Towards measuring social norms

(1)

Towards measuring social norms

Thesis MSc. Economics

Juliette de Wit

Supervised by:

prof. dr. B.P. de Bruin

dr. C. Lisciandra

Faculty of Economics and Business

University of Groningen

(2)

(3)

1

Abstract

The inclusion of social concepts in economic frameworks has become more widely ac-cepted over time. The importance of concepts such as trust, social capital and culture for economic theory has been theoretically and empirically corroborated. This thesis argues that social norms should be added to this list. While social norms have been assessed in an experimental setting, the introduction into other empirical economic models is less accepted. Based on a review of the latter literature, it is found that methodological errors are made when assessing social norms. This thesis aims to map these errors, by provid-ing an abstracted account of an empirical research design. Usprovid-ing this map, a thorough approach towards social norms is developed. Insights from Bicchieri (2005) are used to ground this process. Moreover, the categorized errors function as a guide as to what should not be done. The result is a preliminary attempt to develop an index of social norms that can be used in empirical (economic) research. This index consists of three social norms: honesty, gender equality and work ethic. Whereas these three dimensions are found for the Netherlands, only two of them are extracted for China.

(4)

2 CONTENTS

List of Figures

1 Views on migration across the world . . . 10

2 Overview of research steps . . . 22

3 Organization of section 4 and 5 . . . 23

4 Poverty trends according to international standards . . . 30

5 Graphical representation of dimensions . . . 51

List of Tables

1 Overview of refugee flows . . . 13

2 Descriptive statistics for China and the Netherlands . . . 57

3 Eigenvalues and total variance explained for the Netherlands and China . . . 59

4 Factor analysis for China and the Netherlands. Wave 4 . . . 60

5 Correlation matrix for China . . . 61

6 Correlation matrix for the Netherlands . . . 62

7 Correlation matrices across time . . . 64

8 Features and errors of each step of the research process. . . 69

9 Overview of the selected items from WVS and EVS . . . 70

10 Availability of data across countries and across waves . . . 78

11 Factor analysis across time for the Netherlands . . . 82

(6)

4

1 Introduction

The homo economicus - a rational, narrowly self-interested agent - has had a long-standing tradition within economic science. Alongside this discourse, a broader perspective on human beings has been developed. Though the idea that values, beliefs, customs, norms, and traditions matter in explaining behaviour is not new, it can be understood as a departure from the neo-classical framework. As Tabellini (2008) notes, it is unfortunate that norms have not been included in the paradigm of economics. These factors can induce behaviour that goes against narrow self-interest, for example by forgoing an individual benefit, while the homo economicus is expected to always act in his or her own interest. Efforts have been made to adapt the traditional models so that norms and conventions, amongst others, can become part of the rational decision framework. Staying within this framework, experiments have been used to show that individuals do not always act on their own behalf. Concepts such as fairness, trust, and cooperation have been tested in Ultimatum, Social Dilemma and, Trust games (Bicchieri, 2005). These games have been adopted to assess the nature of social norms, how they change across time and how they are affected by the environment.

A social norm can be defined as: ‘behaviour that is collectively approved or disapproved in a group or population and is enforced by sanctions.’ (Bicchieri, 2014, p. 209). Social norms direct human action, and can result in various types of behaviour. Norms can refer to fairness, equality, or generosity, but also to more specific behaviour, such as a norm that states that you cannot tell lies. They come into existence because a large number of people believe that a specific norm exists under certain conditions, and that other people will conform to the norm under these conditions (Bicchieri, 2005). The willingness to obey a norm is conditional upon the belief that other people will conform. Our interactions are thus shaped and guided by social norms, functioning as rules that guide our behaviour. In guiding our behaviour, norms perform an important social role, facilitating social order and the coordination of collective actions (Bicchieri, 2014). Social norms are not often observable, rendering the hypothesis of revealed preferences as adopted in economics insufficient. Violation of norms however, can be observed and is therefore compatible with revealed preference theory. As Tabellini (2008) notes, norms do affect rational choice. Deliberations that result in a social norm may be rational, and therefore lend themselves well for inclusion in the economic tradition. Moreover, conformation to norms is often automatic and subconscious (Bicchieri, 2005). Only when they are broken or breached by someone else, we may experience anger, outrage, or confusion (Bicchieri, 2005).

(7)

INTRODUCTION 5

of norms, despite having studied the implications of social norms (Tabellini, 2008; Viscusi, Huber, and Bell, 2011). Taking these issues into account, attention to both the theoretical and empirical approach to social norms is warranted.

Based on the conviction that philosophy and economics can benefit from each other, the key objective of this thesis can be described as follows: to provide a theoretically thorough analysis of conceptual and empirical mistakes which are made in the economic literature that deals with social norms, and to develop a preliminary method which can be used to measure norms. I attempt to make some progress in providing a conceptual and methodological framework to measuring social norms grounded in philosophical theory.1 This attempt will be the stepping stone towards a non-na¨ıve way of measuring social norms, that will enable their inclusion in future empirical (economic) research. Non-na¨ıve refers in this context to an approach which is both conceptually thorough (e.g. philosophically grounded) and statistically (e.g. economically) well executed.2 Possibly, this method can be employed for social concepts other than social norms.

This thesis is part of a larger project that addresses the emergence of social norms over time across diverse groups. It is the response to current events in Europe, where large streams of refugees have been arriving over the past years. Fears regarding the unknown effects on society when a large group of people with different norms, values, and cultural backgrounds enters a country with distinctive norms and values revive. Current literature provides insufficient tools to analyse these dynamics concerning social norms, whilst those insights may be essential in gaining a deeper understanding of current events. A combination of the literature on social norms and migration economics may yield interesting conclusions, as well as policy recommendations. In the execution of this idea, it is fundamental to gain a deeper understanding of social norms in general, and within economics in particular. This issue is addressed in this thesis.

In doing so, this thesis follows a tradition in social sciences to measure social concepts.3 Within social sciences, the measurement and operationalisation of socio-economic variables is a widely discussed topic. In trying to capture concepts such as trust, social capital, well-being, and human development, (Lochner et al., 1999; Glaeser, Laibson, Scheinkman, and Soutter, 2000; Lillbacka, 2006; Diener and Suh, 1997) a diverse set of methods has been de-veloped to generate reliable and highly accurate indicators. Criteria such as methodological soundness, intuitive validity, consistency, reliability of data sources and, internal reliability guide the choice and development of indicators (Kovacevic, 2011). Nevertheless, these nor-mative principles do not offer unequivocal advice, leading several authors to observe that

1_{A large body of literature on social norms can be found in social sciences. As the primary focus of this}

thesis is on in the philosophical nature of social norms, this literature is not covered explicitly. Implicitly however, insights gained by Bicchieri (2005) also draw upon the social science literature.

