain 1, the effect of that predictor is not important in comparison with the unexplained heterogeneity between sites. The opposite is true, if the IOR is narrow and does not contain 1. Therefore, the IOR adds complementary information to the conventional odds ratio to interpret the importance of the institutional variables.

Results

Descriptive Results

Descriptive statistics for the sample are presented in Table 1. The sample includes 1,087 members from 107 AmeriCorps*State and National program sites. Approximately, 70% of the AmeriCorps members in this sample completed their terms of service, which is consistent with the national volunteer retention average (CNCS, 2007c). Looking at the demographic composition of the sample, members range in age from 17 to 80 with an average age of 28. Totally, 75% are female. Eighteen percent are married, and 39% have at least one child living in the household. Seven percent of the respondents report having a disability. In terms of racial and ethnic composition, the sample is about half White and one-quarter African American. In all, 20% of the respondents have a high school degree, and another 43% have a college degree or above. A majority have prior experience in employment and volunteering.

Table 1.

Characteristics of a Sample of AmeriCorps Members by Service Completion Status (N = 1,087)

Individual-level categorical variables	Completers frequency (%)	Leavers frequency (%)	Total	Chi-square
Gender
Female	591 (76%)	223 (73%)	814 (75%)	χ²(df = 1) = 0.71
Male	191 (24%)	82 (27%)	273 (25%)
Race/ethnicity
African American	185 (24%)	103 (34%)	288 (27%)	χ²(df = 3) = 13.00**
Hispanic	126 (16%)	49 (16%)	175 (16%)
White	407 (52%)	128 (42%)	535 (49%)
Others	64 (8%)	25 (8%)	89 (8%)
Marital status
Married	135 (17%)	65 (21%)	200 (18%)	χ²(df = 1) = 2.39
Unmarried	647 (83%)	240 (79%)	887 (82%)
Have children
Yes	258 (33%)	169 (55%)	427 (39%)	χ²(df = 1) = 46.23***
No	524 (67%)	136 (45%)	660 (61%)
Disability status
Yes	36 (5%)	36 (12%)	72 (7%)	χ²(df = 1) = 18.39***
No	746 (95%)	269 (88%)	1,015 (93%)
Prior employment
Yes	769 (98%)	294 (96%)	1,063 (98%)	χ²(df = 1) = 3.84
No	13 (2%)	11 (4%)	24 (2%)
Prior volunteering
Yes	676 (86%)	238 (78%)	914 (84%)	χ²(df = 1) = 11.60***
No	106 (14%)	67 (22%)	173 (16%)
	Completers	Leavers	Total
Individual-level continuous variables	M (SD)	M (SD)	M (SD)	t test
Age	28.28 (10.47)	28.91 (10.89)	28.46 (10.59)	t = −0.87, df = 1,085
Income	5.30 (3.36)	5.04 (3.41)	5.22 (3.37)	t = 1.11, df = 1,085
Education	6.95 (2.37)	5.87 (2.51)	6.65 (2.46)	t = 6.67***, df = 1085

p < .01. ***p < .001.

Bivariate tests examined any differences between completers and leavers in terms of demographics and other covariates. According to the results presented in Table 1, there are no statistical differences between completers and leavers in age, gender, marital status, prior employment, and income. However, completion is related to not having a child in the household and not having a disability (χ² = 46.23, df = 1, p < .001 and χ² = 18.39, df = 1, p < .001, respectively). In terms of racial and ethnic composition, completion is associated with being White (χ² = 13.00, df = 3, p < .001). Completion is also associated with advanced education (t = 6.67, p < .001) and prior experience volunteering (χ² = 11.60, df = 1, p < .001).

The 377 members who did not complete their terms of service were asked why. The primary self-reported reason is for personal and/or health reasons (38%), followed by dissatisfaction with the program (20%), financial reasons (15%), to take a job (8%), because they were asked to leave the program (5%), and other (5%). Those with disabilities have federal protections for their accommodation. With that said, members with a disability report (n = 76) varying levels of accommodation with the majority considering the accommodation to be excellent or good (33% and 32%, respectively) and fewer considering it to be fair or poor (14% and 21%, respectively).

