The effectiveness of social norm information in stimulating financial help-seeking behavior: A field experiment

ABSTRACT Early contact with people who need financial help is key in preventing financial derailment. It can prevent avoidable debt accumulation and accompanying psychological problems. In this field experiment, we investigated whether a descriptive social norm message mentioning an absolute number of customers who had previously reached out for help motivates customers of a Dutch mortgage lender to get in contact when they expect financial problems in the near future. The descriptive social norm letter did not result in more contacts. The results of our research contribute to the literature on social norm communication and debt prevention. Moreover, we present insights into the challenges associated with experimental field research on social influence and provide recommendations for future field work.


Introduction
Financial hardship is a worldwide problem.A recent poll by the European Parliament revealed that 58% of people have experienced financial difficulties since the start of the coronavirus disease (COVID-19) pandemic.Fourteen percent of the respondents indicated that they had difficulties paying bills, bank loans, or rent (Banks, 2020).However, even before the COVID-19 outbreak, in the Netherlands, for example, 1 in 5 households had been experiencing financial problems (Schonewille & Crijnen, 2018).Oftentimes, financial problems grow quickly.Typically, two-thirds of a problematic debt consists of the fines and interests that creditors charge according to the Dutch Association for Public Credit (NVVK, 2018).These extra costs can be prevented when an arrangement to pay off one's debt is made on time.
Typically, accounts receivable departments only contact customers with outstanding debts.However, shifts toward preventive debt management can be observed (Authority Financial Markets Financial Conduct Authority, 2016;Roberts, 2001;[AFM], 2013).The goal of preventive debt management is to establish contact with customers at an early stage to find solutions that suit the customers' current financial situations (e.g., payment breaks, payment arrangements, or a job coach when someone is unemployed) to ensure that debts can be prevented or do not increase unnecessarily (AFM, 2013;FCA, 2016).To arrange this, it is essential to establish contact between a company and a customer whose financial situation seems to have worsened (AFM, 2013).
Although much attention and focus have been given to preventive debt management by both companies and financial regulators, getting in touch with customers dealing with financial problems is still difficult (AFM, 2013;FCA, 2016).Many customers do not seek contact, even when companies proactively reach out to them (Keizer, 2016).There can be several reasons for not seeking help.People with financial problems often try to deny their situation for as long as possible (i.e., the ostrich effect; Karlsson et al., 2009) or expect to work them out in the near future (i.e., optimism bias; Sharot, 2011).Shame also plays an important role in the reluctance to seek contact with one's creditor (Keizer, 2016;Walker, 2013).In addition, many people feel guilty about their debts, even when external influences such as job loss or incapacity for work cause the debt.They feel that they have to solve it themselves because it is their responsibility (Tiemeijer, 2016).In addition to the ostrich effect, optimism bias, and shame and guilt, those with financial problems often have to deal with multiple challenges, causing stress and resource depletion (Mani et al., 2013).This can lead them to ignore their financial situation and not seek help (Babcock, 2018).Another reason people do not seek contact with their creditors is that customers are unaware of the desired action, the kind of help that is possible (Tiemeijer, 2016), and the extent to which companies help their customers in financial need (AFM, 2013).
When people are unaware of the correct behavior in a certain situation, they are more likely to look around and determine what others with whom they identify are doing in a similar situation (Cialdini & Goldstein, 2004).Among the first researchers to demonstrate this was Sherif (1937), who showed that people align themselves with others' judgments when put in an ambiguous situation.Since then, much research has been conducted on the influence of the social environment on people's thoughts, feelings, and behavior (e.g., Asch, 1952;Cialdini, 2009).In their focus theory of normative conduct, Cialdini et al. (1990) make a distinction between two types of social norms: descriptive and injunctive.While descriptive norms pertain to the perception of what most others do in a given situation, injunctive norms are a person's perception of what most others think one should do in a certain situation.Therefore, the latter concerns what behavior most members of a group approve or disapprove.Both types of social norms have been found to influence people's choices (Manning, 2009;Mollen et al., 2013).In the financial domain, studies have shown that using a social norm message can persuade customers to pay their social security and tax arrears (Hallsworth et al., 2014;Larkin et al., 2019).The results of these studies demonstrate that using a descriptive social norm message indicating that the majority exhibit the desired behavior (e.g., 'Nine of 10 people in the UK pay their tax on time.') is highly effective in persuading people to pay their taxes.
In seeking contact when experiencing financial problems, most people do not (yet) perform the desired behavior (i.e., seek contact with the creditor).Therefore, communicating a descriptive social norm message claiming that the majority seek contact with a company when experiencing difficulties in paying bills is not an option for ethical reasons.Moreover, a relatively new type of descriptive norm message, the utilization of trending or dynamic norms (e.g., Mortensen et al., 2019), in which the number of people engaging in a behavior is increasing, is also not an option because this trend is not yet visible in the Netherlands.
