Consumers and their behavior: state of the art in behavioral science supporting use phase modeling in LCA and ecodesign

Behavioral scientists possess a variety of tools to arrive at precise observations of behavior and other relevant variables. While the purpose of many studies in environmental psychology and related fields is undoubtedly to understand the driving factors of sustainable behavior, the employed research methods can be equally useful to observe and describe the behavior as such. Many studies in environmental psychology and related fields rely on self-reported behavior in the form of surveys. With surveys, researchers may access a large sample of people and question them not only about their behavior but also about psychological variables such as their attitudes, beliefs, and expectations. Surveys allow for a high degree of external validity, in the sense that they capture how people think “out there”. However, the way a questionnaire is designed is important and may substantially affect the findings. Significant research efforts have been made towards the development of empirically validated questionnaires for specific variables of interest, such as intentions to perform a specific behavior or attitudes towards the environment. As a result, several probed questionnaires are available in the literature (e.g. Milfont and Duckitt 2010; Klöckner and Blöbaum 2010).

Surveys are particularly interesting in the context of a LCA, because variables that can be assessed by means of a questionnaire can typically be gathered on a large enough scale to be representative of the population of interest. For example, Laitala et al. (2012) employed representative surveys to assess how people use and maintain their clothes. Moreover, several large-scale projects exist that regularly survey representative samples and make data accessible for other researchers free of charge (e.g., Eurobarometer surveys). Thus, in principle, surveys can allow LCA researchers to get some indication of the types of user behavior (and their distribution) in the general population.

A potential downside of surveys is that they often measure only intention but not the actual behavior. This is particularly relevant in the context of sustainable behavior where people are likely to provide socially acceptable answers, rather than the truth. Moreover, research has shown that thinking “green” is often not the same as acting “green” (Davies et al. 2002; Pickett-Baker and Ozaki 2008). Possible reasons for this include the fact that people tend to overstate their engagement in behaviors that are socially desirable (see for example Nederhof 1985), which is the case for environmentally relevant behaviors or—when people are asked to report about past behavior—simply memory problems (Corral-Verdugo 1997). For example, the answers collected from a recent Eurobarometer survey containing questions such as “In the past month, have you separated most of your waste for recycling” from representative samples of all European Union member states (Special Eurobarometer 416) might be more informative on how people like others to see them, rather than on actual behavior.

Therefore, significant efforts have been made to develop more direct tools to measure behavior, which is facilitated by the advent of new technologies. Modern communication technology allows for example to prompt users repeatedly during the measurement period to answer questions about current (or recent) behavior (Shiffman et al. 2008; Klöckner and Blöbaum 2010). Although still relying on self-reported behavior, the repeated sampling minimizes biases due imperfect recall and is more accurate.

Rapid developments in sensor technology and computing power further allow observing behavior directly on a larger scale. Smartphones, for example, now contain GPS sensors as well as accelerometers, which collect detailed information about physical activity and transportation (Zafiroglu et al. 2012). Logan and Healey (2006) describe how activities of daily living can be classified by sensors, and the vision of the internet of things described for example in (Gubbi et al. 2013) suggests that detailed large-scale data on user behavior of appliances such as fridges and washing machines might be available soon. Many research facilities now possess facilities to create virtual environments and observe or model the behavior of humans in the interaction with a very specific and controlled environment, which may be interesting for testing new designs in the form of a digital prototype (Gabbard and Swan 2008; Carsten and Jamson 2011; Bennadji et al. 2015).

While for now direct observation of behavior is typically too costly and cumbersome to perform on a scale that is representative of the general population, data from relatively small samples can still provide the required information to determine the characteristics of main user scenarios (e.g., average, upper or lower bound) realistically or at least to make a judgment whether current assumptions are reasonable (Scott 2005). Moreover, direct observation of a small sample can be combined with a representative survey, which is more suited to arrive at accurate estimates of the distribution in the population. Combining the two can help in the development of relevant survey questions, which can then allow extrapolating what is seen in direct observations of behavior to a larger scale (Falk et al. 2013).

One important assumption about the use phase in LCA concerns the possibility for behavioral spillover or rebound effects. That is, environmental impacts of a product may not only be generated by the product itself, and the use of the product, but also by a change in user behavior that occurs as a consequence of one or several attributes of the product (Hertwich 2005; Hofstetter et al. 2005). According to Vivanco et al. (2015), who analyzed how rebound effect has been modeled in several LCA case studies, consumer behaviors are central elements for assessing direct and indirect rebound effects, whereas market and economic behaviors are used for structural and transformational rebound effects.