Institutional facilitation variables were aggregated to the program site–level, based on individual members’ perceptions by site. Descriptive statistics for these variables are presented in Table 2. All variables are treated continuously at this level as they represent averages across all members, and in this table, across all sites. The average levels of perceived institutional facilitation tend toward the positive across the program sites. One exception is the members’ level of satisfaction with their supervisors with an average value of two, indicating “somewhat dissatisfied.”

Table 2.

AmeriCorps Members’ Perceptions of the Institutional Facilitation of Their Service Experience (n = 107 program sites)

	M (SD)	Minimum	Maximum
Attended orientation	0.92 (0.17)	0	1
Number of days of attendance at orientation	5.76 (4.52)	0	30
Brought all members together for a meeting or training	4.85 (0.85)	1	7
Member involvement in planning service activities	3.71 (0.54)	1	5
Level of relationship between activities and career or job interests	2.52 (0.30)	1	3
Development of a relationship with a mentor	0.63 (0.22)	0	1
Level of satisfaction with supervisor	2.00 (0.69)	1	5
Reflection on AmeriCorps experience by keeping a journal or writing	2.25 (0.41)	1	3
Reflection on AmeriCorps experience by talking with others	2.79 (0.18)	1	3

Note: The response options for each variable differ.

Multilevel Model Results

The Model 1 (empty or unconditional model) provides the partition of variance in AmeriCorps retention between two levels. The program site–level variance is 0.61 and significantly different from 0. The ICC of 0.16 indicates that about 16% of total variance is due to program site–level variance. Furthermore, the MOR is higher than 1 (MOR = 2.11) indicating that the program site–level variance is not negligible. Both the ICC and MOR reveal that between-program variance is important to understand AmeriCorps service completion and justify modeling the data in line with a multilevel framework (see Table 3).

Table 3.

Multilevel Logistic Regression Results (Random Intercept Model): Individual and Institutional Predictors of Retention in AmeriCorps*State and National (N = 1,087)

	Model 1 (empty)		Model 2 (with individual-level predictors)			Model 3 (with program site–level predictors)
	EST	SE	EST	SE	Odds ratio	EST	SE	Odds ratio
Individual-level variables
Age			.026*	.013	1.03	.026*	.012	1.03
Female			.205	.182	1.23	.216	.183	1.24
Race/ethnicity (White)
Hispanic			−.083	.359	0.92	−.114	.333	0.89
African American			−.294	.293	0.75	−.241	.274	0.79
Others			−.060	.301	0.94	.117	.342	1.12
Married			−.063	.288	0.94	−.239	.304	0.79
Having a child(ren)			−.921***	.242	0.40	−1.025***	.242	0.36
Having a disability			−1.086**	.343	0.34	−1.233***	.307	0.29
Education			.134***	.034	1.14	.118**	.034	1.13
Income			−.002	.027	1.00	−.013	.028	0.99
Prior employment			.691	.522	2.00	.504	.527	1.66
Prior volunteering			.143	.204	1.15	.103	.219	1.11
Program site–level variables
Attended orientation						−1.448	1.002	0.23
IOR						[0.13 ~ 0.42]
Days of orientation						.016	.027	1.02
IOR						[0.57 ~ 1.80]
Cohort meeting or training						.039	.115	1.04
IOR						[0.59 ~ 1.84]
Involvement in planning						.640***	.181	1.90
IOR						[1.07 ~ 3.36]
Relation to career/job interest						1.522**	.451	4.58
IOR						[2.58 ~ 8.13]
Relationship with mentor						1.931**	.648	6.90
IOR						[3.89 ~ 12.23]
Satisfaction with supervisor						.233	.267	1.26
IOR						[0.71 ~ 2.24]
Solitary reflection						−.013	.278	0.99
IOR						[0.56 ~ 1.75]
Reflection with others						1.282*	.601	3.60
IOR						[2.03 ~ 6.39]
Random effect
Program site–level variance	0.61		0.44			0.10
Intraclass correlation	0.16		0.12			0.03
Median Odds Ratio	2.11		1.88			1.33

Note: EST = parameter estimates; IOR = interval odds ratio.

p < .05. **p < .01. ***p < .001.