2_{A critical assessment of current literature is needed to start a discussion on social norms. That is to say, a}

discussion which addresses the conceptualization and measurement of social norms within empirical economic papers. Such a discussion can function as a starting point towards the development of a thorough and widely accepted methodology.

3_{In 1897, Emile Durkheim showed that an inverse relationship between social integration and suicide}

(8)

6

no consensus about measurement of social concepts exists (Bjørnskov, 2006; Lochner et al., 1999). Moreover, it seems that in many cases the lack of consensus on the methodologi-cal front stems from problems regarding the definition of the concept of interest (Lillbacka, 2006). The development of a more widely accepted methodology for socio-economic concepts is warranted, especially since its subject matter is of multifaceted nature. By using surveys, this thesis aims to develop a measure of social norms as a preliminary foundation for a wider methodology (Rauhut and Winter, 2010).

Institutional economics has conceptualized and measured norms ever since the 1990s with Douglas North’s seminal 1991 paper Institutions. In his paper, North proposes a dis-tinction between formal and informal institutions. The latter includes norms, values, and customs (1991). Hence, any analysis of institutions, defined as “rules of the game” or “hu-manly devised constraints”, incorporates a discussion of norms (North, 1991; Beugelsdijk and Maseland, 2010). Informal rules of the game are however difficult to measure. This may explain why (social) norms have been ignored in empirical studies, whereas other social constructs such as culture, have been included. Questions surrounding institutional quality and economic growth, and more broadly the relation between institutions and culture, have been widely assessed. The type of empirical research that is conducted in this field relates to the meta-question posed by this thesis: how to conceptualize and measure social constructs? Trust, social capital, values, culture have all been reviewed in other publications (for exam-ple Beugelsdijk and Maseland (2010) or Bjørnskov (2006)). This thesis adds social norms to that list.

Finally, this thesis is an attempt to combine insights from diverse fields such as sociology, philosophy, and economics. While all three fields have dealt with social norms in their own way, not much effort has been put into combining their insights into an overarching frame. Rauhut and Winter (2010) note that the methods for measuring social norms in economics and sociology have taken separate paths. Whereas experiments dominate the field of eco-nomics, sociology draws mainly upon surveys. Both approaches have their advantages and disadvantages; as such it is surprising that little effort has been put in combining the obtained insights. According to Rauhut and Winter (2010), different conceptions and research inter-ests in social norms underlie the discrepancy between economic and sociological approaches. To overcome this discrepancy at a conceptual level, insights from philosophy may prove to be useful. A solid theoretical and conceptual discussion of social norms and their evolutionary development can be found there (Bicchieri, 2005). The origin of social norms and how they evolve over time has been thoroughly explored; Cole, Mailath, and Postlewaite (1992) note that the latter feature misses in economic models that include norms.

The key objective of this thesis - the analysis of current approaches to social norms in economics, as well as the development of a non-na¨ıve account of social norms - is addressed in section 4 and 5 of this thesis. Sections 2 and 3 pave the way for these discussions.

(9)

INTRODUCTION 7

of the larger project is, in my opinion, essential to understand the critical contribution of this thesis. It derives its value from its position in a larger whole. In addition to the overview of the Ph.D. projects, a small review is given of the current literature on migration and refugees. Key to note here is that social norms are not incorporated in this literature.

Section 3 That efforts have been made to incorporate social norms in economic theory is shown in section 3. A selection from diverse applications has been assembled to give an overview of the non-experimental literature. Firstly, the incorporation of social norms into migration theory is addressed. Secondly, the application to other fields is described. It is by no means my intention to give an all-encompassing overview of this literature, but merely an idea of what has been done, what is possible, and what is still missing.

Section 4 It will become clear that current non-experimental economic literature misses a conceptual and methodological approach towards social norms. Based on the overview that is given in section 3, section 4 expands upon the errors that are present in current approaches towards social norms. This is done in a structured manner, following different stages of the research process: definition, operationalisation, measurement, estimation and interpretation. Using insights from Cartwright and Runhardt (2014), who analyse the research process from a philosophy of science point of view, features of each stage of this process are identified. Examples from the literature are used to indicate that errors are currently made. Since this thesis has different audiences, for some readers this discussion may be self-evident. It is however included to facilitate reading of readers with other backgrounds.

Section 5 The categorization of errors enables me to identify the challenges that are to be faced when developing a measure of social norms. In working towards a non-na¨ıve measure of social norms, a preliminary attempt is made to face these challenges. This attempt is preliminary for several reasons: 1) a measure that includes three social norms instead of a wide variety of norms is developed; 2) the analysis only concerns two countries - China and the Netherlands; 3) several limitations with respect to the econometric approach are not taken into account, but are merely indicated. Preliminary results are obtained by first developing a conceptual view on social norms, based on insights from Bicchieri (2005). She has developed a theory of social norms, which is well-suited to economics. The operational definition that Bicchieri develops is adopted to define social norms of gender equality, honesty, and work ethic. Hereafter, I operationalise these three social norms, and describe how I attempt to measure them using factor analysis. Measurement is executed using data from two different surveys, the World Values Survey (WVS) and the European Values Study (EVS). Finally, the results are presented and interpreted, after which I indicate the limitations of this empirical study.

(10)

8

(11)

SKETCHING THE LARGER PICTURE 9

2 Sketching the larger picture

For millennia, violent conflicts across the world have caused large movements of people, looking for a safer place to stay. Our times are also characterized by such movements, where, for example, the conflicts in Rwanda and Burundi in the 90s, the continuing conflict in Afghanistan, and, in recent years, the conflict in Syria have all led to displacements of people living in these regions. It should be noted that, alongside the often pressing political and humanitarian aspects of violent conflicts for nearby regions, regions at a further distance have become more and more involved over the course of time. Despite the recurrence and severity of such conflicts, it seems that our knowledge is insufficient when dealing with large refugee movements, or more generally with migration and its impact on receiving countries. Unequivocal policy advice based on theory are difficult to give, despite the great need for them. The current so-called ‘refugee crisis’ can function as an example of this.4

Focusing on the example of Syria, the displacement of millions of Syrian citizens does not only affect the destination countries in the neighbourhood of Syria, but also countries further away. Consequently, in addition to the destabilizing effects of the civil war on the region of Syria, tensions and fears have arisen in Western European countries upon the arrival of large streams of refugees. These developments can be observed in the news, in discussions held in the chambers of parliament, and in everyday conversations, where topics of migration, integration, and refugees occur on a daily basis. Sentiments of unease and fear of dealing with current events can be found when reading between the lines.6 In particular, it is said that our norms and values are at stake. Norms of equality, for example, may be perceived differently depending on one’s cultural and religious background, and will potentially change for both natives and migrants in the course of integration. This change can be accompanied by clashes, between various groups and on different grounds. In the Netherlands for example, where it is estimated that 90.000 people will seek refuge in 2016, problems in the refugee camps are identified with respect to the treatment of homosexual refugees (Pelgrim, Versteegh, Kuiper, and Brummelman, 2016). While the Netherlands is often perceived as a country in which everyone enjoys the same rights, not all homosexual refugees experience these rights in the refugee camps where they are temporarily living. Cases of Syrian refugees being bullied and threatened by other refugees have featured in Dutch media outlets in the past months. This implies that being in the Netherlands does not lead to immediate acceptance of the Dutch laws and norms by all new inhabitants, despite the fact one now lives in an environment where these are perceived as “the norm”.

