1.1 The problem of household food waste
  Globally, 1.3 billion tons per year of all food that is produced for human consumption are wasted (see e.g. Buchner et al., 2012; Gustavsson, Cederberg, Sonesson, Van Otterdijk, & Meybeck, 2011; HLPE, 2014; Lipinski et al., 2013; Monier et al., 2011; Priefer, Jörissen, & Bräutigam, 2016; Thyberg & Tonjes, 2016 for an overview). Except for economic and social issues, these high amounts of unconsumed food also greatly contribute to global environmental problems, e.g. climate change (for example, by causing 3,3Gt unnecessary greenhouse gas emissions in 2007), water scarcity as well as increasing deforestation and species extinction (Food and Agriculture Organization of the United Nations (FAO), 2013).

  Furthermore, it should be noted, that especially industrialized countries are characterized by high amounts of food waste, i.e. food that is discardedbeing suitable for human consumption (see e.g. Buchner et al., 2012; FAO, 2013; Gustavsson et al., 2011; Lipinski et al., 2013; Lucifero, 2016; Parfitt, Barthel, & Macnaughton, 2010; Priefer et al., 2016; Stuart, 2009; Thyberg & Tonjes, 2016). In consequence of inadvertence or conscious decisions for discarding, food waste – especially household food waste– is closely related to individual behavior of private consumers (FAO, 2013; Gustavsson et al., 2011; Lipinski et al., 2013; Parfitt et al., 2010; Priefer et al., 2016; Stuart, 2009). Thereby, studies conducted in several (industrialized) countries (e.g. Germany, Great Britain, USA etc.) unanimously indicated private households as main contributors of food waste (e.g. Buchner et al., 2012; Lipinski et al., 2013; Lucifero, 2016; Monier et al., 2011; Priefer et al., 2016; WWF Schweiz [Switzerland], 2012). Focusing on Germany, households are assumed to cause 61% of all food wastage, resulting in nearly 6.7 million tons per year and an average amount of 82 kg per person. Out of these 82 kg, nearly 47% (i.e. 38 kg) represent food that was completely edible to that time when it was discarded or it would have been edible by consumption in time (Kranert et al., 2012). Therefore, appropriate intervention approaches are needed in order to prevent these high amounts of household food waste – in Germany and beyond.

1.2. Appropriate intervention approaches to promote household food waste-prevention
1.2.1. Predictors of household food waste-prevention: Identifying appropriate intervention approaches to promote household food waste-prevention generally depends on considering relevant influencing factors that are, somehow, causing or related to household food waste (i.e. predictors of household food waste). Consequently, these predictors will also determine household food waste-prevention. Thereby, household food waste as well as its prevention is influenced by a variety of different perceptual (e.g. perceived environmental consequences of household food waste or the perceived own ability to prevent it; see e.g. Farr-Wharton, Foth, & Choi, 2014; Quested, Marsh, Stunell, & Parry, 2013 for details), motivational (e.g. social/ moral norms to prevent household food waste; Graham-Rowe, Jessop, & Sparks, 2014; Graham-Rowe, Jessop, & Sparks, 2015; Parizeau, von Massow, & Martin, 2015; Quested et al., 2013; Visschers, Wickli, & Siegrist, 2016; Watson & Meah, 2013) and behavioral predictors (see e.g. Quested et al., 2013; Schmidt, 2016 for further overviews). Focusing on behavioral predictors, an important aspect has to be mentioned: Household food waste as well as its prevention cannot be seen as (the result of) a single behavior, but as a combination of multiple behaviors influencing its likelihood (Mondéjar-Jiménez, Ferrari, Secondi, & Principato, in press; Neff, Spiker, & Truant, 2015; Priefer et al., 2016; Quested et al., 2013; Thyberg & Tonjes, 2016). Thereby, these food waste-related, or as termed by Schmidt (2016) food waste-preventing behaviors, represent an extensive class of different behaviors reaching from preparation of grocery shopping to grocery shopping itself, storing and preparation of food at home as well as to its final discarding.