It has long been acknowledged that product attributes can lead to changes in behavior and demand. A famous example of such indirect effects is the change in consumption that can occur if more household income is available, e.g., because a product is cheaper than another or because more energy-efficient appliances lower the overall household energy expenditure. Money is not the only scarce resource that can mediate changes in consumption: Time, space, and skills have been suggested, among others, to bear the potential to affect consumption (Hofstetter et al. 2006). For example, if certain time-consuming activities such as transportation become faster, more time becomes available to do either more of the same (e.g., travel to more places, see Chen and Mokhtarian 2006) or more of something else. This effect is also thinkable for fairly small but frequent actions, such as making coffee with a more automatized coffee machine.

Demand changes can not only result from changes in money or time budgets but have also been described in the behavioral literature as a result of interventions (Truelove et al. 2014; Dolan and Galizzi 2015). For example, an intervention designed to lower water consumption was effective at achieving this goal, but at the same time increased energy consumption (Tiefenbeck et al. 2013). Likewise, if people perceive a product as more environmentally friendly, they may be likely to use it more. On the other hand, they may instead feel motivated to use it in a resource efficient way or reduce energy consumption with other products as well (Lanzini and Thøgersen 2014; Steinhorst et al. 2015). Thus, both positive and negative spillover effects of intervention are thinkable.

As outlined by Girod et al. (2011), such indirect effects are incompatible with the frequently made assumption that demand will be constant across different products that are compared within an LCA study. For example, Humbert et al. (2009) compare different preparation modes for one cup of coffee. Given that preparing capsule espresso coffee is very convenient, the availability of a capsule espresso machine may well alter demand. Likewise, if one product is cheaper than another, this frees up household income for other potentially impactful consumption. As long as the comparison is made on equally sized functional units, such differences are neglected. Girod et al. (2011) suggest modeling the change in consumption directly, for example by assuming that households will simply consume more of what they already consume.

Again, in order to model behavior accurately, it is useful to support these assumptions by an understanding of how people actually respond to increased access to faster services or more income, and observational data can be used in order to inform estimates. For example, Thiesen et al. (2008) suggest that presumably not all households will respond in the same way to a marginal increase in income, but their response may be dependent on characteristics such as their income level. Therefore, estimating changes in demand from aggregate, e.g., country-level, data may be inaccurate. They then use data from household expenditure surveys to estimate separately for different levels of income how a small change in income might affect consumption (Thiesen et al. 2008). This allows for a more precise estimation of the environmental impact generated by the additionally disposable household income. As illustrated by Thiesen et al. (2008), this effect can be substantial.

A major hallmark of sustainable behavior is that its benefits will mainly become evident in the future. Therefore, it is important to understand how people decide when some or all of the outcomes of their decisions are going to happen in the future. Outcomes that materialize in the future tend to be perceived as less important than the ones that occur immediately, and people differ from each other in how much they care about future outcomes. Temporal discounting is important for a wide range of behaviors with environmental impact. The reason for this is twofold: the environmental benefits of choosing an energy saving option today are perceived as less important simply because they are distant in time and moreover somewhat uncertain. This may have important implications especially for interventions aimed at encouraging the use of more efficient (but with a higher purchase price) house appliances (Newell and Siikamäki 2013). Secondly, even the monetary costs of paying for energy consumption is typically incurred temporally separated from consumption. Using the car is not associated with paying at the moment it is used, but only when the car needs fueling up. Likewise, individuals pay for their energy and resource consumption in the household in the form of their utility bill, which comes at most every month. This temporal separation makes it less likely that people will reduce their energy consumption in order to save money, because the delayed money savings are simply not so relevant at the time of consumption. The use of feedback allows connecting behavior to its material consequences (e.g., financial consequences) by signaling it more closely in time, thus breaking down the effect of delay.

Often individuals do not receive feedback on the impact of each individual behavior. How much of a utility bill is attributable to which specific action is usually not clear. A recent field experiment demonstrates that a tighter and more specific coupling of feedback to the behavior can help to decrease consumption (Tiefenbeck et al. 2013). A device measuring water temperature and flow in the shower was installed in 700 households. In one third of the households, the device only indicated the current water temperature, but did not provided feedback on consumption. This group served as the control condition. In the other households, the device displayed real-time information about water and energy use. Compared to control households, the households receiving feedback decreased water and energy use in the shower by about 20 % as soon as the intervention was started, and this effect was sustained over 2 months, suggesting that the effect did not wear off in this period. Thus, linking costs with actual consumption represents a promising avenue for design intervention behavior change aimed at reducing the impact of a product. For example, this might translate into providing positive/negative feedback (e.g., amount of energy consumed) to the user when interacting with the product (Wilson et al. 2015).