Another example from the Netherlands can be introduced in order to illustrate the op-position among native citizens as a response to the refugee crisis. In 2015, once it became apparent that many Syrian refugees were fleeing to Europe, a discussion in the Netherlands

4

The term ‘refugee crisis’ is more appropriate than migration crisis, a term also coined by the media to refer to the influx of refugees in the past months. The UNHCR5 _{advises against the use of the latter term,}

as mainly refugees “seeking protection from war and persecution” arrive in Europe.

6_{The psychological state of refugees after leaving their home country, their home, friends, and family in}

(12)

10 The Ph.D. project

started on where refugee camps could be located and what type of shelter should be provided for. While the government proposed to place small refugee camps across the country and to give sober “yet humane care” in the words of one of the ruling parties (NOS, 2016), it quickly turned out that support for this policy was lacking at lower governmental levels. The discus-sion escalated after citizens of several municipalities revolted against the policy, and refused to accept the issuance from the national government. Not only the Netherlands faces such conflicts, but other countries also experience difficulties in finding ways to shelter arriving refugees. This tendency is illustrated by the Gallup World Poll on migration, which asked two questions: 1) In your view, should immigration in this country be kept at its present level, increased, or decreased? and 2) Do you think immigrants mostly take jobs that citizens in this country do not want (e.g. low-paying or not prestigious jobs) or mostly take jobs that citizens in this country want? Both questions reflect attitudes of native citizens towards immigrants. Analysing the data, it is striking that, in particular, many South-European citizens have developed a more negative view of migrants over the past years. Figure 1 illustrates the change in answers to the first question in the period 2012-2014 across the world.

Figure 1: Views on migration across the world

2.1 The Ph.D. project

(13)

(Jakiela, 2011), it is expected that the interaction between migrants and natives influences social norms. The main aim of this Ph.D. project is to adopt multidisciplinary methods from economics and philosophy to (i) develop new theories of social norm dynamics that can deal with subtle norms underlying fairness, equality and justice, and that can be applied to the context of migration, and to (ii) draw policy recommendations from the findings of (i).

Several open questions arise from these objectives: 1) What happens to norms when two groups with different norms start interacting, as might happen in the case of migration? 2) How do more complex types of norms, such as the treatment of man vs. woman, freedom of speech, and treatment on the labour market, change in the light of events described under 1)? and 3) What can be done in terms of policy making? Current developments in Dutch refugee camps make answers to these questions both pressing and relevant. Currently, the idea is to tackle these questions in four separate projects, with this thesis as a building block. A clear understanding of social norms is vital before analyses of changes due to migration can be made, and policy recommendations can be given. This thesis will be part of the first project. The three subsequent projects will use the insights gained in the first. They are structured in the following manner:

Project 1 – Existing conceptual models from evolutionary game theory and migration economics will be employed to design a theoretical model which can answer questions (1) and (2). A thorough theoretical analysis of a variety of norms linked to equality, justice, and fairness stemming from ethics and philosophy in general (Ensminger and Henrich, 2014), will bring about a more complex and detailed view on current evolutionary game theoretical models. Extending the application of these models to interactions between multiple groups over time will allow for their application in the domain of migration. Techniques that are accepted within ethics and game theory will be adopted to develop a theoretical model.

Project 2 – The theoretical model developed in the first project will be operationalised such that a formal model arises that can mimic dynamic group interactions. This formal approach will yield a model, more specifically an agent based model, that will be tested using computer simulations. This setting allows for the analysis of different scenario’s. Group sizes, people’s beliefs, and sets of norms for example, can be manipulated. In this way, the model can account for (migration) shocks and their impact on the evolution of complex norms within and between groups. Agent based models have become more embedded in the economics literature, where they are used in the domain of marketing and analyses of economic systems. They are also employed to assess norm dynamics, for example to analyse the interplay between networks and the emergence of norms (Froncek, 2015), or integration and segregation in cities (Hatna and Benenson, 2015). This project will build upon approaches taken in this literature. Moreover, it will make use of recent insights gained by Alexander (2015), for example, whose research focuses on the development and formal analysis of more complex types of norms.

(14)

12 Economics of forced migration

of the current refugee crisis from an empirical point of view. Data on migration flows will be employed in an empirical analysis. The use of a panel dataset to gain insights into the dynamics and impact of migration streams on economies is well-established in the literature, and these methods will be employed to yield conclusions regarding social norms and migration. Evidence exists that norms may be guided by policy (Viscusi et al., 2011). Based on this evidence in combination with insights obtained from project 1 and 2 it is safe to assume that the empirical results will allow us to give meaningful policy recommendations. Project 4 – Since society at large is influenced by migration, its businesses will equally experience an impact of migration on norms. Based on the evolutionary view on norms, social norms ought to change when migrants enter a specific sector or company. Therefore, this project will adopt methods employed in business ethics to assess how people deal with norm changes following migration in a business context (Bruin, 2013; Wasieleski and Hayibor, 2009). Hypotheses derived from the research conducted in projects 1 and 2 will be tested, so that relationships between different types of norms and characteristics of companies can be assessed. Data obtained through surveys in specific sectors will enable a thorough analysis of this question.

The Ph.D. project will fill several gaps in the literature and make a contribution to philo-sophical literature on social norms, as well as economics literature in a variety of areas. More specifically, the contributions to the latter field will mainly concern migration economics, in-stitutional economics, behavioural economics, and game theory7. This Ph.D. project will contribute to the literature by providing a conceptually thorough framework of social norms that can be adopted in several fields of economics on the one hand, and by translating this conceptually thorough framework to an empirical methodology, applicable to philosophical and economic research, on the other hand.

In order to gain some deeper insights into the background of this Ph.D. project, a general overview of the literature on migration economics is given in the following section. The goal of this section is to indicate the benefit of incorporating social norms into these frameworks.