  Understanding household food waste-prevention as the result of performing several relevant food waste-prevention behaviors has important practical implications already considered by existing food waste-prevention initiatives (e.g. “Zu gut für die Tonne” [https://www.zugutfuerdietonne.de/] in Germany and, Love Food Hate Waste“ in Great Britain [http://www.lovefoodhatewaste.com/]). Thereby, these initiatives are considering the meaning of performing relevant food waste-preventing behaviors sufficiently for promoting household food waste-prevention by providing appropriate action knowledge (e.g. Abrahamse & Matthies, 2012; Homburg & Matthies, 1998; Mack, 2007), i.e. recommendations on how to prevent household food waste by improving the performance of relevant food waste-preventing behaviors, as an essential part of their information campaigns. From a psychological perspective, providing appropriate action knowledge built on the assumption, that knowledge and skills related to the desired behavior (i.e. household food waste-prevention) will reduce its perceived difficulty and increases the perceived ability to do so (Klöckner, 2015). Typically, existing food waste-prevention initiatives are presenting comprehensive collections of many relevant food waste-preventing behaviors on their websites, in information brochures etc. Assuming that most people already perform at least some relevant food waste-preventing behaviors sufficiently, most recipients will only perceive some of all provided recommendations (i.e. those referring to relevant food waste-preventing behaviors a person is not performing (sufficiently) as relevant, while others are perceived as redundant. But without considering these individual perception processes, desired effects of providing action knowledge (i.e. people are following the presented recommendations and will improve their performance of relevant food waste-preventing behaviors) becomes (more) improbable. To explain this assumption in more detail, we have to consider some insights referring to human information processing.

1.2.2. The relevance of human information processing depth for sustainable interventioneffects: Considering established dual process models of human information processing like the elaboration likelihood model (ELM) of persuasion (Petty & Cacioppo, 1986; Petty & Wegener, 1999) or the heuristic-systematic model of information processing (HSM; Chaiken, 1987; Chaiken, Liberman, & Eagly, 1989) depth of human information processing depends on the cognitive ability as well as on the individual motivation for deep information processing. Except for other influencing factors (see e.g. Eagly & Chaiken, 1993) for an overview), the individual motivation highly depends on the perceived relevance of the information a person should process (e.g. a list of recommendations on how to prevent household food waste by improving relevant food waste-preventing behaviors). This is because humans can be seen as economy-minded souls “(Eagly & Chaiken, 1993; p. 330) that are interested in processing information as most efficiently, or in other words, with at leastefforts as possible (least effort principle; Eagly & Chaiken, 1993). Therefore, recipients of lists containing a lot of information – perceived as only partly relevant – could be less motivated to process all provided information deeply. But without such deep information processing, desired changes of recipient’s attitudes or even desired changes in behavior will become not only less probable, but also less durable (Chaiken, 1980, 1987; Eagly & Chaiken, 1993; Petty & Cacioppo, 1986; Petty & Wegener, 1998, 1999).

  Therefore, promoting household food waste-prevention by providing appropriate action knowledge should be based on high probabilities for deep information processing. Thus, increasing the perceived relevance of such provided action knowledge, for example by strictly limiting to recommendations referring to relevant food waste-preventing behaviors a person (still) not perform (sufficiently), could be a useful way for existing as well as for future food waste-prevention initiatives to (further) promote household food waste-prevention sustainably(1).

  Therefore, the central aim of the current research study was to propose an efficient (online) tool/ measuring instrument that can be used by food waste-preventing initiatives for increasing the perceived relevance of their provided action knowledge in order to sustainably promote (further) household food waste-prevention. In this context, efficient“means the developed tool/ measuring instrument can be easily implemented by initiative’s information campaigns” (e.g. by adding the tool on websites). Additionally, reducing efforts for using the developed tool/ measuring instrument was not only applied to initiative’s perspective, but also to their recipient’s perspective: Thus, a tool/ measuring instrument should be developed, that allows increasing the perceived relevance of provided action knowledge without needing high recipient’s efforts (e.g. by asking an excessive amount of detailed questions regarding their current performance of relevant food waste-preventing behaviors).

