ALTERNATIVE PAYMENT SYSTEMS: A TWO-SIDED, CROSS SECTIONAL, MULTI COUNTRY EMPIRICAL ANALYSIS
On success and occurrence of alternative payment systems
by
Arend P. Lakke
A Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of
Master of Science in Economics
at
The University of Amsterdam, Faculty of Economics and Business October 2014
Supervisor: dr. W.E. (Ward) Romp
Abstract:
This paper presents an empirical analysis of the alternative payment systems LETS and Time banks using survey data to describe potential factors that influence the ‘success’ of these systems covering five countries. Information on the appearance of these systems is linked to unemployment, population density, and political preference data on municipality level. Besides some differences between the covered countries, we find that the number of participants and the age of organizations are significant factors for explaining the amount of currency (in official currency unit-equivalents) outstanding. Furthermore, we find that the occurrence of external aid (including non-financial) has explanatory power in explaining the size of a system in terms of participants. We also find for four out of the five countries that a high unemployment rate is a significant predictor for alternative payment system’s occurrence and four three of the five that population density is a significant predictor, and that these findings are robust after controlling for municipal political preferences.
Table of Contents
1 Introduction ... 3
2 Literature review ... 4
2.1 Defining alternative currencies... 4
2.2 The current state of the literature ... 5
2.2.1 LETS and Time banks ... 7
2.2.2 Local Exchange Traded System (LETS) ... 7
2.2.3 Time banks ... 8
3 Data collection ... 9
3.1 Field data ... 9
3.1.1 Sample pre-selection method ... 10
3.1.2 The questionnaire and method of collection ... 11
3.2 Non-field data ...12
3.2.1 Non-survey data collection ... 14
3.2.2 Potential issues with the non-survey data ... 15
3.2.3 Control variables ... 16
4 Data description & analysis ... 16
4.1 Age of the organizations ...17
4.2 Size measure of the organizations ...19
4.3 Governmental aid ...20
4.4 The use of software ...21
4.5 Size ...22
4.6 Number of participants ...24
4.7 Unemployment and population density as explanatory factors for initiative occurrence ...25
4.7.1 Empirical estimation on full dataset ... 27
4.7.2 Estimation on country level... 28
4.7.3 Empirical estimation including political preference ... 33
5 Conclusion and suggestion for further research ... 34
6 Reference list ... 35
1 Introduction
In day-to-day life, we usually use official money to settle transaction involving goods and services. There are, however, other means that are used for making payments that receive more and more attention among various disciplines. Currently, a total of 3418 local currency projects are identified in a recent cross-country mapping exercise by Seyfang & Longhurst (2013).
The aim of this paper is twofold. First, we aim to expand the work done by Seyfang and Longhurst by gathering additional information about a share of the 3418 identified local currency projects in an effort to attain a primer on potential forces that can explain the success of alternative currency projects. This will be done using survey data from a selection of the identified projects. From this additional information, we will describe two, variables empirically that might indicate the success of such system. The first is the value of the outstanding units within that payment system, of those systems that have provided a conversion factor from their system’s units to official currency. Furthermore, we will look at the number of participants within such as system as another proxy of an initiative’s success. Second, we will expand the data by locating a share of these currency projects and linking them to regional unemployment and population density figures in the expectation that both have the power to explain local currency project occurrences. This will be done using logit regression techniques. In both cases, we will focus on two most-established types of alternatives (LETS and Time banks) and the countries in which these alternatives are numerous enough to have meaningful statistical power. We will contrast these findings with the social motives that laid ground to the initial emergence of these currencies.
The current literature uses terms as community currencies, local currencies, complementary currencies or auxiliary currencies. We will use the term alternative currencies to indicate the units used in payment systems that are different from traditional or official payment systems. We do not use this term with the idea to diversify between complementary currencies and alternative currencies, as the latter is sometimes thought of as deceptive because such currencies are often not intended to substitute official currency.
We choose this term because, from an economic point of view, this is essentially what they are likely to do, especially when their importance grows in size.
Section two will provide a brief literature review, also including a description of LETS and Time banks purely for the purpose of this research. Section three describes the data collection phase and section four provides an extensive empirical analysis of the data, both from the results of the survey as well as on the occurrence of alternative payment systems in an economic view. Section five concludes.
2 Literature review
The purpose of this section is threefold. For the sake of the readability of this paper we have chosen to seek a brief wording for all that captures non-traditional money means of payments. But for the purpose of this research we only focus on two sub-groups of these, which require us to define exactly what we mean with ‘alternative currencies’. Secondly, we need to briefly set out the current state of the literature around what we call alternative currencies to be able to place our efforts and value-add accordingly. The third purpose is to zoom in on LETS and Time banks more deeply as these are the two types of alternative currencies we will have the opportunity to investigate further in our field research.
2.1 Defining alternative currencies
Schroeder et al (2011) define complementary currencies according to their characteristics such that community currencies are those currencies that 1) circulate in relatively small geographical areas, 2) are non-governmentally operated and issued, 3) are not or only limitedly convertible and 4) are zero or negative interest-bearing. They therefore do not include national currencies or enterprise barter systems (Philippe, 2011). Social currencies, local currencies, and community currencies are different nomenclature for the same thing, although it depends on the region which term is used. In non-Anglo-Saxon countries, the terms Trueque or Tausch may be used to name their complementary currencies, which are translated to ‘barter’. In Anglo-Saxon countries, however, this term is used for commercially oriented systems (Schroeder et al, 2011). It appears to be difficult to find a global and unique convention for denominating those currencies (Philippe, 2011).