2.2 Economics of forced migration

Though the terms migrant and refugee are often used interchangeably in public discourse, there is a theoretical and practical distinction between the two. Defining both terms is essential for any further discussion related to either one. According to the UNHCR (2016), migrants decide to move with the goal of improving their lives in terms of work, education, family reunions, or other reasons. If they would like to, they could return home at any time of their choosing. Refugees on the other hand are “persons fleeing armed conflict or persecution” (UNHCR, 2016), looking for safety in nearby regions. International law protects refugees, as is stated in the 1951 Refugee Convention. According to the UNHCR, this distinction is of great importance for individual governments, as migrants are dealt with

7

(15)

according to a country’s own immigration laws and processes, while refugees fall under the international norms and laws of refugee protection (UNHCR, 2016). In the literature, a similar distinction is adhered to. The latter group, that is the group of refugees, is often referred to as “forced migrants”, where forced migration can be defined as “the movements of refugees and internally displaced people (those displaced by conflicts) as well as people displaced by natural or environmental disasters, chemical or nuclear disasters, famine or development projects” (Ruiz and Vargas-Silva, 2013, p.773)). ‘Forced’ in this sense refers to the nature of displacement, which is involuntary in the case of refugees and voluntary in the case of migration. I use both terms, forced migrants and refugees, interchangeably.

Following the outbreak of the civil war in Syria in 2011, a large stream of refugees has been set into motion. Driven out of their homes by violence and insecurity, many Syrians have decided to move elsewhere: to countries bordering Syria or to countries further away. As indicated in the introduction, the refugee crisis is not only a highly debated topic in the daily news, but also within politics and academia. In the latter, an increasing amount of attention is given to the impact of Syrian refugees on receiving countries’ economies. From the start of the refugee crisis, analyses of various aspects of, for example, the Turkish labour market have become more abundant (Ceritoglu et al., 2015; Akg¨und¨uz et al., 2015; Carpio and Wagner, 2015; Bahcekapili and Cetin, 2015). Scrutinising past decades however, conflicts comparable to the civil war in Syria can be distinguished. As table 1 illustrates, millions of people have been displaced over the past decades as a consequence of diverse conflicts. These conflicts have been studied in the literature as well, culminating in insights with respect to the impact of streams of forced migration on both destination and home countries. While this literature is especially interesting when conducting empirical research, or when focusing on the interplay between (forced) migration and social norms, the background theory is of value for this thesis, as it sketches the context for the discussion of social norms and its conceptualization. Therefore, the following paragraphs will offer an overview of the conclusions reached in this strand of literature.

Table 1: Overview of refugee flows

Years Refugee Flows Number of Refugees

1945-1950 From East Europe to West Ger-many after WWII

12 million

1940s Migration from East Finland to West Finland during WWII

430 thousand

1993-1994 From Burundi to Tanzania 300 thousand

1994 From Rwanda to Tanzania 250 thousand

1992-1995 From Bosnia-Herzegovina to neigh-bouring countries

1.3 million

2011-present From Syria to neighbouring coun-tries (Turkey, Iraq, Lebanon, Jor-dan)

3.9 million

(16)

14 Economics of forced migration

The renewed academic interest in forced migration stems not only from current events, but also from the fact that the nature of civil conflicts lends itself well to empirical research. Whereas studies of voluntary migration often encounter difficulties in disentangling the mo-tives of migrants to move to a certain destination, for example the separation of economic from other incentives, these difficulties are not an issue when it comes to forced migration. In these so-called natural experiments, violence can be identified as the main driver of mi-gration. The host country’s economic development plays only a minor role when choosing a location. As such, the estimation of the (causal) impact of forced migration on the dependent variable of choice is more precise.8

The literature that analyses forced migration has two main fields of research: impact of forced migration on migrants themselves and the impact on a host country’s market. While the former often finds positive effects in the long run for displaced groups in European countries, evidence for developing countries is less positive. There, scholars even find negative effects in the short and medium term (Ruiz and Vargas-Silva, 2013). Mixed evidence is also found in the analyses focusing on the host country’s market, where both losers and winners are identified: Agricultural producers seem to be the winners, taking advantage of the cheaper labour force consisting of refugees and the increased demand for agricultural products (Alix-Garcia, Bartlett, and Saah, 2013). Children of native citizens, on the other hand, are identified as losers, facing long-term health consequences (Baez, 2011).

The civil conflicts in Rwanda, Burundi, and Syria are the main motive for researchers to analyse the impact of forced migration on a on a host country’s market. The nature of these events, together with the availability of a wide variety of data, make them attractive cases to analyse. The impact of Syrian refugees on the Turkish labour market, for example, is a pop-ular topic (Ceritoglu et al., 2015; Akg¨und¨uz et al., 2015). It seems that mainly uneducated, low-skilled workers are negatively affected by the inflow of Syrian refugees (Ceritoglu et al., 2015; Bahcekapili and Cetin, 2015). As labour qualifications affect the composition of the productive capacities of a host country’s economy, Bahcekapili and Cetin (2015) assess the impact of refugees on the local workforce.9 The authors discover that especially the unskilled workforce in regions where refugees have moved to are deprived of their jobs. Moreover, the growth rate of the Turkish per capita income seems to be negatively affected by the refugee shock. Interestingly, the foreign trade balance increased in regions where Syrian refugees took shelter.

The results of Ceritoglu et al. (2015) corroborate the findings of Bahcekapili and Cetin (2015). Employing a difference-in-difference strategy to assess the impact of forced migration on the Turkish labour market, they also find that the informal Turkish workers are most affected. They are more likely to lose their jobs, as unemployment has risen for this group

8

While the properties of natural experiments make them attractive in terms of econometric analysis, several difficulties with respect to data availability, selection biases, and the assessment of the causal effect still exist due to problems encountered when attempting to separate the impact of war from other effects. Therefore, a variety of (econometric) techniques are adopted in this line of research in order to reach an unbiased estimate of the causal impact of forced migration on a dependent variable. Difference-in-difference estimation, propensity score matching and instrumental variables are often-used techniques.

9

(17)

as a consequence of the shock. Wages are, however, not affected significantly. An identifying assumption similar to Bahcekapili and Cetin (2015) is used, as the regions in which refugee camps are located are used as the treatment group. Refugees are clustered in these regions, as such a quasi-experimental design arises. The paper of Akg¨und¨uz et al. (2015) employs a similar identification strategy, assessing alongside employment rates, internal migration patterns, and the dynamics of food and housing prices in Turkey. No impact of the entering Syrian refugees is found on natives’ employment rates, while possibly the increase in inflation – especially on the housing market and to a lesser degree on food prices – does affect the real wage of natives in a negative manner. Moreover, a significant decline in internal migration rates to regions where refugees are located is observed.

Comparable studies have been executed for the case of Tanzania, where in the 1990s a large group of refugees from Rwanda and Burundi arrived. The existence of a large household panel for the region of Kagera, which includes six waves over the period 1991-2010, has enabled detailed analyses of these events. Similar to the studies of the Turkish labour market, aspects such as prices, employment, and wealth are examined. Alix-Garcia and Saah (2010) find in their study that the presence of refugee camps has increased food prices in nearby regions. Depending on the origin of the majority of refugees, either Rwanda or Burundi, the prices of different baskets of products increased. The difference in diet is proposed as an explanation of this result. The impact on natives’ wealth, employment, and consumption patterns is also assessed. Consumption and wealth of the native population was positively affected by the movement of refugees to the Kagera region in the short and long term (Maystadt and Verwimp, 2014; Maystadt and Duranton, 2015). Business in non-agricultural sectors boomed (Maystadt and Verwimp, 2014; Ruiz and Vargas-Silva, 2013). Not all population groups gained from the forced migration shock however, as Baez (2011) finds that especially children are negatively affected. Their health status and human capital accumulation is deprived in the long run.