  To fulfil this central aim, two preliminary studies were conducted in order to develop and evaluate an initial version of the desired tool/ measuring instrument (i.e. the food waste-preventing behaviors scale). Due to some limitations and opportunities for further optimization of this initial version regarding relevant scale characteristics, a third study was conducted in order to extend aswell as to consolidate the initial scale-version and to finally propose a developed version of thefood waste-preventing behaviors scale.

2. Preliminary studies for developing an initial version of the food waste-preventing behaviors scale
2.1. Materials and methods

2.1.1. Data collection and sample recruitment: Initial development and examination of the food waste-preventing behaviors scale was realized by conducting two consecutive data collections using an online-survey implemented in LimeSurvey-software(2). The first assessment (t1) was conducted in September/ October 2014, the second one (t2) in December 2014/ January 2015. For both assessments, participants were recruited in Germany, mainly around the town of Magdeburg: Thereby, participants were recruited through social networks (i.e. Facebook), circular mails/ newsletters (around the university and the technical college of Magdeburg and some universities beyond) and through posters and flyers (placed or distributed at the university of Magdeburg). Furthermore, in order to receive a sample as heterogeneous as possible to avoid examining only a special household type (e.g. students’ households), written invitations were also distributed through articles published in local newspapers and several people were also invited personally by presenting the intervention study on a public event concerning food waste issues in Germany, organized in the town of Magdeburg.

  Initially, 336 people took part in the t1-assement and 279 people took part in the t2-assessment. After eliminating persons, who did not answer the whole online-survey, 217 people formed the final t1-sample and 206 people formed the final t2-sample. Unfortunately, regarding several socio-demographic characteristics (e.g. distributions of gender, age, education, income and household size) both samples were not representative for the German population (German federal Bureau of Statistics, 2014, 2015a; 2015b; see Table 1 for details).

Table 1: Socio-demographic characteristics of t1- (N_t1 = 217) and t2- (N_t2 = 206) sample compared to German population.

2.1.2. Item selection for the initial version of the food waste-preventing behaviors scale: Even though no standardized scale to record the performance of relevant food waste-preventing behaviors was developed before, there are several collections of such behaviors proposed by other authors (see e.g. Kranert et al., 2012; Priefer et al., 2016; Quested et al., 2013; Thyberg & Tonjes, 2016 for overviews). Thus, these collections were used to compile a baseline composition of nearly all conceivable relevant food waste-preventing behaviors. At the end of the collection process, the baseline composition contained more than 70 different behaviors. Considering the above-mentioned efficiency issue regarding the resulting scale (see section 1.2.2), further item selections were necessary in order to develop an applicable initial scale-version. Therefore, only “changeable” food waste-preventing behaviors were further selected for scale development: Thus, behaviors for those household members are not responsible (e.g. an unintended purchase of already spoiled food) were excluded. Additionally, it was tried to cover the whole class of relevant food waste-preventing behaviors (as already mentioned above, starting with preparation of grocery shopping, grocery shopping itself, preparation of food at home, storing of food at home to final discarding of food) with at least items as possible. At the end of this selection process, the initial version of the food waste-preventing behaviors scale contained 14 items (see Table 6 for an overview). All items were recorded by using a five-point Likert scale (ranging from 1 =” (almost) never” to 5 =” (almost) always”), introduced by questions, like: “Before grocery shopping, do you think carefully about the food you currently really need? This is what I do ...”

2.1.3. Data analyses procedure: As the central aim of the preliminary studies was to construct and examine an initial version of the food waste-preventing behaviors scale by understanding the underlying factor structure of the collected data as well as to reduce data sets to a more manageable size by identifying unnecessary items, an exploratory factor analysis was conducted for both data sets (Field, 2012). Therefore, first analyses were done in order to check for sufficient data quality regarding important conditions for conducting these analyses (i.e. necessary sample sizes, Kaiser-Meyer-Olkin measure of sampling adequacy (KMO) and bartlett’s test of sphericity). Afterwards, the main analyses wereconducted for both samples – using principle axis analysis as extraction method and calculated eigenvalues (Kaiser’s criterion) as well as some theoretical preliminary considerations as relevant extraction criteria (see e.g. Bühner, 2006; Field, 2012 for details). Finally, extracted factors were rotated by using an oblique rotation technique (Direct Oblimin; Delta = 0), because relations between the extracted factors were expected (due to the overall contribution of all relevant food waste-preventing behaviors to household food waste-prevention).