It seems to make sense to continue to use the term community currencies for currencies that comply to Schroeder’s four characteristics, and use the term ‘alternatives to traditional money’ or alternatives if we seek to include other means of payment that are not governmentally operated and zero or negative interest-bearing, but that are easily converted to traditional means of payment or that circulate in wider geographical areas. We feel that such an expansion of definition is necessary to avoid exclusion, typically of the newer, digital types of alternatives such as crypto currencies. From an economic point of view, we should be interested in studying these alternative currencies as a group as they all may have potential economic impact, both in the real economy as well as on the functioning of traditional money itself. Furthermore, we see that also the traditional LETS systems sometimes take forms that surpass the definition of community currencies, which brings them closer than ever to other alternatives to traditional money. Direct convertibility to traditional money is one example, but more importantly, we found in our survey that a fair share of LETS organizations especially in Australia make use of Community Exchange System software which allows currency units to be spent outside of the local ‘exchange’ in another one (CommunityExchange, 2014). Even though this paper will only focus on LETS and Time banks, we will use the term ‘alternative currency’, or ‘alternative’ rather than ‘community currency’ or one of its apparent synonyms.
2.2 The current state of the literature
Up until now, the academic literature has deepened out various types of alternatives to traditional money. We find both research in the form of case studies, as well as survey-based research. LETS seems to have received the most attention which is likely due to its popularity (Collom, 2005). Most of this research is performed on a case-by-case basis, and these have always been limited either to the so-called community currencies, or to digital alternatives. Most research on community currencies is performed as single case studies (Pearson, 2003, Sahakian, 2013, Brenes, 2011), or comparative case studies (Mascornick, 2005, North, 2005, Dini, 2012).
Empirical work on community currencies has been rather limited. Collom (2005) appears to be the first one conducting a research towards the environmental factors that could potentially be related to the emergence and survival factors of community currency systems. The most recent work is that by Seyfang and Longhurst (2013) who made an effort to extensively map all the community currency systems known worldwide. Other studies include a research on the LETS population in the United Kingdom as well as the Australian LETS (Williams 1996a, 1996b, 1996c, 1997). Other studies have focused mainly on the participants in LETS schemes in the UK (Lies and Birch, 2000) and Australia (Williams et al, 2001a, 2001b).
The work of Collom (2005) relates to one part of the research presented in this paper as he asks the question in which types of social environments community currencies emerge and survive. The scope of Collom however is limited to paper-based currencies in the United States with 82 known occurrences of which 17 where successful. His explanatory environmental variables are however more advanced than the ones we will deploy in this paper, and it is clear that there is space for further research in this field.
As he notes, social movement theorists have discussed factors that could relate to social movement formation and success, of which the emergence of community currencies can be considered as one. Such factors, he continues, include age, education, race, martial status, homeownership, residential stability, population size and population density. Furthermore, he hypothesizes that as these types of currencies are originated to empower the economically marginalized, low-income, high unemployment and poverty areas are expected to be in greater need of community currencies, therefore also expected to more likely have (had) some.
Furthermore, Collom hypothesizes that cities with large self-employment sectors and more out-of-labor-force residents are likely to be more environmentally conducive for these types of systems. Hirschman (1970), as Collom cites, notes that participants in community currency systems have, to some extent chosen to ‘exit’ the mainstream economic system.
With respect to the relation between alternative currencies and unemployment, Collom (2005) hypothesizes two ways through which these may be connected. That is, through the fact that high-unemployment areas are in greater need for such systems and through the fact that participants do not engage in mainstream economic activity. More specifically, he finds that, among others, economic marginality is an important factor in characterizing the cities in which paper community currencies have been attempted. He finds that mean values of unemployment and self-employment are both significantly higher in areas where alternative payment systems exist or have existed.
2.2.1 LETS and Time banks
We will now zoom into two types of alternatives that have turned out to be eligible for our field research, being the Local Exchange Traded System currency and the Time bank currency.
2.2.2 Local Exchange Traded System (LETS)
The LETS system was initiated in 1983 as emergency money during recessions and spread out rapidly in the following twenty years through Canada, the UK, New Zealand and Australia. Historically, LETS systems used to develop in circumstances of recession when cash shortages hampered economic activity but they have also been adopted as strategic tools to meet social, environmental and economic objectives (Seyfang, 2001). It has been advocated as a tool for community resilience and as an alternative to perceived unsustainable mainstream economic growth, but it should be noted that there are various political values and objectives held by LETS proponents (Seyfang, 2001).
LETS are schemes or systems facilitating the exchange of goods and services within their local community. The number of participants in that particular scheme binds the local community in which exchange can take place. The organizations are operating on a non-profit basis, run by a sub-group of the participants. To cover administrative expenses, members are often charged an annual fee, which may be required payable in the official or alternative currency or a combination of the two. Transactions take place by transferring currency units from the purchaser to the provider of a service or good, which can occur manually using cheques or via commercial or proprietary software systems.