Other forms of norm communication have been applied in prior research.For instance, Cialdini et al. (2006), in their study on the prevention of petrified wood theft, communicated a descriptive-norm sign 'Many past visitors have removed petrified wood from the Park, changing the natural state of the Petrified Forest'.The results of the study showed that norm messages without a statement on the majority can also be effective in inciting people to the undesired behavior.In this study, the descriptive-norm message resulted in more theft of petrified wood compared to the injunctive-norm message stated " Please don't remove the petrified wood from the Park, in order to preserve the natural state of the Petrified Forest" (7.92% vs. 1.67%).
An alternative is to communicate in terms of absolute numbers when the majority of people do not perform the desired behavior.In their study, Mollen et al. (2013) found that communicating that a relatively large absolute number of people make a healthy food choice affected people's food choices in a cafeteria setting.A message that communicated that 'more than 150 students consume salads every day' resulted in significantly healthier food choices, compared to a control group receiving no norm message.
In experimental field studies on social norms, where the majority does not yet display the desired behavior and deception (due to problems with immediate debriefing) is unethical, communicating absolute numbers of people who engage in the desired behavior can be an alternative.In the current field experimental study, we therefore investigate whether communicating that a relatively large absolute number of people seek contact with their creditor as a descriptive norm message can convince people to act in a situation where the majority do not yet exhibit the desired behavior.
The purpose of this field experiment was to investigate whether communicating that an absolute number of people seek financial help as a descriptive norm can encourage customers in financial need to contact their creditors and ask for help.This leads to our research question: Can mentioning an absolute number of people who seek financial help as a descriptive norm persuade customers in financial need to contact their creditor?
The social impact theory (Latané & Wolf, 1981) offers a model that combines two streams of previous research on social influence: the influence of the majority on a passive minority and the influence of active minorities on a silent majority.This theory proposes that majorities and minorities have simultaneous and mutual influences.Social influence is perceived as the result of several social forces operating in a field of social force.The amount of influence depends on the strength (S) of the source person (e.g., power and status), the immediacy (I)/proximity to the target, and the number (N) of source persons.This can be produced by either the majority or the minority.The aim of our study was to persuade customers to seek contact when experiencing financial problems.By indicating that a large number (N) of customers (S) in a similar situation (I) have received help by contacting their mortgage lenders, we aimed to maximize the impact of our message.
According to Cialdini (2009), social proof exerts the greatest influence under two conditions: uncertainty and similarity.We are more likely to follow people who we identify with than those who we do not identify with (Burn, 1991;Schultz, 1999;Tajfel, 1974), and people are more prone to follow others when they are unsure of how to act in a certain situation (Cialdini, 2009).Several studies have shown that descriptive norm information is particularly influential in situations where people are unfamiliar with or feel insecure or ambiguous about the right course of action (Cialdini, 2001;Griskevicius et al., 2006).Moreover, research by Jacobson et al. (2011) indicates that the influence of a descriptive norm on behavior increases when people are under a cognitive load.When people fall into debt, they experience stress, which causes resource depletion (Mani et al., 2013).A wellknown trigger for financial problems are life events accompanied by a decrease in income or an increase in spending (e.g., job loss, divorce, or the inability to work).Such life events have a destabilizing effect on people's financial situations and are therefore important triggers for the emergence of problematic debts (Westhof et al., 2015;Schonewille & Crijnen, 2018).Often, these people are unfamiliar with the situation where they cannot pay their bills anymore and will therefore be sensitive to how they think others behave in this situation (Cialdini, 2009).A descriptive social norm message can guide them to the right action.
This was also demonstrated in a study by Hallsworth et al. (2014).They conducted several field experiments among people who had already declared their income to be taxed but had not yet paid the tax.In their studies, they tested different types of messages and their effects on payment behavior.The results showed that tax payment reminders in which a descriptive norm message was used considerably enhanced tax compliance.Depending on the message, payment increased by 1.4% (general descriptive norm) to 5.0% (local and descriptive norm).In one of their studies, they also examined and compared the impacts of descriptive and injunctive norm messages and found that descriptive norm messages had a 1.44% (p < 0.000) larger effect on payment than injunctive norms.Larkin et al. (2019) conducted an experiment in which they also used a descriptive social norm message to increase tax collection in local government taxation in the United Kingdom.Their results showed that the amount of payment increased by 12% points compared with that in the control group.Although these studies found that descriptive norm messages had positive effects on tax payment compliance, some failed to produce positive results (e.g., Blumenthal et al., 2001).The key difference between the experiments was the targeted group that received the message.Hallsworth et al. (2014) and Larkin et al. (2019) focused on the payment of non-compliant taxpayers, whereas Blumenthal et al. (2001) focused on the reporting of liabilities.The reason Blumenthal et al. (2001) did not find an effect might be that the social norm message was also sent to a group for whom the message was not relevant.They focused on the generic case of all taxpayers, who may or may not be compliant.Thus, communicating a descriptive norm message has a positive effect on people's payment behavior, especially for people with payment arrears (Hallsworth et al., 2014;Larkin et al., 2019).