2.2. Results of exploratory factor analyses

2.2.1. Examining conditions for conducting exploratory factor analyses: Following Bühner (2006), for conducting an exploratory factor analysis, data of at least 200 people should be given. Thus with N_t1 = 217 and N_t2 = 206 both samples fulfilled this requirement s well as sufficient KMO-coefficients (.74 for t1-data and .77 for t2-data) and highly significant results(p < .01) of bartlett’s tests of sphericity. Thus, both data sets were adequate for conducting exploratory factor analyses.

2.2.2. Factor extraction, loadings and intercorrelations: Considering calculated eigenvalues within both data sets (Kaiser’s criterion; see Table 2 for details) simultaneously did not lead to completely clear results: Considering t1-eigenvalues, the analyses proposed a five factors-extraction (EV1 = 3.97, EV2 = 1.95, EV3 = 1.70, EV4 = 1.13, EV5 = 1.05), while only a four factors-solution was proposed for t2-data (EV1 = 4.32, EV2 = 1.90, EV3 = 1.60, EV4 = 1.12, EV5 = .89). Therefore, the final extraction decision was made in order to receive a solution as stable and as reproducible as possible. Thus, the four factor-solution was preferred for both data sets. Based on the extraction of four factors, nearly 67% of variance could be explained in t1-data and 69% of variance in t2-data.

Table 2: Eigenvalues and explained variance of all possible factors, calculated in exploratory factor analysis conducted with t1- and t2-data sets.

  As can be seen in Table 3, factor 1 (Avoiding spontaneous purchases due to different reasons) wasrepresented by four items, characterized by high factor loadings in both data sets (.81, .83, .93 and.81 for t1-data; .83, .82, .94 and .84 for t2-data).

  Factor 2 (Appropriate dealing with leftovers) was represented by two items, also characterized by high loadings (.98 and .96 for both data sets).

  Factor 3 (Appropriate dealing with excessed and expired, but (still) edible food) was also represented by two items, characterized by satisfying or at least acceptable loadings (.84 and .71 for t1-data; -.78 and -.58 for t2-data). Finally, factor 4 (Sufficient preparation of grocery shopping) was also represented by two other items, characterized by comparatively weak, but still acceptable loadings (.56 and .49 for t1-data; .64 and .45 for t2-data).

  As expected, results also revealed relevant, but varying intercorrelations between all extracted factors (see Table 4 for details): The strongest correlations (r = .31, p < .01 for t1-data; r = .35, p < .01 for t2-data) were identified between factor 3 and factor 4. Only weak correlations (r = .09, n.s. for both data sets) were identified between factor 2 and factor 4, while all other correlations varied between r = .14, p < .05 and r = .30, p < .01.

Table 3: Factor loadings of all items, calculated in exploratory factor analyses conducted with t1- and t2-data sets.

NOTE: (*) Before conducting the factor analyses, these items were weighted with the item to record the general frequency of spontaneous purchases; (**) Before conducting the factor analyses, these items were weighted with the item to record the general frequency of preparing too much food for a meal; (***) Before conducting the factor analyses, these items were weighted with the item to record the general frequency of expired food.

Table 4: Factor intercorrelations, calculated in exploratory factor analyses conducted with t1- and t2-data sets.

** p < .01; * p < .05

Table 5: Correlations between the four extracted factors (during the exploratory factor analyses conducted with t1- and t2-data sets) and the remaining items.

** p < .01; * p < .05

2.2.3. Appropriate dealing with remaining items and finalizing the initial version of the food: As already presented in Table 3, there were three remaining items, characterized by unclear results regarding the four extracted factors: As mentioned above, based on its eigenvalue, the extraction of a fifth factor was proposed for t1-data. But this possible factor was nearly completely represented by only one item – an item to record frequency of grocery shopping (initial factor loading: .75 for t1-data). Because neither for t1-, nor for t2-data any other relevant loadings were found, this factor was not extracted. Simultaneously, the underlying item (to record frequency of grocery shopping) showed no relevant loadings on any of the extracted four factors – neither for t1-data, nor for t2-data. Therefore, this remaining item was excluded.