(18)

16

3 How economics approaches social norms

While social norms are variously addressed across diverse fields, their incorporation in a specific part of economics could use more attention. In the following, I will focus on the research that addresses social norms in (macro-economic) empirical frameworks. There, social norms are characterised in an ad hoc manner; a wide variety of methods is encountered. The experimental literature, as well as theoretical literature that deals with social norms, is excluded from this discussion. This section comprises two parts: firstly, following the line of thought set out in the previous section, an overview of the literature in migration economics that incorporates social norms is presented. Afterwards, a similar examination with examples from various fields in economics sets the first overview in a wider context. Both overviews were compiled using social norm(s) as keyword, retaining only those articles of an empirical nature. Hence, articles covering a wide variety of topics are included. Section 4 builds upon this discussion by indicating errors that are made with respect to conceptualization, operationalisation and measurement of social norms. Illustrations from the literature that is discussed in this section will be used.

3.1 Migration

In the search for studies that assess the interplay between social norms and refugees, the literature on forced migration is rather disappointing. As far as I have encountered, social norms have not been included in these models so far. Looking at migration in a broader sense, however, social norms, or norms in general, have indeed been studied.10 Social norms such as generosity and altruism have been incorporated in models of migration-decision making (Helliwell, Wang, and Xu, 2014; Brown, Gareth, and Prayaga, 2014), but also in models that assess the impact of migration on social norms of migrants (De Jong, 2000; Ryo, 2013). The interaction between migrants and natives and its effect on migrants’ social norms is modelled and empirically tested. Other norms, such as fertility and institutional norms (Beine and Sekkat, 2013; Beine, Docquier, and Schiff, 2013), have been included in models assessing the impact of emigration on home country norms - a so-called feedback effect. This second group of papers focuses on the feedback effect of migration on the home country, whereas the other papers study the impact of migration on migrants themselves.

The motivation for migration is addressed in behavioural studies, which attempt to ex-plain move-stay decisions (De Jong, 2000). While economic motivations are often introduced as explanatory factors, attitudes, expectations, norms, and beliefs are increasingly investi-gated. De Jong (2000) hypothesises that an individual’s migration behaviour - to migrate or not - is mediated through subjective expectations about the possible outcomes. Moreover, the intention to migrate determines the decision to migrate.11 In their turn, social norms and expectations influence this intention and therefore the migration decision. These

be-10_{Ass defined in the previous section: migrants decide to move with the goal of improving their lives in}

terms of work, education, family reunions or other reasons. If they would like to, they could return home at a time of their choosing (UNHCR, 2016).

11

(19)

HOW ECONOMICS APPROACHES SOCIAL NORMS 17

havioural studies arguably use a more general meaning of social norms as “critical elements in translating expectations into behavioural intentions and subsequent action” (De Jong, 2000, p.307) or “moral standards attributed to a social group or collective” (Ryo, 2013, p.5). It is, however, not social norms in general that are included in the decision-making models, but family norms (De Jong, 2000), descriptive norms and injunctive norms (Ryo, 2013). The first and last types of norms refer to the encouragement from family members to migrate, whereas the descriptive norms refer to the migration behaviour of family members. Ryo (2013) provides clear descriptions of the concepts she employs. The concepts refer to two dimensions of social norms. The first, descriptive norm, reflects the individual’s perceptions of how his or her reference group behaves. The second, injunctive component, describes how one ‘ought’ to behave. Such a description of what social norms are, is missing in the paper of De Jong (2000). The definition of a family norm can only be inferred within the context of the estimation framework. There, norms are measured by the perception of the individual of his or her family encouraging him or her to stay or move. The normative component, as identified by Ryo (2013), thus misses from De Jong (2000)’s framework. Even though these approaches differ substantially, they both conclude that social norms increase the likelihood of migration.

Other aspects of migration are examined by Helliwell et al. (2014). They do not analyse the intention or decision to migrate, but the impact migration has on social norms of the individual. They conjecture that the migrant is affected by his or her environment, and will adapt over the course of time. Estimation of this relationship, using a dataset consisting of 132 countries, shows that migrants indeed adapt to their new surroundings while also adhering to norms from their home country. Without clearly defining social norms, the authors mention social trust and generosity as the two norms that are examined in their paper. Both constructs are measured using select items from the Gallup World Poll. Social trust is approximated by the question “Generally speaking, would you say that most people can be trusted or that you have to be careful in dealing with people?”(World Values Survey, 2015). This is a question that is well embedded in the literature on trust, in spite of it being criticized for its validity (Glaeser et al., 2000). Generosity is examined using the estimated frequency of donations to charity over the last 30 days. This may be seen as a rather narrow definition of generosity, as it excludes generous behaviour towards family, friends and other acquaintances.

A growing literature in economics examines the transfer of ‘norms’ from the host country to the home country. A feedback effect may exist caused by remittances, return migration, or the development of trade networks (Beine and Sekkat, 2013). Studies have proven that international migration diffuses democracy (P´erez-Armend´ariz and Crow, 2010), fertility (Beine, Docquier, et al., 2013), institutional quality (Beine and Sekkat, 2013), and civil liberties (Docquier and Rapoport, 2003) to the country of origin. A transfer of values, customs, and behaviour thus exist.12 This transfer, however, differs across socio-economic

12_{I intentionally use ‘values, customs and behaviour’ to refer to the type of feedback, as in my opinion it is}

(20)

18 Miscellaneous

groups. For example, the transfer of institutional norms differs between skilled and unskilled workers (Beine and Sekkat, 2013). Some papers fail to clearly describe how the transfer of norms operates, disregarding the reasons behind the differences in transfers. In the paper of Beine, Docquier, et al. (2013) for example, a theoretical argument for the diffusion of norms is not developed in detail, nor is it incorporated in the econometric specification. Moreover, while some of the articles do analyse the development of norms across time (Beine and Sekkat, 2013), some ignore it completely(Beine, Docquier, et al., 2013). Time may be a crucial aspect in the analysis of an essentially dynamic transfer. Time needs to pass for the transfer to become apparent.

The incorporation of (social) norms in models of migration turns out to expand our understanding of migration decisions, as well as our apprehension of the impact of migration on the home and receiving country. The next section demonstrates that the inclusion of social norms is equally valuable in other economic models.