Another robust finding in behavioral studies is that people are strongly influenced by defaults. That is, if one of the available options is labeled as the default that will be implemented unless an active choice is made and another option is selected, this default option will be chosen much more often (Kahneman et al. 1991). A related phenomenon is the status quo bias: people are unlikely to make changes to an existing state or condition even when this would be beneficial. A variety of factors is thought to contribute to this: people may perceive the availability of a default as a signal that this option is superior to the others or that many other people will chose this option. Also, people are generally reluctant to change the status quo because they pay more attention to the advantages of what they have, than to the advantages of the alternative options. Lastly, people can simply be inattentive or even have a strong preference to avoid making a decision, in which case they will also welcome the default option. A natural field experiment at a large Swedish University found that the adoption of a double-sided printing default setting in the printing machines led to a substantial reduction, with a significant and immediate effect in the form of a 15 % drop in paper consumption, and with that effect staying stable over time. Notably, the effect of the double-sided default has been found to be far larger than that of a 10 % tax on paper products, which would produce a mere 2 % reduction (Egebark and Ekström 2013). Therefore, products which include green default settings are more likely to result in reducing the environmental impact of the use phase.

Human behavior is best understood within a social context, as it is often shaped by the presence and behavior of others. This means that drivers of behaviors are not only explained by people pursuing their own self interest, but also by their social preferences (such as reciprocity, altruism) and self-image concerns (such as reputation, self-respect and status) and perceived behavior of others (social norm). Therefore, the social context is another extremely relevant aspect of pro-environmental behavior.

While the integrity of the environment and the availability of natural resources depend on the contribution of each single individual, the benefits of these are shared by everyone, irrespective of their contribution. That is, they represent a “social dilemma”. Recycling of household waste is an example for this: collecting and recycling takes effort and commitment by the individual, but the resulting environmental benefits are enjoyed by everyone in the community and thus also by those who do not recycle. Likewise, the environmental impacts of individuals not recycling their waste are also shared equally by everyone. Thus, environmentally relevant behaviors impose externalities on society. In this condition, if individuals simply maximize their own benefits, pro-environmental behaviors are unlikely to occur. Yet, moral and social motives can influence people’s willingness to reduce negative externalities they impose on others.

A large amount of research on social dilemmas highlights various ways to increase pro-social behavior. For example, Alpizar et al. (2008) have found that people contributed higher amounts to the preservation of a national park if they received a small gift before deciding on their contributions. A large field experiment on carbon offset found that bus customers who were asked to contribute to a carbon-offsetting scheme while buying their tickets contributed more when the company offered to match their contribution (Kesternich et al. 2014). Thus, it appears that inducing feelings of reciprocity and fairness can in principle increase pro-environmental behavior.

People might also act environmentally friendly not (only) because they are intrinsically motivated to do so, but because they care about the way they are perceived by others. Receiving information about the behavior of others can be a strong personal motivation to spur compliance with environmentally responsible behavior. For example, a recent study has shown that reputation influenced the adoption of an energy program that actively reduced individuals’ energy use during periods of high electricity demand, so to protect the entire system from blackouts. Individuals were more likely to adopt the program if their behavior could be observed by others (e.g., signing up with their names on sheets provided in communal spaces), and the effect was stronger for people living in apartment buildings, where interactions with their neighbors are more likely to occur compared to individual homes (Yoeli et al. 2013). Also, people generally perceive the behavior of the majority as setting the norm for acceptable behavior. This is an important insight for framing messages designed to influence people’s behavior. If possible, it is preferable to highlight that a majority engages in a desirable behavior such as recycling, rather than highlighting that some people fail to do so. Goldstein et al. (2008) applied social comparison drivers in a field experiment in hotels showing that messages like “the majority of guests reuse their towels” encouraged more hotel guests to do the same compared to messages simply stressing the environmental benefits of towel reuse. Interestingly, this effect was even stronger when the message was referring more closely to the guests’ immediate context, such as “the majority of guests in this room reuse their towels”, possibly leading to a closer identification of the hotel guests with the majority. This motive has also been successfully employed in a field experiment investigating an energy conservation program. Using data on energy usage from 600,000 households, it was one of the largest randomized field experiments in history. OPOWER mailed home energy report letters that compared a household’s energy consumption to that of similar neighbors. These letters include messages like: “Last month you used 15 % less electricity than your efficient neighbors”. The average treatment effects of OPOWER’s conservation programs ranged from 1.4 to 3.3 % of baseline usage, with a mean reduction in energy consumption of 2.0 % (Allcott 2011; Allcott and Rogers 2014). The cost-efficiency of this conservation program compared favorably with other energy efficiency programs (Allcott and Mullainathan 2010). As highlighted in Schultz et al. (2007), caution should be paid in designing social norm interventions so that they do not reduce pro-environmental behaviors of those that are already “above” the norm. Taken together, studies on social norms suggest that designers might envisage products that enable interactions between users, thereby allowing for feedback based on social comparison and eliciting feelings of reciprocity and fairness. This might be more feasible for those appliances that automatically collect information on user behavior (e.g., sensors classifying daily living activities).

