Can Incentive Effects in Web Surveys Be Generalized to Non-Western Countries? Conditional and Unconditional Cash Incentives in a Web Survey of Ghanaian University Students

Abstract

Because research on the impact of web survey incentives has exclusively focused on Western settings, it is unclear to what extent current insights translate and generalize to non-Western societies, which are usually characterized by very different economic conditions, cultural traditions, and survey climates. The current article presents the results of a web survey incentives experiment among almost 4,440 Ghanaian university students who were offered conditional and unconditional incentives of different values (in the form of telephone credit). Our analyses partly replicate Western findings: Higher value incentives produce higher participation rates and unconditional incentives outperform conditional ones in the lower value conditions. In the case of relatively high incentives, however, conditional outperforms unconditional incentives. No differential effects of incentives on response quality were found.

Keywords

incentives experiment conditional versus unconditional incentives nonresponse response quality web survey Ghana

In a few years, research into the impact of incentives on survey participation will celebrate its centenary. As early as 1931, the first quasi-experimental study on including a 25-cent coin in a mail survey was published by Shuttleworth (1931). Over the decades, countless survey experiments have gathered sound empirical evidence that incentives can be a powerful tool for increasing response rates (for meta-analyses, see Auspurg & Schneck, 2014; Church, 1993; ; Edwards et al., 2002; Göritz, 2006b; Hopkins & Gullickson, 1992; Singer, Van Hoewyk, Gebler, Raghunathan, & McConagle, 1999; Yu & Cooper, 1983). In a nutshell, available research convincingly shows that (1) incentives increase response rates, irrespective of the survey mode used, (2) unconditional (prepaid) incentives are more effective than conditional (promised) incentives, (3) monetary incentives are more effective than vouchers and in-kind incentives, and (4) the higher value of an incentive, the stronger is the effect it produces, albeit marginal returns diminish (Singer & Ye, 2013). More recently, researchers have also started enquiring the effect of incentives on indicators of response quality, notably on item nonresponse (Curtin, Singer, & Presser, 2007; Davern, Rockwood, Sherrod, & Campbell, 2003; Medway, 2012).

Although the quantity of empirical evidence is impressive, the available research is geographically—and as a result also culturally—very strongly clustered. The vast majority of studies were carried out in the United States or—to a lesser extent—in other Anglo-Saxon countries, such as Canada, the United Kingdom, or Australia, and a few Western European countries. This raises the question to what extent current knowledge regarding incentive effects can be translated cross-nationally (Blom, Jäckle, & Lynn, 2010; Couper & de Leeuw, 2003). After all, “the meaning of gifts as well as monetary incentives is likely to differ between cultures” (Singer et al., 1999, p. 219), and the scarce available comparative research suggests that even within the Anglo-Saxon world, incentive effects show cross-national variation (Mutti, Kennedy, Thompson, & Fong, 2014). The generalizability of incentive effects beyond the Western world raises even greater question marks, in particular, in the context of web surveys. How effective are incentives in web surveys in non-Western societies that are characterized by very different economic conditions, cultural traditions, and survey climates?

The present study contributes to this field by analyzing an incentives experiment implemented in the National Service Scheme Survey (N3S; Langer, Meuleman, Oshodi & Schroyens, 2016), a web survey conducted in August 2014 among 4,440 Ghanaian university students who were offered conditional and unconditional incentives of different values. The incentives were quasi-cash, in the form of top-up credits for mobile phones, a typical means of transferring small amounts of money in Ghana. We examine whether and to what extent the web survey incentives have an impact on survey participation and response quality. To the best of our knowledge, this study is the first to report the results of a web survey incentives experiment in an African country. It, therefore, sheds light on whether the well-documented findings from Western countries can be transferred to non-Western settings.

This article commences with a brief review of the theory and previous research on the impact of incentives on response rates and data quality in Western countries. In the theory section, we also elaborate on the specificity of the Ghanaian context and derive research hypotheses. Subsequently, we describe the design of the study and present our results. We conclude with a discussion of the implications of our study.

Theory and Hypotheses

Effects of Incentives on Response and Response Quality: Theoretical Perspectives

In the context of unit nonresponse, leverage-salience theory stresses that decisions to accept or refuse partaking in a survey depend on a wide array of factors—such as topic, sponsorship, expected time investment, concerns about privacy, perceived benefits, and community involvement—that have “different ‘leverages’ on the cooperation decision for different persons” (Groves, Singer, & Corning, 2000, p. 306). One of these levers that can be relatively easily operated by survey designers is giving incentives to respondents. Numerous studies have successfully used incentives—varying in form (monetary vs. in kind) and conditionality (unconditional vs. conditional; Church, 1993)—to stimulate survey response.

To explain the effectiveness of incentives on response rates, two theoretical frameworks that stress different mechanisms—namely, self-interest versus norms of reciprocity—compete for attention (Singer & Ye, 2013). First, according to rational choice-based approaches, such as exchange theory (Homans, 1950) or the theory of reasoned action (Fishbein & Ajzen, 1975), individual behavior is rooted in rational cost–benefit calculations (Alwin, 1991, pp. 17–18; Philipson, 1997; Singer, 2011). Providing an incentive to respondents adds to the benefit side and consequently increases the probability that a person will take part in the survey. This rational choice approach stresses the importance of the monetary value and conditionality of incentives. The more valuable an incentive is, the higher the probability that it can tip the balance in favor of survey participation. The incentive should only be provided conditional upon completion of the survey. Once handed over, the incentive disappears from the cost–benefit equation, and it is no longer in the self-interest of the respondent to cooperate.

A second approach contradicts the premise that the effectiveness of incentives can be fully understood by referring to the realm of the economic only. Providing incentives is not a neutral transaction of economic value but establishes a social relation of trust and appeals to the principle of the gift (Mauss, 1923). Incentives establish a norm of reciprocity: They create a moral obligation to reciprocate the favor by participating in the survey (Gouldner, 1960). Even small symbolic values can be effective, and the incentive should be given prior to participation in the survey (i.e., unconditionally; Dillman, 2000).

Incentives do not only affect the decision to participate in a survey but could also have an impact on response behavior during the survey. Reasoning along the two theoretical frameworks presented above, contrasting expectations have been uttered (Singer & Ye, 2013). Some scholars have expressed the concern that incentives might have detrimental effects on response quality because they persuade sample units who would otherwise have refused to respond. Assuming that incentives affect the decision to participate but not the respondents’ motivation, these respondents lack internal motivation and might therefore not fill out the survey carefully, which leads to incomplete data of low quality (Davern et al., 2003; Deutskens, De Ruyter, Wetzels, & Oosterveld, 2004; Göritz, 2006b). Hansen (1980, p. 78) uses self-perception theory to explain such quality-reducing effect of incentives: External motivators (such as incentives) impede respondents from perceiving themselves as cooperative persons, which results in less effort during the survey. If, however, incentives activated a norm of reciprocity, their effectiveness would not be confined to the cooperation decision but would instead extend across the whole survey. In this case, incentives would stimulate respondents to provide more complete and adequate information and improve response quality (Davern et al., 2003).