3.2 Miscellaneous

A concept related to social norms - social capital - has been extensively researched in the social sciences. It is a concept that establishes a theoretical framework in which social norms and social relations can be understood and predicted. Coordination arises as a result of these patterns, such that shared goals can be achieved (Narayan and Cassidy, 2001). Whereas social capital has its roots in social sciences where Robert Putnam has introduced it (Putnam, Leonardi, and Nanetti, 1994), it is nowadays accepted in economics as well. Political economics, as well as institutional economists have embraced it. A conceptual gap is however observed by several authors (such as Bjørnskov (2006)). They mention that consensus with regard to definition and measurement is missing. Several attempts to fill this gap have been made, in which some authors have devoted attention to the role of social norms as constituent factor of social capital (Bjørnskov, 2006; Narayan and Cassidy, 2001). It will be interesting to analyse what role is given to social norms in these theoretical frames; this is done in the following paragraphs.

(21)

can be attained. The accrued benefits will not remain within the organization, but will be spread towards new strangers. Efficiency of society is improved as a result. In addition to the positive effects of social capital on interaction, economic dimensions of interrelations are affected. Knack and Keefer (1997) confirm that trust and civic norms are positively related to economic growth, and Beugelsdijk, Groot, and Schaik (2004) show that trust leads to economic growth.

To gain a deeper understanding of social capital, diverse approaches are taken in the literature. One shared characteristic is that many studies have made use of social survey data (Bjørnskov, 2006; Lochner et al., 1999), which are at times complemented with experimental data (Glaeser et al., 2000; Narayan and Cassidy, 2001). The benefit of latter methods is that experimental results can be used to validate survey items. In this manner Glaeser et al. (2000) find that the question that is generally used to measure trust - “Generally speaking, would you say that most people can be trusted or that you can’t be too careful in dealing with people?”(World Values Survey, 2015) - does not measure trust but trustworthiness. They conclude that behaviour cannot be predicted by attitudinal questions as the “trust question”, and that questions asking for past behaviour do a better job. This finding may have implications for other studies that use survey questions, for that it questions the validity of such measures.

Social norms are theoretically proposed as one dimension of social capital (Putnam et al., 1994). The principle factors of social capital (e.g. the dimensions of the construct) can be extracted using factor analysis or principal component analysis (PCA). Such methods make use of observable variables, such as survey items, which are hypothesized to be related to an underlying unobservable factor, such as the dimensions of social capital. The correlation between the observed variables is used to estimate the underlying factor(s). Each extracted factor represents a dimension of social capital. Based on these methods, we may conclude that trust and networks are constituent factors of social capital (Bjørnskov, 2006; Narayan and Cassidy, 2001). Mixed evidence is found however with respect to the role of social norms. A plausible explanation can be found in the diverging measures - and definitions - of social norms that are adopted in the literature. Estimation results depend on the (type of) measure, and may therefore change if various methods are used.

(22)

20 Miscellaneous

and Cassidy (2001), who analyse social capital in Ghana and Uganda. A different measure of social norms is employed however; norms are measured by the following three survey questions:

(1) Generally speaking, would you say that you can’t be too careful in dealing with people or that most people can be trusted?

(2) Would you say that most of the time people are just looking out for themselves, or are they trying to be helpful?

(3) Do you think that most people would try to take advantage, or would they try to be fair? (Narayan and Cassidy, 2001, p.75)

Question 1 is found to be the most important in explaining social norms. The authors state that social norms can be legitimately measured by this question. While measurement of social norms can thus be validated using Narayan and Cassidy (2001)’s findings, another measure would be proposed based on the conclusion reached by Bjørnskov (2006).

Alongside the literature of social capital, literature in other fields of economics has made attempts to include social norms in their (estimation) frameworks. These attempts illustrate that possibilities exist to include social norms in neo-classical models (Cole et al., 1992). The article of Burda, Hamermesh, and Weil (2007) can function as example: social norms are introduced to explain facts that cannot be explained by traditional labour theory. The authors observe that differences in total work between men and women decrease when GDP per capita rises. Traditional explanations of wage differences cannot explain this observation. Social norms, however, can. They induce individuals to use their time in similar manners. As a result men and women spend similar amounts of time working, providing an explanation to the observation of Burda et al. (2007).

(23)

et al. (2012) examine whether religious norms affect financial reporting. They hypothesize that religious people tend to be more honest and risk averse than non-religious individuals, yielding lower likelihoods of financial restatements and overstatements of revenue. Whereas the other artiles have used direct measures of social norms, either by the employment of factor analysis to extract dimensions of social norms, or by the selection of one or two survey questions, Dyreng et al. (2012) use an indirect measure. They choose to approximate the two social norms by religious adherence in the area where a firm’s headquarters is located.

(24)

22

4 The errors made in economics

Every research process can be characterized by several steps. Figure 2 shows a simple conceptualization of this process, where one moves from the left to the right hand side in conducting research. In reality, of course, research is a more reflective process, going back and forth in order to arrive at a research design.13 The process depicted in the upper part of figure 2 is that of an empirical study. In order to construct a measure, specification and de-abstraction are of the greatest importance. The more one moves to the right-hand side of the figure, the higher the increase of empirical concreteness and amount of detail at the theoretical level will be. The bottom part of the figure depicts a more theoretically based research.14 Here, operationalisation is more theoretical in nature, so that the concept can be used in any theoretical model. Of course, empirical applications can also employ such a theoretical operational definition as a guideline for developing an empirical operational definition (e.g. in the case of Bicchieri (2005)). As this thesis is mainly concerned with an empirical approach to social norms, the bottom half of figure 2 is disregarded in the following paragraphs.

Figure 2: Overview of research steps

A set of features identifies each phase of the research process. They prescribe what a proper execution of that step should incorporate. Analysis of the literature of social norms, as reviewed in section 3 of this thesis, shows that various features are not incorporated there. Errors are made on the conceptual, methodological and econometric plane: Guidelines to measuring social norms are missing, definitions are too general, define another concept than intended, or are not at all included. A main error seems to be that no consensus with respect to the measurement of social norms exists. To get an overview of these errors and of the issues that need attention, it is the aim of this section to provide an abstracted and organized overview of each phase. In order to determine the stylized features of each stage, I make use of insights gained from the chapter “Measurement”, written by Cartwright

13_{The idea of a reflective equilibrium refers to a state of balance among beliefs. This equilibrium results}

from an interaction between our judgements and any principle or rule that we believe in on the one hand, and theoretical deliberations regarding our judgements or beliefs on the other hand. Any of these items may be revised to arrive at a coherent state. New beliefs may be added along the way. The practical aim is to guide our behaviour in moral issues. Alongside its conceptual use in ethics and political philosophy, the idea of a reflective equilibrium is also adopted in logic and epistemology (Daniels, 2013).

14

(25)

THE ERRORS MADE IN ECONOMICS 23

and Runhardt (2014). They describe three requirements for measurement, as well as the challenges faced in measuring social constructs. Well-executed research requires the following three activities: defining, operationalising and measuring. In addition to these three steps, I have identified two additional steps that are of importance in empirical economics: estimation and interpretation. These steps are discussed in more detail later on.