An important feature of LETS is that credits need not be earned before they are spent. The currency is not issued by an overhead authority but by its members, by the act of spending (Fitzpatrick and Caldwell, 2001, Seyfang & Longhurst, 2013)
Theoretically, the amount of units in LETS systems is therefore unlimited and economic activity is no longer restricted by a lack of money (liquidity)1. As the system does not account for interest, a negative balance means nothing but a commitment to the group to render goods and services in the future and the accumulation of savings is incentivized in a less severe way than the case of traditional money.
According to Seyfang & Longhurst (2013), growth in number of LETS systems peaked in the late 1990s followed by Europe. Furthermore, LETS systems have evolved into national specific adaptations. Government money has played a role in the early 1990s in the UK and still does in Europe (Seyfang & Longhurst 2013). It is perhaps due to the fact that LETS systems seemed to have a peak in the mid-1990s, that there is not much work from the recent years focused on the system.
2.2.3 Time banks
Time banks, as in the form in which they are most commonly known, were created in the USA in 1986 by Edgar Cahn, as a means to activate the skills and resources of the economically marginalized in deprived neighborhoods. The model first spread across the USA and then to the UK in 1997. Currently, the USA and UK Time banks networks are nationally coordinated, with national administrations and common websites2.
1
Indeed, this unlimited number of units in a LETS system is something that is hardly seen in practice. We had the ‘luck’ that due to initially unclearly posed question in the questionnaire, many organizations responded to the ‘number of units’ question with the upper and lower limits of units an individual can have in his or her account instead of the number of units that were in circulation at a given time. These boundaries were instituted to limit excessive saving or borrowing from the community.
2
Time banks and closely comparable currencies, as they are categorized by Seyfang & Longhurst (2013), are part of the largest group of community currencies (service credits) that makes up 50.2% of the total number of projects identified. Alike LETS systems, members can enroll and advertise the services they would like to offer or receive. An important difference between Time banks and LETS is that the currency is by definition time-based, whereas at LETS we see that the currency may also be linked to official currency or work without any of such base. Typically, the currency unit is in hours and participants earn credits by spending time offering services to someone else. Most importantly, this implies that everyone’s service or work is worth the same, which is not a prerequisite in LETS systems. Furthermore, Time banks generally tend to enjoy more extensive government ruling on encouraging participation of the unemployed, as participation in such schemes is seen as volunteering, while LETS participants are sometimes seen as earning money income from the transactions for which they may be taxed under certain circumstances (Belastingdienst, 2008).
3 Data collection
In this section we will discuss the method of data collection in detail. We make use of two types of data: field data, obtained via a questionnaire sent out by e-mail and data obtained from public databanks on unemployment and population figures (which we will name the ‘non-field data’). As this subject did not receive much large scale field attention so far, we designed a questionnaire aimed at obtaining information on the age and size of the organization, which are both regarded as success factors, and the extent to which the organization receives government benefits or makes use of software, which are considered as control factors. The creation of a custom questionnaire brought about a number of practical issues, which are topic of discussion in the first part of this chapter.
3.1 Field data
We built a questionnaire in order to be able to collect data from numerous initiatives that may or may not refer to themselves as local currency initiatives, alternative payment systems or community currency systems. Because some of the initiatives are much more widespread than others in terms of numbers (or concentration) in one country, and because we do not have the time and resources available to contact all of
the 3418 currencies as counted by Seyfang & Longhurst (2013), it made sense to only include a selection of these local initiatives in our initial mailing. As also language plays a bounding role in this single-man performed research, we applied a number of criteria for our sample building.
3.1.1 Sample pre-selection method
The following criteria were used to select which organizations were to be included in the sample:
1. The members of the contacted organization are likely to feel comfortable communicating in English, French, Spanish, Italian or Dutch;
2. We would generally only consider the most important type of alternative of each country but also consider initiatives that are large enough in number, to have some statistical explanatory power, given the rule of the thumb n > 30. 3. E-mail addresses must be obtainable with reasonable effort and are close to
complete for the complete population within each country according to the empirical data by Seyfang & Longhurst (2013).
The first criteria is unavoidable given the time and resource constraints with which we have to work, and rules out a couple of ‘big fish’ such as the German ‘Tauschringe’ and ‘Seniorengenossenschaften’ with about 380 occurrences and the Japanese ‘Fureau Kippu’ / Time banks with 391 occurrences as well as various local currencies (133). The second criteria is meant to rule out relatively unimportant occurrences to remain a reasonable amount of focus and the third is the most stringent one as it appeared quite tedious to obtain e-mail addresses from some of the larger groups of occurrences. Time banks in the US are well organized, with a central website and system in which numerous data can be requested on a real-time basis, but obtaining the e-mail addresses in order to obtain the data necessary for this research would have been extremely time-consuming. The Italian e-mail address database on Time banks was far from complete, and the UK Time banks governing body could not share some of the requested because of confidentiality reasons. On the other hand, we decided to include LETS Australia as we found more occurrences there, than described in the work by Seyfang & Longhurst (2013).
For the remaining groups of initiatives there was a website available which contained an accessible database of e-mail addresses, sometimes smartly decoded (Spain) or in a Google maps environment which made the collection of the e-mail addresses more difficult than expected at first glance. A breakdown of the selection process is provided in appendix 1.