Empirical Evidence From the Western World

In the Western world (mostly the United States), the theoretical propositions on incentive effects have been tested elaborately among the general population as well as more specific samples—such as adolescents (Jäckle & Lynn, 2008), university students (Shettle & Mooney, 1999; van Veen, Göritz, & Sattler, 2016), or even earthquake engineers (Birnholtz, Horn, Finholt, & Bae, 2004). Several reviews (Simmons & Wilmot, 2004; Singer & Ye, 2013) and meta-analyses (Armstrong & Lusk, 1987; Church, 1993; Göritz, 2006b; Heberlein & Baumgartner, 1978; Singer et al., 1999; Yu & Cooper, 1983) reveal a general pattern that confirms components of both theoretical frameworks. Incentives tend to motivate respondent cooperation, irrespective of whether the survey is interviewer mediated (Singer et al., 1999), administered via mail (Church, 1993) or web (Göritz, 2006b), and of whether data collection is cross sectional or longitudinal (Jäckle & Lynn, 2008).

The exact effectiveness of an incentive, however, depends on its form, conditionality, and value. Prepaid (unconditional) incentives are found to be considerably more effective at increasing response rates than promised (conditional) incentives (Singer et al., 1999; Yu & Cooper, 1983), even when keeping the overall costs equal (e.g., Blom, Gathmann, & Krieger, 2015). Some studies even report that promised incentives do not yield higher response rates than no incentive comparison groups (e.g., Church, 1993). Equally in line with the reciprocity argument is that incentives with small, symbolic value already increase response substantially. Higher values, however, tend to produce stronger response-inducing effects, which can be understood from rational choice theory. Empirical evidence on the functional form of the relationship between incentive value and response rates is mixed. While some studies report that response increases linearly with incentive values (Church, 1993; Gelman, Stevens, & Chan, 2003; Yu & Cooper, 1983), others report diminishing marginal returns (Dillman, 2000; Singer et al., 1999) and are no longer effective beyond a certain threshold (James & Bolstein, 1992). Finally, cash incentives have stronger motivating effects than incentives in kind of the same value (Singer et al., 1999; Yu & Cooper, 1983), which supports the economic argument.

The effect size of incentives also depends on the survey mode. Regarding unit nonresponse, incentives tend to be most effective in mail surveys. Church’s (1993) meta-analysis reveals that prepaid incentives of small value (mean value $1.38) increase response rates by 19.1 percentage points, which implies a 65% bump in response rate (effect size d = .347). Singer, Van Hoewyk, Gebler, Raghunathan, and McConagle (1999) report that, across 39 interviewer-mediated surveys, every dollar of incentive paid yields more or less one third of a percentage point in response rate. A possible explanation for this considerably weaker effect might be that interviewer-mediated surveys generally have higher baseline response rates.

With regard to web surveys, a meta-analysis by Göritz (2006b) shows that the odds of starting and completing a web survey increase with 19% and 27%, respectively (corresponding to effect sizes d of .10 and .13), when an incentive is provided. The comparatively weak effects in web surveys might be related to the difficulty of establishing trust over the Internet combined with the fact that it is practically difficult to provide cash incentives to online respondents. Web-suitable unconditional incentives (such as payments through PayPal or vouchers for online purchases) often require relatively lengthy procedures and high levels of Internet literacy (Bosnjak & Tuten, 2003) or an offline prerecruitment phase, where addresses of sample units, to which the unconditional incentive can be sent, are collected (Scherpenzeel & Toepoel, 2012). The few studies on unconditional cash incentives (Bosnjak & Tuten, 2003; Scherpenzeel & Toepoel, 2012) show mixed results.

Whereas there is abundant evidence for response increasing effects of (unconditional monetary) incentives, fewer studies look into the relationship between incentives and response quality. Often-used indicators of response quality are (the absence of) item nonresponse or the length of answers in open-ended questions (Singer & Ye, 2013). Various studies report that the use of incentives does not affect item nonresponse (e.g., Curtin et al., 2007; Davern et al., 2003; Göritz, 2004; Petrolia & Bhattacharjee, 2009; Teisl, Roe, & Vayda, 2005), while others find that incentives motivate respondents to complete all questions (Bosnjak & Tuten, 2003; James & Bolstein, 1990; Medway, 2012; Singer, van Hoewyk, & Maher, 2000). In any case, the available research does not support the argument that giving out incentives harms response quality. If they have any effect on response quality, incentives seem to strengthen rather than to undermine respondents’ motivation. The effect of different types of incentives on response quality remains unclear.

Potential Differences in the Ghanaian Context

A crucial limitation in the field is that available research tends to be restricted to the United States and other Anglo-Saxon countries. Although recent efforts have been undertaken to generalize the findings to other Western-European countries (e.g., to Germany; see Pforr et al., 2015) or Russia (Avdeyeva & Matland, 2013), studies that test the effectiveness of incentives beyond the Western world—in particular with respect to web surveys—are very scarce.

In this study, we explore to what extent current knowledge on incentive effects can be generalized to non-Western contexts, in particular to the case of web survey incentives in Ghana. Due to the specificity of the Ghanaian context, there are plausible theoretical reasons to expect that web survey incentives may produce different effects from those in Western contexts (for similar arguments, see Avdeyeva & Matland, 2013). In particular, we anticipate that differences in economic conditions, the cultural traditions and customs, and the general survey climate in Ghana might have an impact on the effectiveness of incentives. First, concerning the economic conditions, it is worthwhile to highlight here that Ghana is a developing country with a Gross National Income (GNI) per capita of US$1,480 (Atlas Method, current US$) and an estimated poverty head count ratio of 24.2% in 2015 (World Development Indicators). While university students (i.e., our research population) generally constitute a relatively privileged section of Ghanaian society, a considerable proportion of them nonetheless faced serious socioeconomic hardship during their youth and quite a few students continued to have difficulties making ends meet. Illustratively, 39.3% of students interviewed as part of our survey mentioned that during their childhood, they had gone without food “occasionally” or “frequently” because their family lacked money. In addition, 62.6% of interviewed students mentioned that they had struggled in past years to pay for their university education. Given these difficult and uncertain economic circumstances and living conditions, we would expect from a rational choice or exchange theory perspective that people would be more willing to participate in activities through which they can generate additional income. Furthermore, compared to the relative standard of living, the incentives used are quite large, and as a result, the conditional monetary incentives may well have stronger effects in Ghana compared to Western settings (Avdeyeva & Matland, 2013, p. 176).