In Model 2 where only individual-level variables are included, several predictors are statistically significant. As hypothesized, education is positively related to completion: For one categorical increase in the education measure, the likelihood of completion increases by 14% (OR = 1.14, p < .001). Prior experience is not predictive. However, age is positively associated with AmeriCorps completion, holding constant other variables and program site–level random effects (OR = 1.03, p < .05). Conversely, members who have children are about 60% less likely to complete their terms of service than those who do not have children (OR = 0.40, p < .001), and those who have any kind of disability are about 66% less likely than those who do not have a disability (OR = 0.34, p < .01).

Turning to Model 3, which includes both individual and institutional variables, the results for the individual-level variables are consistent with those presented in Model 2. Over and above the individual-level variables, four institutional, program site–level variables are statistically associated with the log odds of AmeriCorps completion. Members who are highly involved in planning their service activities have higher odds of completion (OR = 1.81, p < .001). The degree of relationship between AmeriCorps activities and long-term career or job interests is positively associated with AmeriCorps service completion (OR = 4.58, p < .01). Development of a relationship with a mentor is positively associated with AmeriCorps completion (OR = 6.90, p < .01). Of the two variables measuring reflection on the AmeriCorps experience, reflection by talking with others is positively associated with completion (OR = 3.60, p < .05).

The IORs for the four statistically significant institutional variables are narrow and do not contain 1 within the intervals. This further reveals that the effects of those four institutional variables (involvement in planning service activities, relation to long-term career or job interests, development of a relationship with a mentor, and reflection with others) are significantly large relative to the unexplained program-level effects. The importance of these institutional facilitators for retention is also reflected in the ICC and MOR in Model 3. The ICC and MOR are substantially reduced after including institutional, program site–level predictors in Model 3 as compared with those in Models 1 and 2, which demonstrates that a substantial proportion of program-level variance is accounted for by the institutional facilitation features assessed in this study.

Discussion

AmeriCorps is a national volunteer service policy that is implemented at the local level. Variation in implementation is inevitable but not without consequence. The results of this study suggest that retention of AmeriCorps members can be explained in part by this institutional variation across sites. Members are more likely to complete their terms of service if the program sites involve the members in planning their service activities, match the volunteer activities to the members’ career interests, help members develop a relationship with a mentor, and facilitate members’ reflection on the service experience with others. Consistent with an institutional perspective, these features leverage member participation. But these institutional facilitators are not part of the provisions of the AmeriCorps legislation. In this way, they are not required, though they may be crucial to the effectiveness and efficiency of the policy.

Prior to discussing the implications of these findings, study limitations are worthy of note. Results are only generalizable to individuals serving in the AmeriCorps*State and National program during the time period of the study. The surveys and items measuring institutional facilitation were not developed with this theory in mind. As such, better measures may exist. Not all items were measured consistently or with the highest levels of measurement, thus, variation is minimal on some items. Moreover, although the measures assess higher level, structural features, they were still measured at the individual level and aggregated to the program site-level. It would have been ideal to have objective measures on each facilitator from each program site. Related, measures of program completion are self-report, though social desirability bias is unlikely to have had a major effect as the attrition rate for this sample is similar to national rates. Another limitation of the current study is the possibility of reverse causality as facilitation variables were measured based on self-reports from individuals at the same time as the dependent variable. Finally, there are likely unmeasured individual and program characteristics that are important predictors of retention.