3.1.2 The questionnaire and method of collection
As we are interested in expanding our understanding of the life dynamics of these alternatives, but also have to be wary of the time constraints of the potential respondent and their own constraints with respect to data availability, we decided to keep the questionnaire short and simple, and to ask for the following:
1. The year at which the local organization was initiated;
2. The way at which units of the alternative change hands (manually, or via software);
3. The current number of users of the alternative;
4. The number of units of the alternative currently issued;
5. The equivalent worth measured in the local currency, if applicable; 6. Whether any government aid or subsidy is being or has been received.
The first two questions aim to build an understanding of the age of the organization as well as the extent to which it has developed in a way to facilitate expansion and security. Furthermore, non-software users could potentially have reduced ability to provide the data asked at question four. The third question is straightforward, as it directly corresponds to a measure of ‘success’. The fourth question, posed in multiple ways, as we will describe in more detail below, was formulated in this way as to provide comparability with other currencies such as paper-based currencies and crypto-currencies. An alternative way of grasping the extent to which an alternative is being used is by means of number of transactions or the total amount of transacted units. These data would however likely be unobservable for paper-based currencies or crypto-currencies due to a lack of data records while the number of outstanding units is very much observable. We chose not to rule out such comparability in order to facilitate further expansion of this dataset. The fifth question helps making the data obtained in question four comparable but also tells us something about the users’
views on how these alternatives should be perceived in the context of the ordinary economy. The sixth question again is a straightforward question directly related to a potential control factor related to the ‘success’ of an initiative.
The questionnaire was subsequently translated into French and Spanish and in both cases also the English version was attached, except for the questionnaire sent out to the Dutch organizations. To enhance the commitment of readers of the e-mail, combined with maximizing efficiency, the e-mails were sent out using automation software, which allowed an extent of personalization, such that every e-mail was enhanced with the zip code and city of the addressee organization. Because especially the fourth question appeared to be ill-posed, particularly in the Dutch, French and Spanish e-mails, which were more or less sent out at the same time, and to a lesser extent also in the English version, a subsequent personally addressed e-mail was issued after a response came in. Because many of the organizations do not perceive their alternative as a currency with the like circulation, and because the number of positive units in a LETS based system should theoretically always equal the number of negative units, respondents often replied that they did not understand the question or replied with the minimum/maximum number of units a participant in the system can possess. The additional e-mails addressed these issues and therefore contributed majorly to the quality of the data. All standardized e-mails are included in appendix 2.
As the first e-mails were sent out as soon as the first complete country set of e-mail addresses were collected, it was after one week since the first e-mail until all e-mails were sent. To those e-mail addresses that did not respond after a month, a reminder was sent.
3.2 Non-field data
In addition to the field data we made use of datasets on population density and unemployment, obtained from governmental websites and linked to the initiatives based on their zip codes. While some organizations clearly had activity outside of just one zip-code region (obvious examples include the capital cities of each country as these typically contain multiple zip-code regions), allowing for these complexities would require averaging obtained figures of multiple zip-code areas. As the actual spillover activity outside the zip-code area of the organization as found in the data
source is not necessarily large, and as it seems more likely that the lion’s share of the activity takes place in the largest city at which the organization appears to be active, we chose to ignore these complexities and therefore also to avoid potentially arbitrary averaging of the location-specific data. In the few cases where an organization was listed as being active in multiple towns or villages, the zip code of the most highly populated town was used as the main zip code. We should note here that these corrections could cause an upward bias in the coefficients found between initiative occurrence and population density, should they be made in cases where they should have not.
Another issue stems from the fact that we used the contact details of each organization as reference point for the location or epicenter of the organization’s activity. This however is not an optimal as the surveyed organizations typically do not have an official office. The contact details and corresponding zip-codes therefore likely belong to the residence of one of the members in charge of external communications which is not necessarily also the center of all the organization’s activity. We have seen obvious examples, where the name of the organization refers to a large city while the zip code and city listed in the contact details are those of a small village in the proximity of the city. Such obvious discrepancies have been corrected, as the local data searched for might be sensitive even to these small differences in location.
The measures used in the analysis are as follows:
1. Population density, measured as the number of inhabitants per square kilometer for the most recent year available;
2. Unemployment, either measured as:
a. The degree of unemployment, which is the average percentage of the workforce (age-bounded) that has been unemployed during 2012 (France, Belgium, Australia).
b. The difference between the work force and the participating work force3 (The Netherlands).
3
Post-employment temporary benefits (“WW-uitkeringen”) are not the same as unemployment benefits as they are typically temporary and convert to unemployment benefits after a number of years of unemployment, based on years worked.
c. The number of unemployed people divided by the number of inhabitants of that region, both obtained from different yet compatible datasets (Spain)
3.2.1 Non-survey data collection
We have assembled a dataset containing unemployment and population density information on regional level for all the countries we have observations for, and have combined our information on where the initiatives are situated with the unemployment and population density information.
For each country, the regional data used is the most granular data freely available on the internet and is obtained from government statistical websites. As the countries differ in their definitions of statistical areas, which can be understood by looking at the number of statistical areas compared to the number of inhabitants or the size of the country, we again make use of country dummies to capture these state-specific effects. France seems to have the most granular data at hand; with over 36,000 statistical areas specified while The Netherlands has only 403. This is going to affect the variances within the countries, and because this also affects the statistical region to initiative ratio, this will also affect the calculated marginal effects of the explanatory variables which will be discussed in the next chapter.