Present study
Compared with the study of Blumenthal et al. (2001), in our study, we target people classified as having a high risk of payment arrears by the mortgage lending company by using their actual payment behavior in the previous year to ensure that messages are sent only to people for whom it is relevant.In contrast to previous studies in this domain (e.g., Hallsworth et al., 2014), the majority of customers in our study do not show the desired behavior of seeking contact when they experience financial problems.A statement that the majority shows the desired behavior (i.e., descriptive norm) can therefore not be used in the present field experiment.We therefore investigate whether communicating a specific type of descriptive norm message in terms of absolute numbers affects whether customers with more than two payment arrears in the past 12 months will contact the company to make a payment arrangement.To ensure similarity and proximity to the source, we emphasize the help-seeking behavior of many (N) other customers of the mortgage lender (I).
To investigate the effect of a descriptive social norm message on the number of customers who will contact a creditor for help, two different letters are sent to a specific group of customers (see Appendix 1).In one letter, an absolute number of people who reached out for help is included as a descriptive social norm message.The other letter has a similar content, except that it does not contain a social norm message.
The hypothesis is that a descriptive social norm message reporting an absolute number will elicit more contacts than a control message that does not contain a descriptive social norm message.

Participants and design
The study was conducted with 297 customers of a Dutch mortgage lender who were classified as high-risk on the basis of their payment history.These customers had failed more than twice to pay their bills on time in the past 12 months.In addition, they had no outstanding debt when the experiment started and had not received any help from the debtors' department in the past year.All 297 customers met the inclusion criteria.We selected customers who had a high likelihood of needing help but who had no previous experience with the bank's accounts receivable departments.Three customers were excluded from the analyses: two customers who had no arrears in the last 12 months were mistakenly selected, and one customer was under investigation for fraud, resulting in the final sample of 294 customers.From among the 294 customers, we eventually managed to get in touch with 157 people.Seven made the phone call themselves, and the other 150 customers were contacted after 14 days after receiving the letter by the mortgage lender's outbound call center.When contact with a customer was made, additional questions were asked.Not everyone wanted to cooperate or answer all questions.As a result, the number of people who answered the different questions varied.See the corresponding n values per measure in the Measures section.
The study started in mid-April 2021 and ended in mid-June 2021.Customers were randomly assigned to one of the two conditions in a single-factor between-subjects design (type of letter: descriptive social norm vs. no-norm control).Before we analyzed the data, the research was preregistered (https://osf.io/sgby3/?view_only= 8d9e5b25271046e4a9eba0276e421e7c) and approved by the ethics committee of the second author's university (2021-PC-13959).