Second, important cultural differences may also contribute to differences in incentives effects. Traditionally, Ghanaian society has been characterized by a strong informal system of reciprocal exchange between family members, friends, neighbors, ethnic group members, and so on, as a necessary means of survival (Hydén, 2006). Furthermore, Ghana has been identified as one of the most religious countries worldwide (Ruiter & Van Tubergen, 2009), which may positively influence societal norms concerning helping one’s fellow man and reciprocating favors, gifts, and assistance. These societal norms may in turn make people more inclined to participate in surveys upon receiving an unconditional monetary incentive. Furthermore, Ghana was found to be one of the least-trusting societies worldwide (Delhey, Newton, & Welzel, 2011), which might not only decrease response rates but also undercut the effectiveness of conditional incentives (Avdeyeva & Matland, 2013).

A third factor which may affect the impact of incentives on survey participation is the survey climate, that is, the societal-level factors that contribute to the willingness to participate in surveys (Groves & Couper, 1998, p. 155). In this respect, it is important to note that prospective respondents in developing countries are not overburdened with surveys to the same degree as prospective respondents in Western settings. Consequently, we would expect relatively higher response rates in these non-Western settings, which in turn, however, may be associated with weaker incentives effects.

Finally, one factor that may mediate the impact of incentives on survey participation in a Ghanaian web survey context is the difference in Internet infrastructure found in most sub-Saharan African countries compared to Western countries. While the Internet in Europe and Anglo-Saxon countries developed on the back of a landline telephone infrastructure, sub-Saharan countries rely more heavily on mobile phone technologies not only for phone communication but also for the Internet (Aker & Mbiti, 2010; de Bruijn, 2009). Figure 1 shows that in our survey of Ghanaian university students, more than half of the respondents (53.4%) used mobile devices to participate in the web survey. In comparison, during the same period in Germany, only 19.9% of the participants aged 20–25 in the German Internet Panel used mobile devices (Blom et al., 2016). At the same time, methodologists have found mobile web participants less likely to respond to survey requests (e.g., de Bruijne & Wijnant, 2013; Mavletova, 2013). Since sample members, however, were invited to the survey via both e-mail and text message, the mobile mode is made particularly attractive. Incentives effects for mobile web surveys are largely unexplored even in Western countries (cf. Mavletova & Couper, 2016). Given the specific mobile survey setting in African countries, the effects of conditional and unconditional cash incentives may differ from effects found in Western countries.

Figure 1.

Participation on mobile and computer devices in the National Service Scheme Survey of Ghanaian students.

Hypotheses

Due to the relative scarcity of previous research on survey incentives in non-Western contexts, the character of our study is necessarily exploratory, and contrasting hypotheses can be derived. Building upon the findings obtained in a Western context, we expect that unconditional incentives yield higher response rates than conditional ones (Hypothesis 1), that higher incentive values will lead to higher response rates (Hypothesis 2a), and that additional value increases have diminishing returns in terms of response (Hypothesis 2b). Response quality will be higher in conditions with unconditional incentives (Hypothesis 3) and with incentives of higher values (Hypothesis 4). Due to specific contextual factors, however, the effectiveness of incentives in Ghana may differ from Western contexts. On the one hand, because of the unfavorable economic context, conditional incentives could possibly be more effective than unconditional ones in Ghana (Hypothesis 5). Finally, the favorable survey climate could lead to a high baseline response rate, with little differences between the various incentive conditions (Hypothesis 6).

Data and Method

Data Set: The N3S

The N3S surveys prospective participants in Ghana’s National Service Scheme (NSS; Langer et al., 2016). The NSS is a compulsory national service program, which requires all Ghanaian students who obtain a bachelor degree to complete 1 year of national social service, usually outside their area of origin. The purpose of the N3S is to study the extent to which Ghana’s NSS program contributes to improving intergroup relations and fostering stronger national identities among its participants. To gather information before and after participation in the NSS, we implemented a three-wave online panel of two cohorts of Ghanaian students. We sampled third- and fourth-year bachelor students from three public universities with a wide ethno-regional spread (University of Ghana in Legon, Kwame Nkrumah University of Science and Technology in Kumasi, and University for Development Studies with campuses at Tamale, Nyanpkala, Wa, and Navrongo) and a variety of bachelor programs—agriculture, administration, chemistry, computer sciences, economics/development studies, engineering, English, nursing, geography, and politics (see Figure 1 for details on the sampling process).

In a first stage, we visited a mandatory third- and fourth-year course for each of the bachelor programs. At the beginning of the class, we asked students to fill out a four-page paper-and-pencil questionnaire. The main purpose of this in-class survey was to obtain e-mail addresses and phone numbers from the students so that they could be invited to participate in the web survey. The other questions in this background questionnaire give valuable information on the students, which we use in the analyses below. In total, 5,825 students completed the questionnaire. Only a very small number of students refused to participate in the survey either by not accepting a questionnaire or by handing it back empty.

In a second stage, the 5,570 students who filled out the background questionnaire were invited by e-mail and text message to participate in an online survey. The incentives experiment reported here concerns participation in this second stage of the survey, that is, the first web survey.

This two-stage procedure has the considerable advantage that background information is available for respondents as well as nonrespondents in Stage 2. A drawback is that our results do not concern the whole population of university students but are conditional on participation in the first stage. However, based on information provided by the professors about class attendance and the low refusal rate, we estimate to have covered 80–90% of all students enrolled in the selected programs. Two hundred and fifty five students did not provide a valid e-mail address, and 1,130 invitation e-mails bounced,¹ leaving us with a sample of 4,440 students. Our final sample consists of 64.6% male students and has a mean age of 23.29 years (standard deviation = 2.84).

Experimental Design

Cash incentives—especially prepaid ones—are notoriously difficult to deliver to respondents (Wetzels, Schmeets, van den Brakel, & Feskens, 2008), especially in web surveys (Bosnjak & Tuten, 2003; Couper, 2000). However, the Ghanaian social context offered an interesting opportunity: quasi-cash incentives as mobile phone credit, so-called top-up. In Ghana, top-ups are easily transmittable and used almost universally as a mode of economic exchange. Top-ups thus are a functional equivalent of cash in Ghana.