Apart from these granularity differences, the preciseness of the data may be harmed by a couple of factors including the conversion process the data has gone through in order to be matched with the data on initiative occurrence. While all the initiatives that were selected for the survey were supplied with a postal code, these postal codes sometimes were not in use anymore, which required a manual correction. Furthermore, in the case of Spain, such correction was partly obstructed by the fact that municipality names were available both in Catalan and in Spanish, and that the city or municipality names that came with the contacts were not always directly traceable to either one of these. Belgian municipality data lacked a postal code link, which required us to manually connect the alternative organizations’ names, written in either Flemish, Welsh, or French to the statistical data. Finally, the Australian statistical areas are created in such way that each area more or less contains a comparable number of citizens, which leads to the situation where each city has a
statistical area of the center, and another one of the area surrounding the center, each of which that may be split up in Center, North, East, South or West parts. This caused some problems when the data had to be connected to postal code areas or city areas, in which initiatives were active. We chose to drop the instances in which an obvious connection to the location of an initiative and appropriate unemployment and population density data was not at hand, causing a dropout of one-third of the initiative occurrences (see table 1 and 3). In cases where an initiative was located in a city with multiple statistical areas, the ‘center’ area was chosen as best fit, as this most closely aligns with the treatment that initiatives in the other countries received. This, again, potentially causes an upward bias in the estimated coefficients in the regression.
3.2.2 Potential issues with the non-survey data
We should mention a number of caveats related to the characteristics of the non-survey data and their applicability for our purposes. We will quickly glance over the problems of cyclicality and age distribution.
Reverse causality
We will test whether unemployment and population density figures are linked to the occurrence of alternative currency organization. While we expect that this type of organization emerges in environments with higher population density and higher unemployment, we should make note of potential endogeneity because the occurrence of large and successful alternative currency organizations in a region might cause more unemployment itself. This is the case under the assumption that working in an unofficial barter economy, providing services to each other and making payments using an alternative currency, could replace participation in the official economy. While a complete replacement of the official economy is not impossible4, it seems unlikely that such a situation would have stayed unnoticed so far. For this reason, we will not control for endogeneity.
4
For an example of an independent village using an alternative currency: http://www.earthaven.org
Composition effect
Furthermore, both unemployment measures include all ages while the local initiative may have a different age-distribution than the region it is situated in, which means that other age-cohort specific unemployment measures could be more appropriate. Controlling for this requires more in-depth information about the age composition, which is not part of our survey query.
3.2.3 Control variables
Because different countries have different economic situations, the actual levels of unemployment may not be the best measure to use in cross-country analysis. In order to be able to perform multivariate regressions, we therefore introduce a number of control variables to capture country specific effects among which the potential measurement differences in unemployment and population density figures. We do this by introducing a dummy variable for the countries Australia, Belgium, France and Spain such that their coefficients describe the country specific effect compared to The Netherlands, where the sample size is closest to that of the population.
4 Data description & analysis
In this section we will dive deeper into the obtained data, to see whether we can see any evidence for the expected relations between the data. We note that the number of e-mail addresses found closely corresponds to the number of initiatives found, which indicates that our data would not suffer much from coverage error, nor would we have problems validating the appropriateness of our sample as it is just about the same as the population (table 1).
The data does however suffer from a nonresponse error. Especially when it comes to discussing initiative success of any kind, it is important to keep in mind that it is likely that unsuccessful initiatives are nonresponsive. They would for instance not check the listed e-mail address anymore or would not feel the need or have the information available to respond to the questionnaire, given that it relates to something that is in their eyes nonexistent. But, we did receive eight responses to the questionnaire with the notion that the initiative is no longer up and running, or that it has never been operational. Such signs seem to improve our success measures’ accuracy, albeit not in a way that we could use this in our analysis at this point.
Our measures of success for the alternative, and any of the correlation we may or may not find are therefore contaminated by the non-response error (survivorship bias) as we only included responses from the survivors, which are already successful in a sense.
In the last part of this section we will ignore the data obtained through the questionnaire for a moment and turn to the information we have on where alternative currency organizations exist or may have existed. We will hypothesize that LETS/Time bank organizations are more likely to occur in areas with higher unemployment and higher population density.
4.1 Age of the organizations
Spanish and Belgian initiatives are much younger than the others as can be seen from table 1. This seems in line with the findings of Seyfang & Longhurst (2013) who report that Belgian national coordination from 2009-2011 initiated a doubling of all projects, as well as with their notion that the initiatives are still expanding in number, and that the Spanish number of initiatives is growing particularly in the last few years. They also note that the French SEL/LETS systems, of which our dataset consists, are likely to be in decline and suffer from the same problem that it is unknown how many of the SELs are still active.
When we look at our other activity measures, such as the number of participants in table 1, we can easily see that the distribution of number of participants in Spain and The Netherlands is skewed as the average numbers are much higher than the medians. Especially in these two countries, large outliers push up the averages. The Netherlands has one organization with 4500 members and one of 900 members which are both very large compared to the median. In Spain an organization with 1100 participants and one of 760 participants push the figures up. Furthermore, all the distributions seem positively skewed to some extend, indicating that each country has a couple of exceptionally large organizations.