Individual Predictors and Inclusion

One of the goals of AmeriCorps is to provide the opportunity to serve to those who desire it. As such, provisions are made to remove barriers through health benefits, child care subsidies, and stipends. However, in this study, those who have a disability and those with children are statistically less likely to complete their terms of service than those without a disability and those without children. About a third of those with a disability report separately that their disabilities were not accommodated well (poor or fair) while they were in the program. When all respondents who left before completing their term of service were asked why, the overwhelming response was “personal and/or health reasons.” These data suggest that barriers to participation remain. Completion by all who serve is not realistic. However, if there are systemic issues with facilitation of completion that relate to members’ inherent characteristics, then more must be done to foster inclusion. Program sites should be sensitive to and yet diligent about potential barriers and ensure that members know what institutional resources and supports are available to them.

Related, advanced education is associated with retention, which is consistent with other research. However, those with less education may be the ones who stand to benefit the most from participation in volunteer service (Spring, Dietz, & Grimm, 2007). Additional research is needed to ascertain the specific motivations and participation challenges for this population. They may particularly benefit from a more explicit link between their national volunteer service activities and employment possibilities postservice, especially given the recent economic downturn. Although this begs the history of national volunteer service and its contested function as a “jobs program” (Sherraden & Eberly, 1982), this may be a reality faced by a portion of those who are interested in or who serve in AmeriCorps. This also begs the question of fit between the individual and the role. Some attrition is inevitable and even desired, as those individuals who do not perform well or do not thrive—in spite of facilitation—may sacrifice the program’s end outcomes by staying merely to complete their term of service.

Matching the Institution to Individuals

One can infer from the results that the members who enrolled in AmeriCorps during this time period had instrumental expectations regarding their service experience, meaning that beyond any goals of community impact, they had expectations regarding their personal skill and knowledge development. In short, they expected to be “facilitated.” Research is emerging that identifies differences between the generations in their motivations to serve and what they expect from the experience (CNCS, 2007a). The primary age demographic in this sample and in AmeriCorps overall is young adults. Consistent with life course theories regarding civic engagement, these respondents are on the cusp of Generation X and the Millennial Generation. They may view their service as a developmental experience to help launch their vocational and career pursuits (Morrow-Howell & Tang, 2007).

This type of generational analysis is important, especially as the Serve America Act calls for overall increases in membership but also increased membership among Baby Boomers. It speaks to a potential need to tailor the institutional support to the individual member. Although the young adult may want activities matched to their career interests as well as a career mentor, the Baby Boomer may be more interested in substantive training on the issue being targeted, to further her or his lifelong learning (Morrow-Howell & Tang, 2007). This is speculation but the point remains that there may be larger generational forces at play here, which could be amplified by increased outreach and enrollment of older adults.

Modifying the Model

These findings suggest that the institutional model for AmeriCorps may need to standardize strategies such as mentoring and reflection, especially for those members who aim to connect the experience to their career development. On a large scale, these strategies may be viewed as time-consuming and costly but so is volunteer turnover. How can such seemingly “high touch” facilitators be taken to scale? National service may have something to learn from service learning. Service learning pairs learning with community service in a reflective academic context. It is increasingly implemented at primary, secondary, and postsecondary levels and may be the most institutionalized form of service in the United States (Billig & Eyler, 2003).

National service is implemented in a community context but without an explicit learning component. This study suggests that facilitated learning promotes national service retention. Service-learning programs and courses have found that the volunteers’ service experiences can be “transformative” in terms of life purpose and career pursuits when they integrate and reflect across their experiences (Eyler, 2002). This outcome is facilitated, almost always through structured dialogue with program leaders, supervisors, and/or other volunteers. The CNCS embraces service learning, and their online National Resource Center has resources on service learning with suggested models for reflection in AmeriCorps specifically. However, as this study demonstrates, the integration of these models is not widespread or uniform. The promotion, use, and impact of these resources by the managers of AmeriCorps members should be examined. More widespread technical assistance may be warranted.