Our experimental design varies both the value and the conditionality of cash incentives (see Figure 2). Regarding the value of incentives, students either received a mobile top-up of 5, 10, or 20 Ghanaian Cedi’s (GHS; 50%, 25%, and 25% of the sample, respectively). These amounts correspond to approximately €1.20, €2.40, and €4.80. Five GHS is the price of a dinner in a campus restaurant, while 20 GHS buys a bus ticket to travel from Accra to Kumasi (roughly 250 km). Given that university students are a privileged section of Ghanaian society, the incentives we offered were relatively small compared to their spending patterns. With respect to conditionality, the three value conditions were fully crossed. For students receiving an unconditional incentive, the mobile phone top-up—together with a text message indicating that the top-up was meant as a token of gratitude for their participation in the survey—was sent to the respondents’ phones just before they received the survey invitation by e-mail. Respondents in a conditional incentive condition received a text message saying that they would receive a top-up upon completing the survey. Note that the experimental design does not contain a no incentive comparison group. The reason is that, given the previous research in the Western world, there were strong expectations that giving no incentives at all would lead to lower response, which could have damaged the data quality of the N3S research project.

Figure 2.

Design of the National Service Scheme Survey—including distribution of students and classes over the incentive conditions.

Because classmates are likely to communicate with each other and differences in amounts might be perceived as unfair (Singer & Ye, 2013), we decided to randomize classes of students enrolled in the same bachelor program rather than individual students over the six experimental conditions. In order to obtain a balanced distribution with respect to relevant covariates, 64 class groups were allocated to the six conditions (see Figure 2) using covariate-adaptive randomization (Hu, Hu, Ma, & Rosenberger, 2014). Specifically, the size of the group, the bachelor year, the university, and the discipline were taken into account in this randomization. As a consequence, the experimental group sizes in terms of individuals differ slightly from the projected sizes.

Indicators

In the context of a web survey, it is useful to distinguish between three indicators of survey participation. A first indicator is the starting rate, that is, the percentage of the invited sample units that started the survey (i.e., clicked on the invitation link and loaded at least the first page of the survey). Second, we look at the completion rate (for a similar approach, see Göritz, 2006a). This is the percentage of starters that actually complete the survey (i.e., reach the final survey page). As such, completion is the counterpart of dropout. The distinction between starting rate and completion rate makes it possible so disentangle incentive effects on different states of the survey process. Third, we analyze the response rate, defined as the percentage of invited sample units that completes the survey (i.e., reaches the last page). The response rate is a general indicator that combines incentive effects of the different phases in the survey process (starting and completing) and reveals the impact of incentives up to the end of the questionnaire.

Response quality is operationalized by means of two concrete indicators. First, this study includes the most often-used measure of response quality in incentive research, namely, the prevalence of item nonresponse. Concretely, we calculate the percentage of missing answers over a set of 85 items that were presented to all respondents (thus not included in routings or filters). Second, we analyze the occurrence of so-called straightlining, that is, of selecting the same response option for a set of items that are presented jointly in a matrix question, such that the responses form a straight vertical line on the screen. Straight lining is often seen as an indicator of satisficing (Kaminska, McCutcheon, & Billiet, 2010). The N3S contains 11-item batteries, each consisting of multiple items that use the same answer scale and are presented on the same screen. For each respondent, we calculated the percentage of scales on which straightlining behavior was observed. To avoid confounding response quality and survey participation, the indicators of response quality were calculated only for respondents who completed the survey (N = 2,892).

In order to control for the differential composition of the class groups, the model controlled for the following background variables: gender, age (in years), ethnicity (i.e., a self-classification into the major ethnic groups in Ghana), university, year of study (third- vs. fourth-BA year), and the answer on the question whether they look forward to their national service (on a scale from 0—not at all to 10—very much). Especially, the latter two variables are of interest. Given fourth-year students were invited to take the survey 4 weeks before the start of their national service, and given that the national service is the topic of the N3S, these variables indicate topic salience and attitude toward the topic of the survey. Descriptive statistics for all variables used can be found in the Appendix.

Results

Effects of Incentives on Survey Participation

Table 1 and Figure 3 present the starting, completion, and response rates for each of the six incentive conditions. First, at face value, the incentive conditions show substantively relevant variation in starting rates. The difference between the condition with the lowest proportion of starters—as expected the promised 5 GHS condition—and the condition with the highest starting rate (prepaid 10 GHS) is 14 percentage points, and the corresponding odds ratio (OR) equals 1.91. Notwithstanding a notable exception, the pattern of differences in starting rates confirms current knowledge. Generally speaking, unconditional incentives seem to be more successful in motivating respondents to start the survey: For each of the values, the starting rate for the prepaid incentive is higher than the one for the promised incentive (although the difference in the 20 GHS conditions is marginal). Second, higher amounts indeed seem to lead to higher starting rates. At the same time, the starting rates suggest diminishing returns to value increases. For the conditional incentives, moving from 10 to 20 GHS yields about the same increase in starting rate as moving from 5 to 10 GHS, while the value increase is twice as large. In case of the unconditional incentives, moving from 10 to 20 GHS even leads to a decrease in the starting rate. It is not clear whether this surprising result is due to sampling fluctuations or reveals a substantively relevant pattern. James and Bolstein (1992) also report that their highest prepaid incentive—a $40 check included in a mail survey—results in lower response than lower denominations. James and Bolstein (1992, p. 447) refer to Dillman (1978) to argue that incentives can become too large: “The closer the monetary incentive comes to the value of the service performed, the more the transaction tends to move into the realm of economic exchange and the easier it becomes for many people to refuse it” ( p. 16).

Table 1.

Indicators of Survey Response and Data Quality by Incentives Condition.

Condition	Starting		Completion		Response		Item Nonresponse (%)	Straightlining (%)
Condition	Rate	OR	Rate	OR	Rate	OR	Item Nonresponse (%)	Straightlining (%)
Prepaid 5 GHS	64.99	1.28	86.53	0.81	56.24	1.16	4.36	20.12
Prepaid 10 GHS	73.49	1.91	89.30	1.05	65.63	1.72	3.51	19.95
Prepaid 20 GHS	69.53	1.57	92.74	1.60	64.48	1.63	3.78	18.87
Promised 5 GHS	59.25		88.85		52.64		4.15	20.28
Promised 10 GHS	63.23	1.18	93.18	1.71	58.91	1.29	3.27	21.11
Promised 20 GHS	68.31	1.48	94.58	2.19	64.61	1.64	4.68	20.73
N	4,440		2,892		4,440		2,660	2,660

Note. OR = odds ratio, compared with the theoretically worst incentive condition (conditional 5 GHS). GHS = Ghanaian Cedi.

Figure 3.

Starting, completion, and response rates by condition.

The completion rates reveal a very different dynamic. First, completion rates tend to be slightly higher for the promised than for the prepaid incentives (although differences are very small). Apparently, prepaid incentives are more successful in stimulating Ghanaian students to start the survey, while promised incentives are a slightly stronger motivating factor not to drop out. This seems to contradict a meta-analysis of Western web surveys that found that prepaid and promised incentives have a very similar effect on completion (Göritz, 2006b, p. 64). Second, the differences in completion rates are neatly aligned with the value of the incentives: The conditions with higher incentives show higher completion rates.