  Another remaining item was initially used to record time spend for grocery shopping. But regarding the four extracted factors, this item showed no relevant loading for t1-data and only a marginal loading on factor 4 (.39) for t2-data. A comparable loading pattern was found for the last remaining item, that was initially used to record the usage of appropriate measures to extend shelf-life of food at home: For t1-data no relevant factor loadings were identified, whereas for t2-data at least a marginal relevant loading on factor 3 (-.34) was found. To further decide how to deal with these two remaining items, middle product-moment-correlations (following Bortz & Döring, 2006) between these items and the four extracted factors were calculated (see Table 5). This procedure was chosen, because such correlations– at least at a minimum – should be given for an appropriate integration of these items into the full scale. Thereby, the item to record time spend on grocery shopping showed significant correlations with factor 3 (r = .24, p < .01) and factor 4 (r = .21, p < .01) in t1-data and nearly the same correlations were found for t2-data (factor 3: r = .17, p < .05 and factor 4: r = .24, p < .01). Instead, for the other item (to record usage of appropriate measures to extend shelf-life of food at home) significant correlations were found with all extracted factors in t1-data (factor 1: r = .18, p < .01; factor 2: r = .24, p < .01; factor 3: r = .32, p < .01; factor 4: r = .30, p < .01). For t2-data, highly significant correlations were found again with factor 3 (r = .22, p < .01) and with factor 4 (r = .21, p < .01).

  Based on these results integrating only the item to record usage of appropriate measures to extend shelf-life of food at home into the full scale was seen as useful, while the item to record time spend on grocery shopping was excluded.

2.3. Examining relevant scale characteristics and discussion
  Following the exploratory factor analyses, relevant scale characteristics were calculated in order to further examine quality and applicability of the initial scale version as well as of its four subscales in both data sets (see Table 6). Therefore, mean valuesand standard deviations were calculated as relevant indications for item distribution. Additionally, reliability estimates were calculated for the full scale as well as for each subscale: Cronbach’s Alpha estimates were calculated for subscales consisting of three or more items (e. g. Schermelleh-Engel & Werner, 2012),while Spearman-Brown correlations were calculated as alternative reliability estimates for subscales consisting of two items (Eisinga, Grotenhuis, & Pelzer, 2013).

  Considering mean values of each subscale (ranging from 2.72 (.97) to 4.26 (.82) for t1-data and ranging from 2.84 (1.02) to 4.35 (.90) for t2-data), results suggest, that some relevant food waste-preventing behaviorswereperformed more frequently (e.g. behaviors related to sufficient preparation of grocery shopping) than others (e.g. behaviors related to appropriate dealing with leftovers). Consequently, some relevant food waste-preventing behaviors could be characterized by a higher number or stronger specific behavioral barriers (see e.g. McKenzie-Mohr & Smith, 2008 for details)than others – as already proposed for example by Graham-Rowe et al. (2014).

  Considering scale reliability in both data sets, mostly high or at least sufficient reliability estimates were calculated for the full scale (α_t1 = .81 and α_t2 = .83) as well as for its subscales with α_t1 = .92 and α_t2 = .93 for factor 1, r_t1 = .94 and r_t2 = .94 for factor 2, r_t1 = .58 and r_t2 = .51 for factor 3 and with r_t1 =.30 and r_t2 = .37 for factor 4 (George & Mallery, 2003; Gliem & Gliem, 2003). Considering the comparatively low reliability estimates for factor 3 and 4, two different explanations seem possible: On the one hand, lower estimates could be the result of an inappropriate integration of two items into one superior factor. On the other hand, lower estimates could also be influenced by the chosen reliability estimation measure (for example, calculating retest reliability could lead to higherreliability values; see e.g. Schermelleh-Engel & Werner, 2012). But only based on available data, no alternative reliability estimation measures could be calculated(3). Thus, by only considering preliminary data, no clear indications can be delivered to further examine both explanations. Strong associated with this point, another issue regarding the calculated reliability estimates should be mentioned: For the remaining item (to record usage of appropriate measures to extend shelf-life of food at home), that was finally integrated into the full scale, no reliability estimate could be calculated (because retest reliability estimations were not feasible, as already mentioned above). Thus, neither good nor bad statements can be made regarding the reliability of this item.