Procedure
All 297 letters were sent to customers via the Dutch postal services on 16 April 2021.Customers who did not contact the mortgage lender within 14 days were contacted by call center agents to ask if they needed help with their finances.All customers who participated in the field experiment were given a special code in the debtor system.This allowed for all actions taken to be registered (e.g., whether the customer had contacted the mortgage lender after receiving the letter or whether the company's call center agents called after 14 days).When contact was made with a customer, additional exploratory questions were asked to measure the customer's attitude toward the letter and the mortgage lender and their knowledge of the financial help options the mortgage lender can offer.In addition, the customers were asked if they thought that they or others would seek contact when they would experience payment difficulties in the future.

Manipulations
To test the effect of a social norm message on the degree of contact, two letters were written: one with social norm information and an almost identical letter without it.As the company had spoken to 8,376 customers in the previous year and helped this group of customers, this number was included in the descriptive norm letter.
The purpose of the letters was to persuade customers to contact the mortgage lender when they needed help because they expected or actually experienced financial problems.The only difference between the two letters was the descriptive social norm information about how many customers contacted the mortgage lender in the past year.The descriptive social norm message included the following sentence at the end of the first paragraph: 'Because 8,376 customers contacted us, we were able to help them.'In addition, in the middle part of the letter, a reference was again made to the customers who had contacted the mortgage lender: 'Just like the aforementioned customers, please contact us.' See Appendix I for the two versions of the letter.

Dependent variable
The dichotomous dependent variable initiative contact was assessed through the mortgage lender system.Customers could either seek contact themselves (n = 7) or not seek contact at all (n = 287).

Exploratory variables
For mortgage lenders, it is important to get an idea of whether people will contact them if they experience financial problems at present or expect problems in the near future, as in those cases, financial derailment can be prevented.For this reason, a standardized explorative questionnaire was used by the call center agents when they talked to the participating customers.This gave the mortgage lender a better idea of what possible barriers could be to seeking help.They asked the same questions to both the customers who initiated the contact themselves and those who were called by the agents.The questionnaire measured the following variables:

Attitude mortgage lender
All customers (N = 157) were asked what they thought of the entire campaign (letter and follow-up call; 82.2% positive, 10.2% neutral, and 1.9% negative).Only a few customers (5.7%) did not provide an answer.

Attitude letter
All customers who read the letter (n = 97) were asked what they thought of the letter (85.6% positive, 9.3% neutral, and 2.1% negative).A few customers (3.1%) did not answer the question.

Knowledge
All customers who were contacted were asked (N = 157) if they were aware of all services that the mortgage lender offers to help them even before payment is in arrears.Forty-nine customers indicated that they had no idea of the help that was available (31.2%).
(Perceived) future help seeking All customers were asked two questions about seeking contact with the mortgage lender when facing financial difficulties in the future.The first question was whether the customers thought that they themselves would seek contact when they would experience payment difficulties in the future.One hundred thirty-six customers answered this question (n = 136).The second question was whether customers thought that other people would seek contact with the mortgage lender when they would experience payment difficulties in the future.Ninety-four customers answered this question.

Hypothesis test
To test our hypothesis that a social norm message would elicit more customer-initiated contact, we performed a chi-square test.The manipulated independent variable was the type of letter, and the dependent variable was whether the customer took the initiative to contact the mortgage lender.A significant difference between conditions was found [χ 2 (1) = 7.781; p = .005;95% confidence interval (CI), .41-.53].However, the proportional odds assumption was violated.According to this assumption, the frequency in each cell must be > 5.Only seven of the 294 customers contacted the mortgage lender themselves.All these customers received the no-norm control letter.Therefore, none of those who received the descriptive social norm letter contacted the mortgage lender of their own accord.This is in contrast to the expectation.However, as the assumption of this test was violated, no conclusions can be drawn regarding the effectiveness of descriptive norm messages on contact seeking by customers.In conclusion, the hypothesis could not be tested.