The combined effect of both processes—starting and not dropping out—is visible in the overall response rate. Three conditions, namely, prepaid 10 GHS, prepaid 20 GHS, and promised 20 GHS, produce very similar response rates and outperform the worst condition (promised 5 GHS). The remaining two conditions—prepaid 5 GHS and promised 10 GHS—associate with medium response rates.

However, the results presented in Table 1 and Figure 2 should be interpreted with great caution. First, it remains to be seen to what extent the reported differences in survey cooperation are statistically significant. And second, the reported figures are flat averages, without taking into account the fact that class groups—with varying compositions in terms of individuals’ characteristics—are randomized over the experimental conditions rather than individuals. In order to calculate the significance of the incentive effects, taking into account the clustered nature of the data into account and controlling for differential compositions of classes, we make use of multilevel analysis (Hox, 2010). Table 2 presents the results of multilevel logistic regression models for each of the three dichotomous indicators of survey cooperation (namely, start, completion, and response). The models reveal statistically significant differences in the three indicators of survey cooperation between the incentives conditions, although controlling for individual characteristics leads to certain changes in the pattern of differences.

Table 2.

Multilevel Logistic Regression Models for Starting the Survey, Completion, and Overall Response (Estimated Odds Ratios and Their 95% CIs).

	Model 1: Starting Rate		Model 2: Completion Rate		Model 3: Response Rate
	Exp(B)	95% CI	Exp(B)	95% CI	Exp(B)	95% CI
Intercept	6.16***	[3.01, 12.61]	6.20*	[1.37, 28.13]	3.64***	[1.81, 7.34]
Incentive
Prepaid 5 GHS	1.23	[0.94, 1.62]	1.01	[0.71, 1.43]	1.21	[0.94, 1.56]
Prepaid 10 GHS	1.63**	[1.17, 2.28]	0.85	[0.56, 1.30]	1.48*	[1.09, 2.01]
Prepaid 20 GHS	1.43*	[1.07, 1.93]	1.39	[0.88, 2.18]	1.47**	[1.12, 1.95]
Promised 5 GHS (ref. cat.)				[1.02, 2.92]
Promised 10 GHS	1.09	[0.80, 1.49]	1.72*		1.17	[0.88, 1.57]
Promised 20 GHS	2.00***	[1.33, 3.00]	3.15***	[1.74, 5.69]	2.37***	[1.63, 3.45]
Gender
Male	1.39***	[1.20, 1.61]	1.32*	[1.01, 1.73]	1.41***	[1.23, 1.63]
Female (ref.)
Age	0.94***	[0.92, 0.97]	1.01	[0.96, 1.07]	0.95***	[0.93, 0.97]
Ethnicity
Ga-Dangme	0.97	[0.78, 1.22]	1.12	[0.73, 1.72]	1.00	[0.80, 1.24]
Ewe	1.16	[0.95, 1.41]	1.42	[0.95, 2.11]	1.24*	[1.02, 1.50]
Guan	0.79	[0.54, 1.16]	1.33	[0.57, 3.14]	0.87	[0.60, 1.27]
Gurma	1.19	[0.79, 1.79]	1.16	[0.49, 2.75]	1.22	[0.82, 1.81]
Mole-Dagbani	0.99	[0.75, 1.30]	1.06	[0.61, 1.85]	1.00	[0.77, 1.30]
Grusi	0.92	[0.51, 1.65]	1.68	[0.39, 7.20]	1.00	[0.57, 1.77]
Mande	1.23	[0.54, 2.83]	1.86	[0.24, 14.20]	1.34	[0.60, 3.00]
Akan (ref.)
Year of study
Third BA	0.61	[0.50, 0.74]	0.92	[0.69, 1.21]	0.64***	[0.53, 0.77]
Fourth BA (ref.)
University
University of Ghana—Accra	1.19	[0.93, 1.52]	0.79	[0.51, 1.23]	1.10	[0.87, 1.39]
Kwame Nkrumah University of Science and Technology—Kumasi	1.12	[0.85, 1.49]	0.46***	[0.29, 0.71]	0.91	[0.70, 1.19]
University for Development Studies—Tamale (ref.)
Looking forward to National Service Scheme	1.00	[0.97, 1.03]	1.02	[0.97, 1.08]	1.00	[0.98, 1.03]
Variance class level	0.220***		0.000		0.169***
Intra-Class Correlation	0.063		0.000		0.049
N	4,089		2,709		4,089

Note. BA = bachelor; CI = confidence interval; GHS = Ghanaian Cedi..

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

Regarding the starting rate (see Model 1), three conditions—prepaid 10 GHS, prepaid 20 GHS, and promised 20 GHS—significantly outperform the reference condition (promised 5 GHS). Although the prepaid 5 GHS condition yields a slightly higher starting rate than the promised 5 GHS condition (OR = 1.23), it cannot be ruled out that this difference is due to chance. Controlling for individual background of the respondents, a promised incentive of 20 GHS seems to have the strongest effect on survey participation. However, pairwise comparisons show that there are no significant differences in starting rate between prepaid 10 GHS, prepaid 20 GHS, and promised 20 GHS. Furthermore, females, younger students, and fourth-year students (for whom the survey topic on the national service is more salient) are more likely to start the survey.

Promised incentives with higher values (10 and 20 GHS) lead to significantly higher completion rates than the promised 5 GHS group (see Model 2). None of prepaid incentive conditions, however, deviate significantly from the baseline. Apparently, the prospect of receiving a sizeable top-up motivates Ghanaian students to complete the survey, whereas just having received a top-up produces no such effect. Regarding effects of background variables, completion rates are lower among females and students at the University of Kumasi (for which we have no explanation). Contrary to the starting rate, the completion rate shows no significant variation at the class-group level. This suggests that the decision to start the survey has a class dynamic: Probably, the decision to start with the survey is made public in the group of peers via communication between classmates, which can reinforce the norm of reciprocity. The decision to (dis)continue the survey, conversely, is done privately and boils down to a purely individual decision.

These differentiated patterns of starting and retaining are reflected in the impact of incentives on overall response rates (Model 3). As a result of the high starting rate combined with a high completion rate, the promised 20 GHS condition has the highest response rate (OR = 2.37). The response rate does not only have a higher proportion completers than the baseline condition of 5 GHS promised; pairwise comparisons indicate that the response rate of the promised 20 GHS group is higher than in any other incentive condition. Furthermore, prepaid incentives of 10 and 20 GHS turn out to increase overall response rates significantly beyond the level of the promised 5 GHS condition.