Maximizing Minimal Provisions

Prior research suggests that orientation, training, and supervision are predictive of volunteer service retention. However, as noted previously, the volunteer field is diverse, and one study does not necessarily generalize to other programmatic contexts or populations. For this national service program and corresponding sample, these provisions are not statistically significant predictors. However, this does not mean that they are unnecessary for retention. They may be merely basic. Moreover, the Corporation may have done an effective job at establishing this baseline and resourcing the program sponsors to provide these supports and to do so consistently. Member selection, orientation, training, and supervision can also be explicitly linked to the facilitators that are found to be predictive. For example, member retention may be maximized by screening applicants on what they expect from their experience, or as part of orientation, the individual members can plan some aspect of their tasks, and as part of training, the members can be allowed to dialogue and reflect on their experience as a group. Pairing strategies and integrating these findings throughout the volunteer engagement process can maximize resources, helping to take these predictive facilitators to scale.

To retain AmeriCorps members and thus meet the projected member goal levels called for in the 2009 Serve America Act may require more universal implementation of institutional features that leverage the members’ development, including ownership over planning activities, activity and interest matching, mentorship, and group-based reflection. Although these features are not inconsistent with the intent or design of the policy or volunteer management overall, they require the implementation of provisions and further development of national resources that increase the capacity of all program sponsors to retain members. In short, we should not go into the expansion of AmeriCorps with business as usual. The consequences of member turnover are costly if we do. This study suggests that efforts to promote retention in national service should not generalize from research on volunteerism overall but be tailored to this unique volunteer service role. Widespread adoption of these institutional features may increase the effectiveness and efficiency of national volunteer service policy.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The author acknowledges the Ford Foundation for support of the analysis of these data.

Notes

Bios

Amanda Moore McBride, PhD, is an associate professor at the George Warren Brown School of Social Work, director of the Gephardt Institute for Public Service, and research director of the Center for Social Development, Washington University in St. Louis. Her research focuses on service learning and national and international volunteer service.

YungSoo Lee, MA, is a doctoral candidate at the George Warren Brown School of Social Work, Washington University in St. Louis. His research interests include the intersection of social policy, family and child well-being, and civic engagement and volunteering.

References

Billig

S. H.

Eyler

(2003). The state of service-learning and service-learning research. In Billig

S. H.

Eyler

(Eds.), Deconstructing service-learning: Research exploring context, participation, and impacts (pp. 253-264). Greenwich, CT: Information Age Publishing.

Cheung

F. Y. L.

Tang

C. S. K.

Yan

E. C. W.

(2006). Factors influencing intention to continue volunteering: A study of older Chinese in Hong Kong. Journal of Social Service Research, 32, 193-209.

Clary

E. G.

Snyder

Ridge

(1992). Volunteers’ motivations: A functional strategy for the recruitment, placement, and retention of volunteers. Nonprofit Management and Leadership, 2, 333-350.

Cnaan

R. A.

Amrofell

(1994). Mapping volunteer activity. Nonprofit and Voluntary Sector Quarterly, 23, 335-351.

Cnaan

R. A.

Cascio

T. A.

(1999). Performance and commitment: Issues in management of volunteers in human service organizations. Journal of Social Service Research, 24, 1-37.

Corporation for National and Community Service. (2007a). Keeping baby boomers volunteering: Volunteer retention and turnover, research brief. Washington, DC: Corporation for National and Community Service.

Corporation for National and Community Service (2007b). Serving country and community: A study of service in AmeriCorps. Washington, DC: Corporation for National and Community Service.

Corporation for National and Community Service. (2007c). Volunteer retention, issue brief. Washington, DC: Corporation for National and Community Service.

Edward

Kennedy Serve America Act of 2009, H.R. 1388, 111th United States Cong. (2009).

10.