Exploratory analyses
Attitude.We also conducted an exploratory analysis on the customers whom the mortgage lender had contact with (N = 157).These were customers who initiated contact within 14 days and those who were contacted by the mortgage lender through phone calls.We used an analysis of variance to compare the two groups (descriptive social norm vs. no-norm control) to determine if attitudes toward the mortgage lender and the letter differed.Our analyses revealed that receiving a descriptive social norm message did not result in a different attitude toward the mortgage lender (p = .98)or the letter (p = .94)compared with receiving the no-norm control letter.

(Perceived) future help seeking
In our final analyses, we first coded the two open answers to reflect whether customers thought they, or others, would contact the mortgage lender in the future when experiencing financial problems (yes or no).The first analysis of future help seeking was performed to determine whether customers themselves would contact the mortgage lender when facing financial difficulties in the future.We performed a chi-square test to determine whether differences existed between those who received a no-norm control letter and those who received a descriptive social norm letter.No significant difference was found between the two conditions (p = .95).
The second analysis of future help seeking was performed to determine whether customers thought that others would contact the mortgage lender when facing financial difficulties in the future.We performed a second chi-square test to determine whether differences existed between those who received a no-norm control letter and those who received a descriptive social norm letter.A significant difference was found [fjfj 2 (1) = 18.362; p < .001;95% CI, .04-.33].Those who thought that other customers would seek contact in the future when experiencing financial difficulties were fewer among those who had received the descriptive social norm letter than among those who had received the non-descriptive social norm letter (for the percentages, see Table 1).This contrasts what one may have expected given that the social norms message communicated that a large number of people have contacted the mortgage lender in the past.

Discussion
In the current study, we investigated and compared the effects of descriptive social norm and no-norm control letters on contact-seeking by mortgage lender customers for preventive debt management.The descriptive social norm message included the absolute number of customers who contacted a mortgage lender for help in the past year.The nonorm control letter was similar, except that it did not contain a social norm message.We hypothesized that a descriptive social norm message reporting an absolute number of customers who had received financial help in the past year would elicit more contacts than a non-norm control message.The descriptive social norm message did not elicit more contacts than the no-norm control message.Additional analyses show the two letters did not result in differences in attitudes toward the mortgage lender and the letter.And that those who believed that other customers would contact their mortgage lender in the future when facing payment difficulties were fewer among those who received the descriptive social norm letter than among those who received the no-norm control letter.An explanation for this might be that respondents did not perceive an absolute number of 8,357 customers who previously contacted the mortgage lender for help as high and therefore concluded that not that many people would seek contact in the future when facing payment difficulties.In our study, we did not perform a pretest nor include a manipulation check question, so we could not determine how customers interpreted the absolute number.Moreover, the response rate to the question on whether they thought others would contact the mortgage lender was rather low (33%).The responses were rated using a dichotomous answering scale.Future research should obtain a larger response rate and use a more detailed scale such as a slider ranging from a minority to a majority.As a result, a variance analysis can be used so that smaller differences between conditions can be observed.By including a clear manipulation check question that asks participants whether they think that a minority or majority contacts the mortgage lender in case of (expected payment difficulties), the success of the manipulation can be better determined.Even better would be to confirm this beforehand by conducting a pretest.
Mentioning the absolute number of 8,357 clients who had contacted the mortgage lender and had received help by the mortgage lender did not have the desired effect possibly because the customers had difficulties estimating if this number is high or low.To assess this, they need a frame of reference, which most customers likely do not have.In many other social norm studies, proportions or percentages are more frequently communicated (Hallsworth et al., 2014;Larkin et al., 2019), which are easier to interpret in terms of a majority or minority.Future research would ideally examine the effect of mentioning the proportion of customers who had contact and had received financial help for their payment arrears from the mortgage lender.However, in many field experimental studies this would likely entail deception, when companies do not have this data or when a majority does not yet perform this behavior, which can be unethical without the possibility for an appropriate debriefing.In this study we did not gave a debriefing and speak to almost half of the customers who participated.If you were to apply deception, you cannot be sure whether this group of customers will be debriefed correctly.And that is the main reason not to apply deception in a field experimental study.
In addition, previous studies on social norm communication in this domain have examined payment behavior instead of seeking contact as an outcome measure (Hallsworth et al., 2014;Larkin et al., 2019).Although we could not formally test our hypotheses, the percentages in our study may indicate that mentioning an absolute number as a descriptive social norm does not effectively persuade clients to seek contact when facing payment difficulties.Previous research has shown that customers can be ashamed of their financial situation, which means that the threshold to contact for financial help might be higher than that for paying an outstanding bill and complying with a payment request (Keizer, 2016;Walker, 2013).
Social proof is thought to exert the greatest influence under two conditions: uncertainty and similarity (Cialdini, 2009).People are more likely to follow those who look like or are like them (Burn, 1991;Schultz, 1999).As people are ashamed of their debts (Keizer, 2016;Walker, 2013), it may well be that they do not want to identify with other people who also experience financial problems because this negatively affects their self-esteem.This would be in line with the self-categorization theory (Turner & Reynolds, 1987) and the finding Wenzel (2004) demonstrated in his study on tax compliance that instead of conforming and behaving to the norm, people reject it and disidentify themselves from the group holding this norm.This could have happened in our study where people started to exhibit the opposite behavior of not seeking contact when needing help with their financial problems.In a follow-up study, it would be interesting to determine the extent customers with arrears identify with other customers with payment problems.
For future research, an interesting investigation would be to clarify whether and to what extent the different barriers such as the ostrich effect (Karlsson et al., 2009) and optimism bias (Sharot, 2011), and knowledge of solutions (Tiemeijer, 2016) play roles in customers' hesitation to ask for financial help when they actually need it.Perhaps denying and assuming that everything will be resolved in the future are easier options than facing someone's own financial problems.
The results of the two attitude questions showed that customers were positively surprised about the help that a mortgage lender can and wants to provide when someone needs financial help.It might be interesting for future research to use this customer feedback as a testimonial on the website of the mortgage lender, so that it serves as a form of social proof, to convince other customers to contact mortgage lenders for financial help.