Effects of Incentives on Response Quality

Incentives can effectively motivate Ghanaian students to start taking a web survey and to prevent them from dropping out. However, might incentives have unfavorable consequences for the quality of responses provided? In this section, we systematically compare two indicators of response quality—item nonresponse and straight lining—across incentive conditions (see Table 1).

Our results indicate that there is little reason to believe that incentives produce harmful effects on data quality. The percentage of items left blank across the survey varies less than two percentage points across the incentive conditions. Furthermore, this variation in item nonresponse shows no clear pattern across incentive condition. The lowest proportion of item nonresponse is found for the promised 10 GHS condition, while the highest level is found for the promised 20 GHS, which we would have expected to trigger very similar effects. Furthermore, for straightlining between-condition differences are rather small and not clearly patterned. The lowest prevalence is found among respondents in the prepaid 20 GHS condition, who show straightlining behavior on 18.87% of the scales. The highest percentage of straightlining is found among respondents in the promised 10 GHS group with 21.11% of scales.

In order to take the clustering of the data (students within classes) and the compositional differences of classes into account, we again analyzed the proportions of item nonresponse and straight lining per respondent with multilevel models. Multilevel regressions confirm that there is no relationship between the incentive conditions and the indicators of response quality investigated. Neither item nonresponse nor straightlining shows any significant differences across the incentive groups (for detailed results, see Appendix Table A2).

Conclusion and Discussion

Numerous studies into the impact of incentives on survey participation have convincingly shown that (1) incentives increase response rates, (2) unconditional (prepaid) incentives are more effective than conditional (promised) incentives, (3) monetary incentives are more effective than vouchers and in-kind incentives, and (4) the higher value of an incentive, the stronger is the effect it produces, albeit marginal returns diminish (Singer & Ye, 2013). While unconditional prepaid incentives can be difficult to implement in web surveys, overall these findings apply to web surveys as well as other survey modes. In the past decade, scholars have expanded their scope of interest beyond the impact of incentives on survey participation and have started investigating the effect of incentives on indicators of response quality as well, notably on item nonresponse (see, e.g., Curtin et al., 2007; Davern et al., 2003; Medway, 2012). One major shortcoming of this prolific field of study, however, is the fact that all of these studies and empirical analyses have exclusively focused on Western settings. A crucial question, therefore, is: To what extent can we generalize and translate insights concerning incentive effects derived from these Western settings to non-Western societies, which are usually characterized by very different economic conditions, cultural traditions, and survey climates? The current article addresses this important question by analyzing the results of a web survey incentives experiment conducted among 4,440 Ghanaian university students who were interviewed about their upcoming participation in their country’s NSS.

Several important findings and conclusions emerge from our empirical analyses of the incentives experiment. First, the higher the value of an incentive, the more powerful it is to motivate respondents to start the web survey (confirming Hypothesis 2a), although we also find indications for diminishing marginal returns to increasing incentive values (confirming Hypothesis 2b). Second, in the conditions with incentives of lower values, unconditional incentives tend to outperform conditional incentives in encouraging respondents to start the web survey (thus partially supporting Hypothesis 1). These findings largely replicate previous research performed in Western context (for similar conclusions regarding a postal survey in Russia, see Avdeyeva & Matland, 2013). One particular finding, however, contradicts prevailing Western knowledge. Unexpectedly, we found that the providing high (20 GHS) conditional incentives yielded the highest response rate (which is in line with Hypothesis 5). The expectation of a high incentive that will be offered upon completion of the survey not only motivates Ghanaian students to start but also prevents them from breaking off the survey. This deviating finding can probably be explained by the difficult economic context of Ghana and the constant need for many students to find financial resources to cover living expenses and college fees. The fact that a personal relationship between the research team and the students was already established prior to the web survey (during the in-class background survey) and word-of-mouth recommendations by fellow students might have additionally created the trust level necessary for this conditional incentive to be so effective. Furthermore, we surprisingly detect that the promised 20 GHS condition does not outperform the baseline and is nominally worse than the promised 10 GHS incentive. This suggests that unconditional incentives can become so large that they become perceived as purely economic transactions, making it easier to refuse participation (see also James & Bolstein, 1992). Future research is needed to find out whether this is a reproducible finding or rather a chance fluctuation.

Regarding the quality of the data, providing differential incentives did not produce any significant effects (thus rejecting Hypothesis 3 and Hypothesis 4). Incentive value and conditionality were found to be unrelated to both the amount of item nonresponse and straight-lining response behavior.

Given the considerable differences between Western and non-Western settings in terms of the nature and state of the prevailing economic and livelihood conditions, the existing cultural traditions, customs and norms, and the general survey climate, it is revealing to observe that incentive effects concerning web survey participation appear to be largely similar across both settings. Because giving incentives to respondents is an important and relatively easy lever for survey designers, this finding is encouraging for survey practitioners in non-Western countries.

However, it should be clear that our study design has a number of limitations and that considerably more research is needed to confirm the effectiveness of incentives in non-Western contexts. Most notably, the results presented here are based on a survey among university students. Hence, we do not claim generalizability of the effectiveness of incentives in web surveys to the whole Ghanaian population. In fact, such a generalization would be rather meaningless given that large sections of the Ghanaian society have no access to the Internet. Yet, it remains to be seen to what extent our findings can be replicated among other groups in Ghana who are less educated and well-off but nevertheless have access to the Internet on a regular basis. In line with rational choice theory, one could expect that among poorer individuals, unconditional incentives are even more effective than our results suggest. Future research should also attempt to replicate these findings in other non-Western societies. The findings presented here cannot settle the issue on differences between Western and non-Western societies definitively. A second limitation is that we decided to randomize class groups rather than individuals across the conditions in order to preempt class dynamics and contacts between classmates. Although our models control for differences in demographic composition, this decision has decreased statistical power substantially. Finally, although mobile top-ups are almost equivalent to cash in Ghana, further investigation is needed to check whether the provision of real cash incentives would produce different results.

Footnotes

Appendix

Table A2.

Multilevel Regression Models for Indicators of Response Quality.

	Model 4: Item Nonresponse	Model 5: Straightlining
	Par. Est.	Par. Est.
Intercept	−5.91***	19.04***
Incentive
Prepaid 5 GHS	−0.12	−0.94
Prepaid 10 GHS	−0.57	−0.79
Prepaid 20 GHS	−0.19	−1.10
Promised 5 GHS (ref. cat.)
Promised 10 GHS	−0.72	0.32
Promised 20 GHS	0.76	0.49
Gender
Male	−1.29***	−0.52
Female (ref.)
Age	0.51***	−0.02
Ethnicity
Ga-Dangme	0.05	0.42
Ewe	−0.36	0.32
Guan	−2.25*	−0.52
Gurma	−2.28*	0.56
Mole-Dagbani	−1.7**	−0.46
Grusi	−2.15	−2.05
Mande	0.03	0.26
Akan (ref.)
Year of study
Third BA	1.55***	0.35
Fourth BA (ref.)
University
University of Ghana–Accra	0.29	1.53
Kwame Nkrumah University of Science and Technology—Kumasi	0.04	1.30
University for Development Studies—Tamale (ref.)
Looking forward to National Service Scheme	−0.18**	0.15
Variance class level	0.00	1.54
Residual var.	59.33***	228.54***
ICC	.000	.007
N	2,446	2,446

Note. GHS = Ghanaian Cedi; BA = bachelor of arts.