Table 1: Generic statistics
Australia Belgium France Netherlands Spain Total
Number of initiatives found 67 119 438 91 157 872
Number of e-mail addresses
found5 67 119 424 87 157 131
Number of observations 11 29 37 39 15 131
Response rate 16.4% 24.3% 8.7% 44.8% 9.5% 15.3%
Average age of initiative 12.9 8.1 11.6 12.3 4.3 9.9
Median age of initiative 11 4 11 15 2 8
Average number of participants 110.6 107.4 79.5 199.5 261.7 144.6
Median number of participants 100 90 70 60 115 70
Average size of the initiative in
official currency € 23,263.67 € 30,687.45 - € 7,382.65 - € 19.865,29
% of obs. with known currency
size 54.6% 17.2% 0.0% 59.0%
6.7%
(1 obs.) 34.4% % of obs. with government
financial aid 18.2% 20.7% 8.1% 18.0% 26.7% 16.8%
% of obs. with financial &
nonfinancial aid 27.3% 41.4% 24.3% 25.6% 53.3% 32.1%
% of obs. using software for
administration 81.8% 89.7% 13.5% 59.0% 40.0% 51.9%
5
These numbers exclude double counting of multiple mail addresses per organization and are not corrected for the numerous undeliverable e-mails that were returned because of invalid e-mail addresses.
4.2 Size measure of the organizations
The size figures are controversial. As economists, we like to express the world in terms of money and we have various methods to do so. The size measure is constructed as the sum of all positive balances of the participants, which are then converted to the equivalents in the respective official currency if a conversion factor was provided by the organization. As many organizations do not have the compatibility to make this calculation, or such a conversion factor, the latter sometimes for ideological reasons, this size measure restricts our sample size somewhat, but makes organizations mutually more comparable.
As noted before, the sum of all accounts in a LETS system should equal zero by definition, which means that the correct measure of spendable currency is the sum of all positive balances. We have chosen this measure to enhance the comparability of the figures to other means of payments. In order to let this size figure be an appropriate measure of activity such that a larger size means more activity, the participants and systems have to comply with two (fairly strict) assumptions, being:
1. All organizations adhere to the principle that units are only created by spending them;
2. There are typical creditors, and debtors, and their status is persistent over the year.
The first assumption rules out the organizations that actively induce inflation, by for instance giving credits to newcomers. Such activity would lead us to overstate the activity measure, as the positive balances are not activity induced. The second assumption is necessary to assure that active members do not have close to zero balances, on average, which would lead us to completely miss their activity.
While we have seen in the responses to the questionnaire that some organizations indeed do violate assumption one, by giving newcomers some credits to spend, we should note here that this is never a recurrent event, and that some organizations gradually counter these inflationary pressures by introducing an annual fee for a couple of years. Furthermore, the number of credits that are freely given is typically
small, and this only occurs in situations where the credits are tied to hours. We therefore continue to assume these assumptions are fulfilled and that our measure of activity makes some sense from a theoretical point of view.
We should stress here that typically the French organizations quite unanimously resented the question with respect to convertibility to Euro equivalents as they typically measure their credits in terms of time (which makes them Time Banks effectively) and as some of them strictly prohibit conversion. The Dutch and Australian LETS organizations, however, often had conventions on the convertibility to Euro and Australian dollars. The Spanish respondents were selected as being Time banks. One of these organizations appeared to have a conversion factor for their time units, but seemed to be a LETS organization eventually. These differences are reflected in the differing percentages of organizations with a known EUR equivalent of size as can be seen in table 1.
Sometimes, organizations replied to the conversion factor question (5) with a conversion factor from their local currency to a time measure, either minutes or hours. While this opens the way for further conversion to an official currency equivalent, we refrained from doing so as the improvement of the data is questionable, this is not information asked sample-wide and because this would probably be in ideological conflict with the organizations intentions in providing this information
4.3 Governmental aid
With respect to government aid, the questionnaire was structured with only one question referring to whether an organization received any government aid or not. In the responses, however, it appeared that quite a substantial portion of the organizations didn’t receive reoccurring financial aid, but received some help through having free office room and sometimes a telephone line at their disposal. There were some other occasions in which an organization said to receive some aid, but of which we thought it differed from what we expected to see in response to the question (such as recurrent financial aid). Such diversions included aid from philanthropic organizations, meeting room availability or very small financial contributions ranging from 50 to 500 euro on an annual basis including the payment of an insurance plan by the municipality. We therefore decided to account for these small, non-financial governmental aids in a separate way, creating a dummy variable for plain financial
aid in the form of money recurrently coming into a bank account (AID) and one for more broadly defined aid, including the above but also the small, non-financial contributions (AIDSM).
4.4 The use of software
We can see the use of software in a couple of ways. While it is most appealing to see it as a degree of modernization in order to facilitate transparency and accountability, and to facilitate a larger number of participants, there might also be a case for de-socialization. In situations where a LETS system has a strong communal function, where participants regularly meet and clear their accounts, such software might be undesirable. We did not ask to provide a reason next to the choice to have software or not, which is why we cannot say anything about this but it is important to keep this in mind. What we have regularly seen however, among all countries, is that next to a software system, some organizations tend to have some manual activity for those people that do not possess a computer or Internet connection.
We initially did not have any expectations with respect to the dispersion of software use. We only assumed that software-users would have a higher response rate than non-software users to which we will turn shortly. Software usage, however appears to vary largely from country to country.