Overall learnings for applied research
Working together with partners from practice is imperative to test our theories and assumptions among relevant populations in the field.Although this type of research is very valuable it does present with certain challenges that might hamper the internal validity of the research.Below we would like to present the main challenges we encountered in this research and provide tips to avoid certain pitfalls in future applied studies.

Manipulations and manipulation checks
Key to performing a good experiment is an effective manipulation of the independent variable.In this case the descriptive social norm of contacting the mortgage lender in case of anticipated payment problems.Whereas in a laboratory experiment a researcher can use strong manipulations of the independent variable and can make use of deception to reinforce a manipulation, this in most cases would be considered unethical in actual practice (e.g., 100,000 vs. 1,000 customers).The main reason for this is that you cannot provide a good debriefing in a field experiment.In our research, we also had to adhere to clear guidelines from the mortgage lender, including mentioning only the actual number of people who had contacted the mortgage lender for financial help.This makes sense because financial service providers naturally want to be honest and transparent (not using deception) and do not want the different conditions in to clearly benefit only part of the customers (reducing the strength of the manipulations).To ensure that manipulations indeed have the intended impact, a solution could be to first conduct a laboratory experiment.Where different descriptive social norm messages could more easily be used to investigate the effect of these.When it is clear which type of message works best, this intervention can then be replicated in a field study.
Conducting a pretest can also help to identify if customers perceiving the absolute number of 8,357 as low.We were given the opportunity to conduct this research among customers of the mortgage lender, but it had to be completed within a certain period; thus, we could not perform a thorough pretest.In hindsight of the results of our study, we recommend emphasizing the importance of a pretest to a partner from practice in the future.To then create space in the planning to first carry out a pretest.
To counteract the observer bias, it was decided that the call center agents should be blinded to the experimental conditions customer were in to prevent biased effects.As a result, we did not include a question to check whether the customers perceived the absolute number of 8,357 customers as many or a few.In addition to performing a pretest, conducting a manipulation check can provide insight into whether the manipulation of the independent variable was successful.
The increase internal validity of field experimental studies we recommend to both conduct a pretest and include a manipulation check.