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

Declaration of Conflicting Interests

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

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Research Foundation Flanders (FWO) project “Making Citizens ‘National’: Analyzing the Impact of Ghana’s National Service Scheme (NSS)” (Grant reference G049513N) and the KU Leuven Special Research Fund. The first and third authors gratefully acknowledge support from the Collaborative Research Centre “Political Economy of Reforms” (SFB 884) at the University of Mannheim, which is funded by the German Research Foundation (DFG).

Note

References

Aker

J. C.

Mbiti

I. M.

(2010). Mobile phones and economic development in Africa. The Journal of Economic Perspectives, 24, 207–232.

Alwin

D. F.

(1991). Research on survey quality. Sociological Methods & Research, 20, 3–29. doi:10.1177/0049124191020001001

Armstrong

J. S.

Lusk

E. J.

(1987). Return postage in mail surveys: A meta-analysis. Public Opinion Quarterly, 39, 111–116. doi:10.1086/269031

Auspurg

Schneck

(2014). What difference makes a difference?: A meta-regression approach on the effectiveness conditions of incentives in self-administered surveys. MAER-Net 2014 Athens Colloquium, Athens.

Avdeyeva

O. A.

Matland

R. E.

(2013). An experimental test of mail surveys as a tool for social inquiry in Russia. International Journal of Public Opinion Research, 25, 173–194. doi:10.1093/ijpor/eds020

Birnholtz

J. P.

Horn

D. B.

Finholt

T. A.

Bae

S. J.

(2004). The effects of cash, electronic, and paper gift certificates as respondent incentives for a web-based survey of technologically sophisticated respondents. Social Science Computer Review, 22, 355–362. doi:10.1177/0894439304263147

Blom

A. G.

Gathmann

Krieger

(2015). Setting up an online panel representative of the general population: The German internet panel. Field Methods, 27, 391–408. doi:10.1177/1525822X15574494

Blom

A. G.

Herzing

J. M.

Cornesse

Sakshaug

J. W.

Krieger

Bossert

(2016). Does the recruitment of offline households increase the sample representativeness of probability-based online panels? Evidence from the German internet panel. Social Science Computer Review, Published Online First on June 2, 2016. doi:10.1177/0894439316651584

Blom

A. G.

Jäckle

Lynn

(2010). The use of contact data in understanding cross-national differences in unit nonresponse. In Harkness

J. A.

Braun

Edwards

Johnson

T. P.

Lyberg

Mohler

P. Ph.

Pennell

Smith

T. W.

(Eds.), Survey methods in multinational, multiregional, and multicultural contexts (pp. 335–354). Hoboken, NJ: John Wiley. doi:10.1002/9780470609927.ch18

10.

Bosnjak

Tuten

T. L.

(2003). Prepaid and promised incentives in web surveys. An experiment. Social Science Computer Review, 21, 208–217. doi:10.1177/0894439303021002006

11.

Church

A. H.

(1993). Estimating the effect of incentives on mail survey response rates: A meta-analysis. Public Opinion Quarterly, 57, 62–79. doi:10.1086/269355

12.

Couper

M. P.

(2000). Web surveys. A review of issues and approaches. Public Opinion Quarterly, 64, 464–494. doi:10.1086/318641

13.

Couper

M. P.

de Leeuw

E. D.

(2003). Nonresponse in cross-cultural and cross-national surveys. In Harkness

J. A.

van de Vijver

F. J. R.

Mohler

P. Ph.

(Eds.), Cross-cultural survey methods (pp. 157–178). Hoboken, NJ: John Wiley.

14.

Curtin

Singer

Presser

(2007). Incentives in random digit dial telephone surveys: A replication and extension. Journal of Official Statistics, 23, 91–105.

15.

Davern

Rockwood

T. H.

Sherrod

Campbell

(2003). Prepaid monetary incentives and data-quality in face-to-face interviews: Data from the 1996 survey of income and program participation incentive experiment. Public Opinions Quarterly, 67, 139–147. doi:10.1086/346012

16.

de Bruijn

(2009). Mobile phones: The new talking drums of everyday Africa. Oxford, England: African Books Collective.

17.

de Bruijne

Wijnant

(2013). Comparing survey results obtained via mobile devices and computers: An experiment with a mobile web survey on a heterogeneous group of mobile devices versus a computer-assisted web survey. Social Science Computer Review, 31, 482–504. doi:10.1177/0894439313483976

18.

Delhey

Newton

Welzel

(2011). How general is trust in ‘most people’? Solving the radius of trust problem. American Sociological Review, 76, 786–807. doi:10.1177/0003122411420817

19.

Deutskens

De Ruyter

Wetzels

Oosterveld

(2004). Response rate and response quality of internet-based surveys: An experimental study. Marketing Letters, 15, 21–36. doi:10.1023/B: MARK.0000021968.86465.00

20.

Dillman

D. A.

(1978). Mail and telephone surveys. New York, NY: Wiley.

21.

Dillman

D. A.

(2000). Mail and Internet surveys: The tailored design method. New York, NY: John Wiley.

22.

Edwards

Roberts

Clarke

DiGuiseppi

Pratap

Wentz

Kwan

(2002). Increasing response rates to postal questionnaires: Systematic review. British Medical Journal, 324, 1183–1192. doi:10.1136/bmj.324.7347.1183

23.

Fishbein

Icek

(1975). Belief, attitude, intention and behaviour: An introduction to theory and research. Reading, MA: Addison-Wesley.

24.

Gelman

Stevens

Chan

(2003). Regression modeling and meta-analysis for decision making: A cost–benefit analysis of incentives in telephone surveys. Journal of Business & Economic Statistics, 21, 213–225. doi:10.1198/073500103288618909

25.

Göritz

A. S.

(2004). The impact of material incentives on response quantity, response quality, sample compositions, survey outcome, and cost in online access panel. International Journal of Market Research, 46, 327–345.

26.

Göritz

A. S.

(2006a). Cash lotteries as incentives in online panels. Social Science Computer Review, 24, 445–459.

27.

Göritz

A. S.

(2006b). Incentives in web studies: Methodological issues and a review. International Journal of Internet Science, 1, 58–70.

28.