It is now time to discuss the relation between some of the variables. To recap, we are on the one hand interested in gaining an understanding how the different characteristics of the organizations are related, and will define the number of participants as a dependent variable as well as the size of the currency measured as the sum of all positive balances translated to official currency units. We will look into whether there is a link between these measures of organizational size or success with external aid defined as financial government aid and a more broader defined aid variable also including non-financial non-governmental aid as well as with the degree of modernity, which is approximated by whether an organization uses software for the creation of its marketplace or not. Furthermore, as the age of the organization (in years) might be linked to either the number of participants and therefore the size of the organization, and the extent to which software is used, we will include this variable separately, too.
On the other hand, we have the opportunity to look into the relation of alternative organizational occurrence in a macroeconomic context, which we try to link to population density and unemployment information. In both cases we capture country fixed effects using dummy variables. The regression results are presented in table 2.
4.5 Size
Our regression model looks as follows, where the country dummies are for the countries Australia, Belgium, France and Spain. We note that the number of observations has decreased to 35, which we have to divide over four countries (as no French organization has supplied with a conversion factor to Euros). The statistical power of the regression equation is therefore somewhat low.
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 = 𝛽𝛽0+ 𝛽𝛽1𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 + 𝛽𝛽2𝐴𝐴𝑆𝑆𝐴𝐴 + 𝛽𝛽3𝐴𝐴𝑆𝑆𝐴𝐴𝑆𝑆𝐴𝐴𝑆𝑆𝐴𝐴𝐴𝐴 + 𝑐𝑐𝑆𝑆𝑐𝑐𝑐𝑐𝑆𝑆𝑆𝑆𝑐𝑐 𝐴𝐴𝑐𝑐𝐴𝐴𝐴𝐴𝑆𝑆𝑆𝑆𝑑𝑑
and;
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 = 𝛽𝛽0+ 𝛽𝛽1𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 + 𝛽𝛽2𝐴𝐴𝑆𝑆𝐴𝐴 + 𝛽𝛽3𝐴𝐴𝑆𝑆𝐴𝐴𝑆𝑆𝐴𝐴𝑆𝑆𝐴𝐴𝐴𝐴 + 𝛽𝛽4𝐴𝐴𝐴𝐴𝑆𝑆 + 𝛽𝛽5𝑃𝑃𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑐𝑐𝑆𝑆𝑃𝑃𝑆𝑆𝑐𝑐𝑆𝑆𝑑𝑑
+ 𝑐𝑐𝑆𝑆𝑐𝑐𝑐𝑐𝑆𝑆𝑆𝑆𝑐𝑐 𝐴𝐴𝑐𝑐𝐴𝐴𝐴𝐴𝑆𝑆𝑆𝑆𝑑𝑑
In the first regression, only the Spanish dummy appears to be very significant, which is easily explained by the fact that they have just one organization that was able to provide us with size figures convertible to the official currency, and that this organization was of considerable size. The dummy picks up this full effect. Furthermore, we note that the variables are mutually significant on a 1% level as well as that the variables of interest, aid and software have the expected sign, albeit not close to significance. The second regression, where we include the potentially omitted factor Age and Participants is an improvement over the former judging from the improved adjusted R-squared. Furthermore, its inclusion seems relevant to the extent that both the aid as well as the software dummy coefficients have been rendered insignificant and with a counterintuitive sign.
Table 2: Regression results on survey data
OLS OLS OLS OLS OLS OLS
Size Size Size (NL) Size (NL) Participants Participants
Number of obs 35 35 23 23 131 131 R-squared 0.55 0.72 0.51 0.94 0.09 0.06 Adj R-squared 0.45 0.64 0.43 0.92 0.03 0.03 Constant -1660 (11186) -18488 (12033) -3315 (5004) -513 (2645) 4.94 (108.31) -25.55 (78.39) Spain 184451*** (34205) 193728*** (29357) 95.25 (133.52) France (omitted) -84.24 (101.67) Australia 14621 (15222) 32554** (13459) -119.62 (140.96) Belgium 32805* (17644) 43955*** (14612) -101.99 (108.87) Software 5131 (14231) -5182 (11868) 11309* (6494) 1671 (2533) 89.04 (87.43) 71.53 (71.83) Aid 2965 (22247) 23825 (20788) -43877*** (11897) 1585 (5952) -115.76 (127.07) -104.34 (125.73) AidSmall 16600 (17787) -3621 (16823) 42257*** (10267) 770 (5301) 222.93** (103.00) 228.01** (101.33) Age 1528** (676) 119 (156) 8.73* (5.22) 7.78 (4.80) Participants 19.48** (7.44) 18.44*** (1.70)
The change of sign may be explained by the fact that the age of organizations is correlated with the dummies Aid, AidSmall and Software by 6.0%, 12.6% and -12.8% respectively, indicating that older organizations are less likely to deploy any of the three, or that not deploying any of the three causes organizations to be shorter lived. The number of participants seems also relevant in explaining an organization’s size, which makes sense under our second assumption, which is that there are typical debtors and typical creditors and that their status is stable over the year, meaning that users with a positive balance will keep to have a positive balance, and that more users just means more positive balances (as well as negative ones). The positive, significant sign of the number of users coefficient is likely to pick this up.