Enders

C. K.

Bandalos

D. L.

(2001). The relative performance of full information maximum likelihood estimation for missing data in structural equation models. Structural Equation Modeling, 8, 430-457.

11.

Eyler

(2002). Reflection: Linking service and learning-linking students and communities. Journal of Social Issues, 58, 517-534.

12.

Farmer

S. M.

Fedor

D. B.

(2000). Volunteer participation and withdrawal: A psychological contract perspective on the role of expectations and organizational support. Nonprofit Management and Leadership, 9, 349-367.

13.

Frumkin

Jastrzab

Vaaler

Greeney

Grimm

R. T.

Cramer

Dietz

(2009). Inside national service: AmeriCorps’ impact on participants. Journal of Policy Analysis and Management, 28, 394-416.

14.

Gidron

(1983). Sources of job satisfaction among service volunteers. Journal of Voluntary Action, 12, 20-35.

15.

Gidron

(1985). Predictors of retention and turnover among service volunteer workers. Journal of Social Service Research, 18, 1-16.

16.

Goldstein

Browne

Rasbash

(2002). Partitioning variation in generalized linear multilevel models. Understanding Statistics, 1, 223-232.

17.

Graham

J. W.

Hofer

S. M.

Mackinnon

D. P.

(1996). Maximizing the usefulness of data obtained with planned missing value patterns: An application of maximum likelihood procedures. Multivariate Behavioral Research, 31, 197-218.

18.

Hager

M. A.

Brudney

J. L.

(2004). Volunteer management practices and retention of volunteers. Washington, DC: The Urban Institute.

19.

Haski-Leventhal

Cnaan

R. A.

(2009). Group processes and volunteering: Using groups to enhance volunteerism. Administration in Social Work, 33, 61-80.

20.

Jamison

I. B.

(2003). Turnover and retention among volunteers in human service agencies. Review of Public Personnel Administration, 23, 114-132.

21.

Lammers

J. C.

(1991). Attitudes, motives, and demographic predictors of volunteer commitment and service duration. Journal of Social Service Research, 14, 125-140.

22.

Larsen

(2006, November 24). New measures for understanding the multilevel logistic regression model. Seminar conducted at the workshop “Statistische Methoden fur korrelierte Daten” in Bochum, Denmark.

23.

Larsen

Merlo

(2005). Appropriate assessment of neighborhood effects on individual health: Integrating random and fixed effects in multilevel logistic regression. American Journal of Epidemiology, 161, 81-88.

24.

McBride

A. M.

(2007). Civic engagement, older adults, and inclusion. Generations, 30, 66-71.

25.

McBride

A. M.

Gonazles

Morrow-Howell

McCrary

(in press). Stipended volunteer civic service: Inclusion, retention, and volunteer benefits. Public Administration Review.

26.

McBride

A. M.

Sherraden

(2007). Civic service worldwide: Impacts and Inquiry. Armonk, NY: M.E. Sharpe.

27.

McBride

A. M.

Sherraden

Benitez

Johnson

(2004). Civic service worldwide: Defining a field, building a knowledge base. Nonprofit and Voluntary Sector Quarterly, 33, 8S-21S.

28.

Mesch

D. J.

Tschirhart

Perry

J. L.

Lee

(1998). Altruists or egoists? Retention in stipended service. Nonprofit Management and Leadership, 9, 3-21.

29.

Miller

L. E.

Powell

G. N.

Seltzer

(1990). Determinants of turnover among volunteers. Human Relations, 43, 901-917.

30.

Morrow-Howell

Hinterlong

Rozario

P. A.

Tang

(2003). Effects of volunteering on the well-being of older adults. Journal of Gerontology, 58, S137-S145.

31.

Morrow-Howell

Hinterlong

Sherraden

Rozario

P. A.

(2001). Advancing research on productivity in later life. In Morrow-Howell

Hinterlong

Sherraden

(Eds.), Productive aging: Concepts and challenges (pp. 285-312). Baltimore, MD: Johns Hopkins University Press.