The characteristic features of each step help to identify the the errors that are currently made. These arise when features are not present in a the argumentation of a paper, or when the steps are poorly executed. Table 8 in the appendix gives an overview of each phase, its features, and errors that are found in the literature. This overview will function as a starting point for section 5, as depicted in figure 3, in which it is attempted to develop a non-na¨ıve measure of social norms. Tentative solutions to the errors are proposed to improve upon current approaches. I identify these tentative attempts as ‘challenges’, as the errors challenge us to find solutions.

Figure 3: Organization of section 4 and 5

The remainder of this section follows the process of research: definition - operationali-sation - measurement - estimation - interpretation. Examples from the literature are given, that will help to further expose the errors that are made. Moreover, in order to avoid a mere theoretical discussion, a practical example is adopted which will reoccur at each step. Following Cartwright and Runhardt (2014), the example of poverty in Bangladesh is used. The case of Bangladesh will be expanded alongside the outline of the theoretical framework. I will shortly introduce this example:

Poverty is a much analysed concept, especially within development economics. For example, its relation to economic growth and its link to inequality and human development are topics which have been examined by a diverse range of authors. While poverty is a concept that is self-explanatory for most of us, its application in research is less clear-cut. This is probably caused by the fact that poverty can be defined in multiple ways, and consequentially measured by various methods. A general definition of poverty proposes that poverty is a condition in which basic needs are not met, such as shelter, food, and clothing (Business Dictionary, 2016). Whereas this definition is useful if you are interested in merely getting to know what poverty means, it is far too general if you want to approach poverty from a research point of view. For the latter purpose, a more specific definition is needed. This choice depends on the aim of one’s research.

(26)

24 Definition

curiosity arose from an announcement of the World bank, which stated that Bangladesh reached the first Millennium Development Goal - the reduction of poverty - in 2013. With its 150 million inhabitants, Bangladesh is one of the most densely populated countries across the world. Despite the announcement of the World Bank, it is still considered as a very poor country. In light of these facts, our interest has been caught to find out more about the case of Bangladesh.

4.1 Definition

Many problems that occur when measuring social norms stem from problems encountered on the conceptual or theoretical level. Before measurement of any entity is possible, it should be clear what this entity signifies: What does this concept really mean (Cartwright and Runhardt, 2014)? Answering this question in the most clear and explicit way will make the execution of further steps - operationalisation, measurement, estimation - easier. In looking for the features of a good definition, I follow the line of thought of Cartwright and Runhardt (2014).15 The practice of defining reflects the explicit description of a category or quantity, listing all features that may be of interest. This enables us to determine if a unit belongs to our main category (Cartwright and Runhardt, 2014). This is a fundamental aspect of measuring social concepts. Fundamental in the sense that it should precede the development of any operationalisation (the way a category or unit is represented in scientific research), or measurement (the method needed to carry out a successful measurement of the concept). Cartwright and Runhardt (2014) propose these three steps as the three requirements for measurement in social sciences.

Let us have a careful look at what the process of defining really entails. It is an explicit description of a category. Moreover, features that characterize a concept should be listed. This enables us to categorize units as belonging to this concept. Concepts can be seen as choice sets, for which we want to determine which units belong to that set. For example, not having shelter or food can be identified as features of poverty. These features enable us to identify who is poor and who is not. The more specific we are in describing these features, the easier it becomes to identify poverty.16 _{Several features should be kept in mind by the}

researcher, when choosing a definition.

Firstly, the concept should be clearly formulated so that all dimensions are covered. Cartwright and Runhardt (2014) call this action ‘delineation’, the practice of defining the boundaries of a concept with detail and precision. When doing this, the purpose of the concept, i.e. its application in a (research) context, should be kept in mind. The purpose functions as a guide. As a consequence, a concept can be defined in multiple ways contingent on the aim of the research. While a definition may be fitting in one line with of research, it may not be for another. Because of the nature of socially constructed concepts, which are

15_{For some reason Cartwright and Runhardt (2014) use the term characterization when referring to}

defi-nition. I choose to use the more generally accepted term ‘definition’.

16_{Of course instances exist in which a general definition is preferred over a specific definition. In such cases}

(27)

not natural kinds that can be observed in nature,17clear definitions are of great importance. It helps to clarify the focus of one’s research. For instance, dynamic social norms may be delineated differently than norms that are stable over time. The aim, therefore, defines the formulation of the definition.

Secondly, the nature of social concepts is such that they have no definite boundaries. Nor do all members of a set share one list of characteristics (Cartwright and Runhardt, 2014). It may be easy to identify these aspects for natural kinds; in describing a tree, for example, we can list characteristics such as leaves, branches, and a trunk. For a social construct however, this is less evident. Social constructs do not share explicit characteristics but are connected through family resemblances. Wittgenstein (2010, 4th ed.) introduced the idea of family resemblances to refer to similarities among family members. Not all share the same set of characteristics, but they do have overlapping features, which allow us to identify them as members of the same family. Similarly, another family can be identified by different overlapping features. Family resemblances cause social constructs to be fuzzy.18 Fuzziness refers to the lack of concrete features that all members of a group share. This feature of social concepts makes their identification difficult, as no definite characteristics can be discerned.

Alongside the fact that we must deal with fuzziness and the socially constructed nature of social concepts, a third feature of a proper definition arises. Cartwright and Runhardt (2014) indicate that there is no guarantee that the formulated category will be useful for research. This seems to be a rather vague warning, but it is in fact something that occurs often. It may occur when a definition is too general, so that subjects are wrongly identified as belonging to the defined concept. It could also occur that definitions seem to define the same concept, but actually report different things. This can cause confusion. When, for example, two articles state that they analyse a similar concept, a comparison may show that this is not the case at all. If we had not been aware of this, we might have drawn incorrect conclusions. Moreover, it would have been very difficult to claim anything meaningful using these definitions.

The practice of defining can be illustrated using the example of Bangladesh. A general definition was stated in the introduction. Someone is poor when basic needs are not met, such as shelter, food, access to education, and clothing (Business Dictionary, 2016). While this definition gives a general idea about what poverty entails, it is not detailed enough for guiding our research. More specifically, if we wish to measure basic needs it should be clear what we mean by basic needs. Do we include only access to shelter and food, or access to education as well? The answer to this question should be guided by the aim of our research. Choosing among many definitions can be difficult. The nature and definition of poverty has been a much debated topic among researchers from diverse disciplines. The dimension-ality of poverty in particular has been an issue over the last years. The Human Development Index (HDI) can be seen as a product of the proponents of a multi-dimensional approach

17

The ontology of social constructs is debatable. Some scholars define them as natural kinds, others do not (for instance Searle (1995) or Hacking (1999).