Sample size and characteristics
The benefit of conducting a field study in collaboration with a partner from practice is that is easier to obtain a sample that is relevant to the problem that is studied and that it is possible to measure actual behavior.In many laboratory experiments, researchers can only measure behavioral intentions rather than actual behavior.Particularly given the current subject of help seeking in case of financial problems, simulating reality in a laboratory experiment would be very difficult and field research has great added value.But there are also certain challenges that researchers need to consider regarding the sample size and characteristics.
One problem in the current study was that due to customer privacy, no information could be obtained about background variables such as age, gender, income, and the mortgage amount.As a result, we could not control for these variables in our analyses, and we could not determine whether there were preexisting differences between conditions on several important variables.This is a threat to the internal validity.To obtain more background data clients can be anonymized by the company one collaborated with.Before conducting a study like this a data sharing agreement can be drawn up in which it is delineated which data will be shared and which not.It is good to think about this well in advance to prevent obtaining privacy sensitive information and ensure that you obtain relevant additional information about the participants (e.g., age, gender, mortgage amount).
While we were able to conduct an experiment in a relevant target group, a limitation of this study is the relatively small sample size.Due to the limited number of customers that met the inclusion criteria within the company, it is difficult to capture a somewhat smaller effect size typically observed in social psychology (Schäfer & Schwarz, 2019).By first testing several variants via a pretest among participants who score high on financial stress (Mani et al., 2013), it can be determined which intervention has the most effect on the target group.Subsequently, the best working intervention can be tested in a field experiment.
After 14 days the call center agents were instructed to call participants in the current study.The customers who indicated that they were willing to cooperate were asked several questions about the letter they had received.However, 51% could be contacted.There is therefore a chance that the obtained data might be biased in that only a specific group of customers were spoken to.The call center agents who participated in our research only worked on weekdays during the day.As a result, there is a chance that they were not be able to contact people who are also working at that time.In a future study it is our recommendation to have the followup action take place at different times, e.g., in the weekend and/or in the evening to make sure that you get in touch with as many customers as possible.This way you increase the chance of getting in touch with customers who cannot answer their phone on working days.

Data collection
If data collection is not conducted by the researcher, but by the partner from practice, as was the case in the current study, it is important to make good agreements in advance and to carry out quality checks during the research.In the current study call center agents were briefed on how to log all answers to pre-formulated questions in a central Excel spreadsheet.During the analysis, however, it became clear that especially the open text fields for the exploratory questions were not filled out consistently and completely by all call center agents, making further in-depth analysis difficult.An example of this is the question asked about how customers had experienced the letter.Some call center agents described the customer's response extensively while other call center agents only wrote down the words positive or negative and did not explain this further.As a result, it is not clear why customers were positive or negative and it is difficult to determine what could be changed.To prevent inconstancies in the dataset, it is recommended to check for completeness consistency after a few days and if necessary provide additional instructions in a timely manner.
In the current study only 2% of the customers contacted the mortgage lender themselves.The other customers were contacted after 2 weeks.Some customers indicated that they did not currently need help but would certainly seek help in the future because of the letter they received.In a follow-up study, we recommend that all customers who received a letter be followed up for a longer period.So that it can be investigated whether the manipulation has the intended effect in the longer term, because asking for financial help is not a daily activity.It is recommended to think in advance about the behavior that is being studied and how often this occurs when determining the moment of a postmeasurement.

Conclusion
The results of our research contribute to the already rich literature on social influence.In this study, an absolute number was used as a descriptive social norm message.This method has been rarely used and certainly not in the field of financial decision-making.In addition, there is increasing focus on the prevention of debt (AFM, 2013;FCA, 2016).As this research was conducted in the field among real customers of a mortgage lender, it provides valuable insights into the difficulty of establishing contact with customers and helping them prevent financial derailment.Important lessons have been learned from this collaboration between science and practice that may aid those with similar study designs, to avoid certain pitfalls.

Table 1 .
Percentage of customers who think that others will seek contact for financial help in the future.