Table 6: Relevant scale characteristics, calculated for the initial scale-version and its subscales with t1- and t2-data.

NOTE: All item values were matched in a way that higher values represent a higher likelihood of household food waste-prevention (recoded items were marked with (*)). Generally, Cronbach’s Alpha estimates were used as reliability measure for scales consisting of three or more items. For scales consisting of two items the Spearman-Brown correlation of both items was calculated as an alternative reliability measure (Eisinga et al., 2013).

  Taken together, both preliminary studies contributed successfully to the development of an initial version of the food waste-preventing behaviors scale. Simultaneously, the presented results also indicate the need of further development processes, especially regarding additional factor separations/ options to further extend as well as to consolidate the initial factorial structure of the scale. Therefore, an additional study was conducted in order to contribute to these further development processes by examining a developed version of the food waste-preventing behavior scale.

3. Examining a developed version of the food waste-preventing behavior scale
3.1. Materials and methods

3.1.1. Data collection and sample recruitment: The third assessment (t3) to further extend as well as to consolidate the initial version of the food waste-preventing behaviors scale was conducted from spring to summer 2015. Therefore, a third online-survey was implemented in LimeSurvey-software again. Participants were recruited in Germany, primary by using social networks (i.e. Facebook) and circular mails/ newsletters. Altogether, 348 people took part in the assessment. Out of those, 312 people completed the whole survey and formed the final – unfortunately, as well as the previous t1- and t2-samples, not representative for German population (see Table 7 for details) – t3-sample.

Table 7: Socio-demographic characteristics of participants in the t3-assessment (N_t3 = 312) compared to German population.

3.1.2. Further item selection for the developed food waste-preventing behaviors scale: In order to select additional items to further extend as well as to consolidate the initial factorial scale structure appropriately, the initially composed baseline composition of more than 70 conceivable relevant food waste-preventing behaviors (see section 2.1.2) was considered again. Thereby, the above-mentioned critical points regarding initial factorial structure of the initial scale version (especially referring to initial factors 3 and 4 as well as referring to the remaining single item; see section 2.3) were considered. Keeping these points in mind, finally, seven additional items wereassessed as appropriate additions, primary representing additional food waste-preventing behaviors related to sufficient preparation of grocery shopping, appropriate dealing with expired and excessed, but (still) edible food as well as behaviors related to storing of food at home (see Table 11 for details). Thus, this extended scale-version consisted of 21 items. Again, all items were recorded by using a five-point Likert scale (ranging from 1 =” (almost) never” to 5 =” (almost) always”), introduced by questions, like: “Before grocery shopping, do you think carefully about the food you currently really need? This is what I do ...”

3.1.3. Data analysis procedure: Although, the preliminary studies identified some initial factor structure, also the third study’s aim was to (further) understand (an extended) factor structure of the collected data. Therefore, an exploratory factor analysis was conducted for t3-data again (Field, 2012)– starting againwith examining for sufficient data quality regarding important conditions for conducting these kind of analyses (necessary sample size, Kaiser-Meyer-Olkin measure of sampling adequacy (KMO) and bartlett’s test of sphericity). Afterwards, the main analysis was conducted nearly the same way as it was conducted for t1- and t2-data by using principle axis analysis as extraction method and an oblique rotation technique (Direct Oblimin; Delta = 0; see again Bühner, 2006; Field, 2012 for details). Thereby, the previous results and the mentioned assumptions regarding an extended factor structure of a developed scale were used as additional extraction criteria. Therefore, the extraction of seven factors (again, the initial factors 1 and 2, whereas the initial factors 3 and 4 were assumed to be separated into two different factors respectively and one new factor representing storing of food) was predefined at the beginning of the analysis, while eigenvalues (Kaiser’s criterion) were examined in order to identify relevant deviations of these assumptions.