32.

Morrow-Howell

Hinterlong

Sherraden

Tang

Thirupathy

Nagchoudhuri

(2003). Instructional capacity for elder service. Social Development Issues, 25, 189-204.

33.

Morrow-Howell

Mui

(1989). Elderly volunteers: Reasons for initiating and terminating service. Journal of Gerontological Social Work, 13, 21-35.

34.

Morrow-Howell

Tang

(2007). Youth service and elder service in comparative perspective. In McBride

A. M.

Sherraden

M. W.

(Eds.), Civic service worldwide: Impacts and inquiry (pp. 157-180). Armonk, NY: M.E. Sharpe.

35.

Omoto

A. M.

Snyder

(1995). Sustained helping without obligation: Motivation, longevity of service, and perceived attitude change among AIDS volunteers. Journal of Personality and Social Psychology, 68, 671-686.

36.

Perry

J. L.

Katula

M. C.

(2001). Does service affect citizenship? Administration & Society, 33, 330-365.

37.

Perry

J. L.

Thomson

A. M.

(2004). Civic service: What difference does it make? Armonk, NY: M.E. Sharpe.

38.

Perry

J. L.

Thomson

A. M.

Tschirhart

Mesch

Lee

(1999). Inside a Swiss army knife: An assessment of AmeriCorps. Journal of Public Administration Research and Theory, 9, 225-250.

39.

Persson

D. I.

(2004). Volunteer ombudsmen in the nursing home: Obstacles to retention. Journal of Aging Studies, 18, 205-214.

40.

Raudenbush

S. W.

Bryk

A. S.

(2002). Hierarchical linear models: Applications and data analysis methods (2nd ed.). Thousand Oaks, CA: SAGE.

41.

Schlomer

G. L.

Bauman

Card

N. A.

(2010). Best practices for missing data management in counseling psychology. Journal of Counseling Psychology, 57, 1-10.

42.

Sherraden

(2001a). Civic service: Issues, outlook, institution building. St. Louis, MO: Washington University, Center for Social Development.

43.

Sherraden

(2001b). Youth service as strong policy (CSD Working Paper 01-12). St. Louis, MO: Washington University, Center for Social Development.

44.

Sherraden

Eberly

(Eds.). (1982). National service: Social, economic, and military impacts. New York, NY: Pergamon Press.

45.

Snijders

T. A. B.

Bosker

R. J.

(1999). Multilevel analysis: An introduction to basic and advanced multilevel modeling. Thousand Oaks, CA: SAGE.

46.

Spring

Dietz

Grimm

(2007). Youth helping America: Leveling the path to participation: Volunteering and civic engagement among youth from disadvantaged circumstances. Washington, DC: Corporation for National & Community Service.

47.

Tang

Morrow-Howell

Hong

S. I.

(in press). Institutional facilitation in sustained volunteering among older volunteers. Social Work Research.

48.

Tschirhart

(1998). Understanding the older stipended volunteer: Age-related differences among AmeriCorps members. Public Productivity and Management Review, 22, 35-48.

49.

U.S. General Accounting Office. (1997). National service programs: Role of state commissions in implementing the AmeriCorps program. Washington, DC: U.S. Government Printing Office.

50.

U.S. Office of Management and Budget. (2005). Detailed information on the AmeriCorps state and national grants assessment. Retrieved from http://www.whitehouse.gov/omb/expectmore/detail/10000324.2005.html

51.

Vogenbeck

D. M.

(2007). National service impacts on nonprofit community networks. Research report 07-1. St. Louis, MO: Center for Social Development, Washington University in St. Louis.

52.

Wilson

Musick

M. A.

(1999). Attachment to volunteering. Sociological Forum, 14, 243-272.

53.

Yanay

G. V.

Yanay

(2008). The decline of motivation? From commitment to dropping out of volunteering. Nonprofit Management and Leadership, 19, 65-78.