18

(28)

26 Definition

to poverty.19 This index combines several dimensions of poverty (Kovacevic, 2011). These elements together offer the opportunity to analyse poverty from the perspective of human development: the acquirement of knowledge, the access to a minimum standard of living, and the possibility of living a long and healthy life (Business Dictionary, 2016). Being poor, then, is determined by the absence of one or more of these elements. The development of this index has been accompanied by a discussion on the theoretical level, where people such as Amartya Sen argue in favour of a capabilities approach to inequality and poverty (Sen, Dreze, et al., 1999). Humans need freedom to pursue different options - capabilities - such as education, health, relationships, etc. All combinations of these different options determine a person’s capability set, which is determined given their constraints (Alexandrova, 2014). Basic capabilities yield a threshold that is necessary for survival. Poverty may be defined as the absence of these basic capabilities. The other side of the debate is covered by the one-dimensionalists, who choose to approach poverty from one dimension: GDP per capita. Poverty lines, such as the $1 a day line, stem from this idea.

In addition to specifying the dimensions of poverty, the boundaries of the concept should be determined as well (as indicated by the second feature of a definition). In the case of poverty, absolute and relative poverty can be distinguished. Absolute poverty refers to the minimum nutritional standards that are needed to retain a certain level of physical health. The absolute poverty threshold is constant across countries. Relative poverty refers to minimum levels of living standards, which vary across countries. By limiting poverty to absolute poverty and contrasting this to relative poverty, the boundaries of what we are interested in become clearer. Together with the specification of the dimensions of poverty, research can move on to the execution of subsequent steps.

In conclusion, a proper definition: is clearly delineated, depending on the aim of one’s research, covers all units that belong to the concept, and lists all important features. Keeping these aspects of a definition in mind, it is clear that many definitions do not meet these criteria.

4.1.1 The definition error

Based upon the previously specified features of a ‘proper’ definition, several errors can be identified in the literature on social norms: 1) the definition of a social norm is missing, 2) a concept is not properly defined. The first error may occur when a definition is missing, for example when the measurement of a social norm is used to define what a norm is. The second error arises when a definition does not clearly delineate its borders, or when the definition actually defines something else. Several examples are given as illustration of these errors.

Definition is missing In the articles where a definition of social norm(s) is missing, the proposed measurement is most often used as definition of the social norm(s) of interest. This is the case in ‘The multiple facets of social capital’ written by Bjørnskov (2006), where

19_{It has been argued that the HDI gives a more encompassing view of human development compared to a}

(29)

social norms are proposed as explanatory variable of social capital. In seeking for a measure of social capital, a clear definition of social capital is given in the introduction to the paper. Social norms, however, are introduced without explanation on the first page as one possible dimension of social capital. Only in the empirical part of the paper it becomes clear what social norms are and what kind of norms are analysed by the author. He proposes four questions of the WVS as measure of social norms, that ask whether the following actions can be justified: “claiming government benefits to which one is not entitled, avoiding fares on public transport, cheating on taxes, and accepting bribes in the course of ones duties” (Bjørnskov, 2006, p.25). Several pages later these questions are described as measuring the strength of “internal enforcement of behaviour” (Bjørnskov, 2006, p.32). This is the definition which was warranted at the beginning of the paper, but it can be considered as being too general. Social norms are measured by a specific set of questions, indicating that a type of norm is measured in stead of norms in general. The real meaning of social norms in this paper can therefore only be deduced from the WVS questions, reflecting norms of honesty, trustworthiness or something along these lines.

In a similar vein, generalized norms are identified as a dimension of social capital in ‘A dimensional approach to measuring social capital: Development and validation of a social capital inventory’ of Narayan and Cassidy (2001). Similar to Bjørnskov (2006), Narayan and Cassidy (2001) aim to develop a measure of social capital. A mistake akin to the one encountered in Bjørnskov (2006) is made here. In searching for a definition and measurement of social capital, the definition of norms and their measurement is left unexplained up to a large degree. A description of what social norms are is not present in the paper, while their measurement is. In describing which survey questions are used to approach norms, the readers get some insights in what the authors have in mind when using the term ‘norm’. The selected items address the following themes: helpfulness of people, trustworthiness of people, and fairness of people (Narayan and Cassidy, 2001, p.67). When comparing these items to those selected by Bjørnskov (2006), it can be noted that the two articles ‘define’ and measure ‘social norms’ in two different ways. I will return to this issue in section 4.5.1. One could wonder why this instances are identified as errors. When a concept (social norm) is defined by its measurement, the phases of research are turned around. As stated by Cartwright and Runhardt (2014), measurement of a concept should be devised based on the definition of a concept and not the other way around. Moreover, where theory can provide ways to capture the fundamental nature of what a social norm is, data are more likely to capture the dynamic nature of social norms as norms are apt to change over time. It could therefore be the case that not one does not measure the intended construct when no clear definition is adopted to guide operationalisation and measurement.

(30)

28 Operationalisation

norm - (and/)or categories may seem similar but refer to different things. This occurs when ‘social norm’ is adopted as general term, while research is conducted using a specific type of norm, or when the term ‘social norm’ is adopted to refer to something which is not a social norm. As Reis (2008, p.71) indicates, “[...] it is a fallacy to think different methods measure the same concept just because we attach the same name to them.” Oftentimes authors are unaware of committing this fallacy. As a result we may think that a set of papers analyse similar concepts, namely social norms, while they actually examine different concepts. Results may be incompatible in these cases, as one could imagine that a norm of honesty can have implications which differ from a norm of equality.

Helliwell et al. (2014) identify social trust and generosity as social norms in their paper ‘How durable are social norms? Immigrant trust and generosity in 132 countries’. The paper aims to identify whether social norms are traits that may be transmitted across generations, or are shaped by the environment in which one lives. The authors conclude that both effects play a role, but the environmental impact is stronger than the generation effect. The error that Helliwell et al. (2014) make in my opinion, is the adoption of ‘social norm’ to refer to trust. I restate Bicchieri (2005)’s definition from the introduction: ‘behaviour that is collectively approved or disapproved in a group or population and is enforced by sanctions’. Two elements are of importance to identify certain behaviour as the result from conformation to a social norm: 1) the fact that its is approved by a group, 2) a conditional component that may result in a sanction. While we could observe one group to be more trusting than others, it is not clear that this behaviour has arisen as a result from collective approval, nor from the conditional expectation that others conform to a norm. In describing social trust as an individual trait, the absence of both dimensions becomes clear. Helliwell et al. (2014) thus uses ‘social norm’ to refer to something that is not a ‘social norm’.

The elasticity of a concept such as social norms complicates the development of a proper definition. It can be made to contain many different things, which is why definition is of utmost importance. In this way comparisons across papers can be made and it is avoided that seemingly similar concepts are compared.

4.2 Operationalisation

Towards measuring social norms