3.2. Results of the final exploratory factor analysis

3.2.1. Examining conditions for conducting an exploratory factor analysis: Again in line with Bühner (2006), with N_t3 = 312 t3-data completely fulfilled the sample-size-requirement, as well as data revealed a sufficient KMO-coefficient (.78) and a highly significant result (p < .01) of bartlett’s test of sphericity. Thus, a t3-data set was assessed as also adequate for conducting an exploratory factor analysis.

3.2.2. Factor extraction, loadings and intercorrelations: As can be seen in Table 8, the predefined extraction of seven factors was nearly in line with the number of factors, which would have been extracted based on calculated eigenvalues – at least one factor barely missed the 1.00-barrier. So, this criterion did not lead to strong deviations regarding the predefined number of to extracting factors. Therefore, the seven-factor solution, explaining more than 75% of all variance, was maintained.

Table 8: Eigenvalues and explained variance of all possible factors, calculated in exploratory factor analysis conducted with t3-data set.

  As can be seen in Table 9, factor 1 (Avoiding spontaneous purchases due to different reasons) was represented again by the same four items, characterized by high loadings (.82, .84, .96 and .77), as in the preliminary studies as well as factor 2 (Appropriate dealing with leftovers), which was also represented by the same two items as before, also characterized by high loadings (.94 and .89). As already expected, the initial factor representing appropriate dealing with excessed and expired, but (still) edible food, was now separated into two factors: Thereby, (new) factor 3 (Appropriate dealing with excessed, but (still) edible food) was represented by two items, characterized by high loadings (.70 and .72) as well as factor 4 (Appropriate dealing with expired, but (still) edible food), which was also represented by two items, characterized by really high loadings (.94 and .90). Also in line with the previous expectations, factor 5 (Planning of grocery shopping in advance) was represented by three items, characterized by high or at least sufficient loadings (.49, .87 and .70) and factor 6 (Sufficient overview on food stocks at home) was represented by two items, characterized by satisfactory loadings (.63 and.56). Finally, factor 7 (Usage of adequate measures to extend shelf-life of food at home) was represented by three remaining items. Consistent with the weakest eigenvalue of this factor, all items were characterized by at least acceptable, but comparatively weak loadings (.49, .35 and .47).

Table 9: Factor loadings of all items calculated in exploratory factor analysis conducted with t3-data set.

NOTE: (*) Before conducting the factor analysis, these items were weighted with the item to record the general frequency of spontaneous purchases; (**) Before conducting the factor analysis, these items were weighted with the item to record the general frequency of preparing too much food for a meal; (***) Before conducting the factor analysis, these items were weighted with the item to record the general frequency of expired food.

  As expected, results of this final factor analysis also revealed relevant, but again varying intercorrelations between all seven extracted factors – except for factor 2, which was characterized by only one significant, but weak correlation with factor 1 (r = .13, p < .05; see Table 10 for details). Now, the strongest correlations were identified between factor 7 and factor 4 (r = .39, p < .01 and r =.35, p < .01) as well as between factor 3 and factor 4 (r = .37, p < .01 and r = .35, p < .01). Instead, the weakest correlations were identified between factor 5 and factor 6 (r = .13, p < .05) as well as between factor 7 and factor 1 (r = .14, p < .01).

Table 10: Factor intercorrelations calculated in exploratory factor analysis conducted with t3-data set.

** p < .01; * p < .05

3.3. Examining relevant scale characteristics and discussion: As already done for the initial scale-version, relevant characteristics were calculated again for the developed scale-version in order to further examine the quality and applicability of the full scale as well as of its seven subscales (see Table 11). Therefore, the same measures (mean values, standard deviations and reliability estimates) were calculated for t3-data.

  In line with previous results, considering mean values of each subscale (ranging from 2.77 (.90) to 4.37 (.65)) still indicates general differences in performance frequency of the examined food waste-preventing behaviors – further supporting made assumptions regarding different behavior difficulty because of different amounts of specific behavioral barriers.