Gouldner

A. W.

(1960). The norm of reciprocity: A preliminary statement. American Sociological Review, 25, 161–178.

29.

Groves

R. M.

Couper

(1998). Non-response in household interview surveys. New York: John Wiley.

30.

Groves

R. M.

Singer

Corning

(2000). Leverage-salience theory of survey participation: Description and illustration. Public Opinion Quarterly, 64, 299–308. doi:10.1086/317990

31.

Hansen

R. A.

(1980). A self-interpretation of the effect of monetary and nonmonetary incentives on mail survey respondent behavior. Journal of Marketing Research, 17, 77–83. doi:10.2307/3151120

32.

Heberlein

T. A.

Baumgartner

(1978). Factors affecting response rates to mailed questionnaires: A quantitative analysis of the published literature. American Sociological Review, 43, 447–462.

33.

Homans

G. C.

(1950). The human group. New York, NY: Harcourt, Brace and Company.

34.

Hopkins

K. D.

Gullickson

A. R.

(1992). Response rates in survey research: A meta-analysis of the effects of monetary gratuities. Journal of Experimental Education, 61, 52–62. doi:10.1080/00220973.1992.9943849

35.

Hox

J. J.

(2010). Multilevel analysis. Techniques and applications. New York, NY: Routledge.

36.

Rosenberger

W. F.

(2014). Adaptive randomization for balancing over covariates. Wiley Interdisciplinary Reviews: Computational Statistics, 6, 288–303. doi:10.1002/wics.1309

37.

Hydén

(2006). African politics in comparative perspective. Cambridge, MA: Cambridge University Press.

38.

Jäckle

Lynn

(2008). Respondent incentives in a multi-mode panel survey: Cumulative effects on nonresponse and bias. Survey Methodology, 34, 105–117.

39.

James

J. M.

Bolstein

(1990). The effect of monetary incentives and follow-up mailings on the response rate and response quality in mail surveys. Public Opinion Quarterly, 54, 346–361. doi:10.1086/269211

40.

James

J. M.

Bolstein

(1992). Large monetary incentives and their effect on mail survey response rates. Public Opinion Quarterly, 56, 442–453. doi:10.1086/269336

41.

Kaminska

McCutcheon

A. L.

Billiet

(2010). Satisficing among reluctant respondents in a cross-national context. Public Opinion Quarterly, 74, 956–984. doi:10.1093/poq/nfq062

42.

Langer

Meuleman

Oshodi

A. G. T.

Schroyens

(2016). Can student populations in developing countries be reached by online surveys? The case of the National Service Scheme Survey (N3S) in Ghana. Field Methods, doi:10.1177/1525822X16663702.

43.

Mauss

(1923). Essai sur le don forme et raison de l’échange dans les sociétés archaïques. L’Année sociologique (1896/1897-1924/1925), 1, 30–186.

44.

Mavletova

(2013). Data quality in PC and mobile web surveys. Social Science Computer Review, 31, 725–743. doi:10.1177/0894439313485201

45.

Mavletova

Couper

M. P.

(2016). Device use in web surveys: The effect of differential incentives. International Journal of Market Research, 58, 523–544.

46.

Medway

(2012). Beyond response rates: The effect of prepaid incentives on measurement error (PhD thesis). Graduate School of the University of Maryland, College Park.

47.

Mutti

Kennedy

R. D.

Thompson

M. E.

Fong

G. T.

(2014). Prepaid monetary incentives—Predictors of taking the money and completing the survey: Results from the International Tobacco Control (ITC) Four-Country Survey. Sociological Methods & Research, 43, 338–355. doi:10.1177/0049124113506406

48.

Petrolia

D. R.

Bhattacharjee

(2009). Revisiting incentive effects: Evidence from a random sample mail survey on consumer preferences for fuel ethanol. Public Opinion Quarterly, 73, 537–550. doi:10.1093/poq/nfp038

49.

Pforr

Blohm

Blom

A. G.

Erdel

Felderer

Frässdorf

… Rammstedt

(2015). Are incentive effects on response rates and nonresponse bias in large-scale, face-to-face surveys generalizable to Germany? Public Opinion Quarterly, 79, 740–768. doi:10.1093/poq/nfv014

50.

Philipson

(1997). Data markets and the production of surveys. The Review of Economic Studies, 64, 47–72. doi:10.2307/2971740

51.

Ruiter

Van Tubergen

(2009). Religious attendance in cross–national perspective: A multilevel analysis of 60 countries. American Journal of Sociology, 115, 863–895. doi:10.1086/603536

52.

Scherpenzeel

Toepoel

(2012). Recruiting a probability sample for an online panel effects of contact mode, incentives, and information. Public Opinion Quarterly, 76, 470–490. doi:10.1093/poq/nfs037

53.

Shettle

Mooney

(1999). Monetary incentives in U.S. government surveys. Journal of Official Statistics, 15, 231–250.

54.

Shuttleworth

F. K.

(1931). A study of questionnaire technique. Journal of Educational Psychology, 22, 652–658. doi:10.1037/h0074591

55.

Simmons

Wilmot

(2004). Incentive payments on social surveys: A literature review. Social Survey Methodology Bulletin, 53, 1–11.

56.

Singer

(2011). Toward a benefit-cost theory of survey participation: Evidence, further tests, and implications. Journal of Official Statistics, 27, 379.

57.

Singer

Van Hoewyk

Gebler

Raghunathan

McConagle

(1999). The effect of incentives on response rates in interviewer-mediated surveys. Journal of Official Statistics, 15, 217–230.

58.

Singer

Van Hoewyk

Maher

M. P.

(2000). Experiments with incentives in telephone surveys. Public Opinion Quarterly, 64, 171–188. doi:10.1086/317761

59.

Singer

(2013). The use and effects of incentives in surveys. ANNALS of the American Academy of Political and Social Science, 645, 112–141. doi:10.1177/0002716212458082

60.

Teisl

M. F.

Roe

Vayda

(2005). Incentive effects on response rates, data quality, and survey administration costs. International Journal of Public Opinion Research, 18, 364–373. doi:10.1093/ijpor/edh106

61.

van Veen

Göritz

A. S.

Sattler

(2016). Response effects of prenotification, prepaid cash, prepaid vouchers, and postpaid vouchers an experimental comparison. Social Science Computer Review, 34, 333–346.

62.

Wetzels

Schmeets

van den Brakel

Feskens

(2008). Impact of prepaid incentives in face-to-face surveys: A large-scale experiment with postage stamps. International Journal of Public Opinion Research, 20, 507–516. doi:10.1093/ijpor/edn050

63.

Cooper

(1983). A quantitative review of research design effects on response rates to questionnaires. Journal of Marketing Research, 20, 36–44. doi:10.2307/3151410