A further important insight on behavior in the context of sustainability is that there is a strong influence of habits. Habits are learned behavioral patterns that are performed without much conscious deliberation and instead strongly controlled by the stimulus environment (Dezfouli et al. 2014; Orbell and Verplanken 2015). Indeed, actions such as turning the light on/off, disposing waste, running the bath or shower are often done in a repetitive and almost unconscious way. Using data from a field experiment in Sweden, Eriksson et al. (2008) suggested habitual behavior to be a key factor in choosing means of transport. While habits are useful to free up cognitive resources for more important decisions, they may become particularly problematic when dealing with conditions for behavior change.

Environmental stimuli play an important role in maintaining habitual behavior patterns, and as such they could be employed also to change them. For example, several studies have found that people are more likely to establish new behavior patterns after moving to a new environment (Bamberg 2006; Dolnicar and Grun 2009; Walker et al. 2014). In an attempt to decrease food waste in the well-defined environment of hotel buffets, Kallbekken and Sælen (2013) have identified some simple changes to decrease waste caused by guests loading too much food on their plate. Using a large number of comparable hotels, the researchers were able to systematically test and compare two interventions to a control group. Simply by using smaller plates, people were better able to estimate the quantity of food that they wanted and reduced food waste by 20 %. The same effect was achieved by explicitly inviting people to visit the buffet repeatedly, rather than taking a lot at once. This finding is in line with the observation that people have difficulties predicting their own visceral states, in this case how much food they actually need.

Environmental stimuli (such as the way a particular product is designed) can also stimulate a particular behavior without a history of learning and habit formation if they naturally enable or encourage a particular behavior, a concept known as “affordances” (Gibson 1979). This might be particularly relevant for product design, where modifications in the way affordances (e.g., possibility of interacting with the product) are created can be used to design products that support environmentally friendly behavior (Srivastava and Shu 2013a).

The effect of environmental stimuli on choices has also been successfully applied to increase the proportion of healthy food chosen and consumed in canteens. Making healthier options more easily accessible than less healthy options consistently increased the choice of healthier foods (Hanks et al. 2013). Similar techniques could be tried towards decreasing meat consumption. A recent survey from the Netherlands suggests that, other than commonly thought, a sizable majority of people do not have a strong preference to consume meat frequently (Dagevos and Voordouw 2013). This group of people might be receptive to small changes in the choice environment and choose less environmentally impacting alternatives if this was made easier or more appealing.

Rational models of human decision making assume that individuals will understand and process all the available information. However, in the face of complex decisions and limited attention and time, people tend to use so-called “heuristics”, which are smart rules of thumbs, to simplify the problem. The use of such heuristics is particularly important when understanding how users react to information on the energy consumption and environmental impact of a product. This might well explain why in the field of sustainable consumption attempts to increase consumer knowledge through provision of information have shown little impact on behavior (Davies et al. 2002; Gardner and Stern 2002; Pedersen and Neergaard 2006; Moisander 2007). Moreover, if products differ on several dimensions such as energy efficiency, pollution, durability, and price, it is unlikely that the decision maker will integrate all of these into his choice. Rather, he will attend to a selection of these, for example the ones he understands best or considers most important. For example, in a study of alternative labeling schemes for energy-efficiency appliances, labels providing simple information on the economic value of saving energy represented the most important element guiding more cost-efficient purchases, while the addition of further information (e.g., on carbon dioxide emission or energy use) had little effect (Newell and Siikamäki 2013).

Complex choices can lead to information overload and ultimately demotivation (“choice overload”, (Chernev et al. 2015). Under a “too many choices” condition, individuals become overwhelmed by the number of options and may accept even more the standard option. It is thus important to present relevant information as simple and intuitive as possible: for example, it has been shown that the most common US measure of fuel efficiency, namely the “miles-per-gallon” measure, is not intuitive (Larrick and Soll 2008). As a result, this measure may lead to a significant underestimation of benefits of replacing most inefficient vehicles. By contrast, knowing that people are more likely to understand if a continuous measure (e.g., water use) is broken down into discrete unit, products that present information on resource use in discrete units (e.g., in cups), might work more effectively in reducing resource consumption (Srivastava and Shu 2013b).