  Considering calculated reliability estimates, data revealed really high or at least good reliability estimates for the full scale as well as for nearly all subscales (George & Mallery, 2003; Gliem & Gliem, 2003) of the developed food waste-preventing behaviors scale. Especially regarding newly built subscales for planning of shopping in advance, sufficient overview on food stocks at home, appropriate dealing with expired but (still) edible food and for appropriate dealing with excessed, but (still) edible food, reliability estimates were satisfying. Only one subscale was characterized by an at least acceptable reliability estimate – usage of measures to extend food’s shelf-life at home. Thus, the overall aim of this final study (i.e. further extending and consolidating the initial factor structure of the scale) was generally reached.

Table 11: Relevant scale characteristics calculated for the developed scale-version and its subscales with t3-data.

4. General Discussion

4.1. Summary
  The central aim of the current paper was to propose an efficient (online) tool/ measuring instrument for recording self-reported performances of several relevant food waste-preventing behaviors. Starting with developing and examining an initial scale-version consisting of 14 items, characterized by a four factor solution, finally, a more comprehensive developed scale-version – consisting of 21 items – could be created. Considering relevant characteristics, the developed food waste-preventing behaviors scale seems to be such an efficient and satisfying (online) tool/ measuring instrument that was aimed. Thus, this scale could actually be used by existing and future food waste-preventing initiatives for increasing the perceived relevance of their provided action knowledge in order to sustainably promote (further) household food waste-prevention. But the application field of the proposed scale should not be limited only on a practical area. Instead, the scale could also be used as an appropriate research tool, for example, when conducting science-based intervention studies as it was doneby Schmidt (2016).

4.2. Limitations and implications for future research
   Although, the presented results indicated reaching relevant research goals, there are some limitations that should also be mentioned. One limitation refers to the exclusive use of self-reported measures to record the performance of relevant food waste-preventing behaviors: As a general limitation associated with self-report measures, possible response biases – especially referring to social desirability – having negative impacts on the data quality (see e.g. Krumpal, 2013; Tourangeau & Yan, 2007 for details), cannot be excluded.

  Furthermore, generalizing current results is restricted referring to their transferability on other samples and populations. As shown above, all studies were conducted with convenience and not representative (for German population) samples, mainly consisting of well-educated, young and female students characterized by low incomes.

  Finally and maybe not completely seen as a limitation, but as an opportunity for future research: Although the developed scale is characterized by primary satisfying characteristics (especially regarding calculated reliability estimates), further scale improvements could (still) be indicated – especially regarding the seventh subscale to record usage of measures to extend food’s shelflife at home. For example, by examining alternative item formulations, creation of additional useful items or possibly additional factors/ subscales or further scale characteristic estimations (other reliability estimation measures; see e.g. Schermelleh-Engel & Werner, 2012). Additionally, conducting confirmatory factor analyses in order to validate the proposed factor structure would be another appropriate objective for future research.

5. Conclusion

  So, based on the presented results, an efficient (online) tool/ measuring instrument was developed that could be easily used by currently existing or future food waste-prevention initiatives to further promote household food waste-prevention by restricting provided action knowledge on these a person (still) not perform (sufficiently). Therefore, the perceived relevance of the provided information will increase, expected to increase probability of consistent and durable improved performances of relevant food waste-preventing behaviors. Consequently, such improvements will probably result in fewer amounts of household food waste and so, in relevant reductions of its negative consequences on our environment.

6. Acknowledgement:
 (1) Similar assumptions concerning household’s energy saving-promotion by providing appropriate action knowledge were madeby other authors considering the effectiveness of energy saving campaigns, for example, by Nachreiner and Matthies (2016).
 (2) These two assessments were part of a superior research project focusing on household food waste-prevention (see Schmidt, 2016 for further information).
 (3) Since both assessments were part of a superior research project to further examine household food waste-prevention by conducting an intervention for its promotion, correlations between t1- and t2-data could not be used to calculate retest reliability estimations.

Back to Top: (1), (2), (3)

7. References