We repeat the two regressions for the observations in The Netherlands only, because we have by far the highest response rate as well as size measures for over a quarter of the total population in this country (23 out of 91). The inclusion of the number of participants turns out to be important again, as the sample contains two very large organizations that are likely to have different size characteristics if our assumptions hold. We note that in the full regression, including age and participants control variables, all coefficients have the expected size and that they are mutually significant on a 1% level.
4.6 Number of participants
We regress as follows; with country dummies are for the countries Australia, Belgium, France and Spain. We immediately include the omitted variable Age as its exclusion leads to a worse adjusted R-squared. The results are again shown in table 2.
𝑈𝑈𝑑𝑑𝑆𝑆𝑆𝑆𝑑𝑑 = 𝛽𝛽0+ 𝛽𝛽1𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 + 𝛽𝛽2𝐴𝐴𝑆𝑆𝐴𝐴 + 𝛽𝛽3𝐴𝐴𝑆𝑆𝐴𝐴𝑆𝑆𝐴𝐴𝑆𝑆𝐴𝐴𝐴𝐴 + 𝛽𝛽4𝐴𝐴𝐴𝐴𝑆𝑆
+ 𝑐𝑐𝑆𝑆𝑐𝑐𝑐𝑐𝑆𝑆𝑆𝑆𝑐𝑐 𝐴𝐴𝑐𝑐𝐴𝐴𝐴𝐴𝑆𝑆𝑆𝑆𝑑𝑑
While the variables AidSmall and Age now both seem to have a significant coefficient, which is also in line with our expectations, we should note that the F-statistic for mutual significance indicates that the coefficients of all the variables together are not statistically different from zero. This is in line with the R-squared and adjusted R-squared statistics, indicating that the model is not a good fit. Furthermore,
the country dummies indicate that the number of participants is country specific, as the two very large organizations in The Netherlands and the one in Spain seem to have significant impact on the country dummy coefficients. Furthermore, the age of the organization seems significant, but we should also note that the age of the organizations differs substantially cross-country, which is why it is potentially related to the country dummies.
Most interestingly, we also see that the AidSmall coefficient is positive significant, even after exclusion of the country dummies, which might hint that larger organizations are more likely to seek funds from somewhere, or that fund or aid-receiving organizations are more likely to have more participants.
4.7 Unemployment and population density as explanatory factors for initiative occurrence
Regardless of the responses to the questionnaire, we can look at the information we have about the occurrences of alternative currency organizations and some basic regional data. By doing this, we expand on the work by Collom (2005) who hypothesizes that unemployment is higher in areas where alternative payment systems emerge, as well as that these areas should be more densely populated. Collom only finds statistical evidence for the former hypothesis. Our dataset is more expansive, covering five countries. We do not have the opportunity to address all the explanatory factors Collom uses and will only look at population density, unemployment rates and for a subset of the sample (The Netherlands) we will also include the results from the most recent Second Chamber elections on municipality level, to capture potential omitted social movement resources. We use these voting data as a proxy for these potential omitted factors as the data is easily accessible on municipality level for The Netherlands, as the author can apply its own reasoning in categorizing the political parties in the political spectrum (left versus right, conservative versus progressive) and, most importantly, these data may be linked to unemployment figures and social movement factors as follows.
Following Rehm (2011), we may safely say that conventional wisdom and research states that the ‘haves’ tend to sympathize with the Republicans in the US and the ‘have-nots’ with the Democrats. Rehm also finds that people’s expected future
income is an important indicator in shaping political preference. When we extend this notion to Dutch political parties and their common classification according to Weetmeer (2012), we may make the assumption that left-progressive voters have lower (expected) mean-incomes and/or higher unemployment rates. Furthermore, we may see the emergence of alternative payment systems as a manifestation of progressive thinking, especially when we consider the fact that both LETS and Time banks have their roots in providing a platform to help the economically marginalized.
If we think about political preference in this way, we might hypothesize that left-wing progressive political preference (these two categorizations go hand in hand for almost all6 Dutch political parties) is linked with the occurrence of alternative currency systems, as well as with unemployment rates. We therefore will include political preferences, modeled as dummy variables (left = 1, right = 0 and progressive = 1, conservative = 0), in the logit-regression for The Netherlands.
Before we turn to estimating the model, we should note that the initiatives as from our dataset are not necessarily dispersed among the countries in the same ways. Judging from figure 1, which visualizes the initiatives on their respective country maps by small dots, we see that the payment systems tend to cluster in or around large cities, especially in Spain, Belgium and France while those in The Netherlands seem to be more equivalently dispersed among the country. Australia seems to be the odd one out, but it is important to keep in mind that this country is by far the least densely populated7 of the five, and that most of the population of the country resides at the east coast or at the lower west coast, which coincides with where the initiatives are situated. It is tempting to draw conclusions with respect to the link between population density and initiative occurrence by looking at these pictures, but we can only keep these in mind while interpreting the regression results presented in the next section.
6
Only the Dutch “Partij Voor de Vrijheid (PVV)” is categorized as left-wing and conservative.
7
Although, because of the specific statistical geography of Australia, we will not see this in the data.
4.7.1 Empirical estimation on full dataset
We will now turn to the estimation of the following model:
𝐴𝐴𝑆𝑆𝐴𝐴𝑆𝑆𝑆𝑆(𝐼𝐼𝑐𝑐𝑆𝑆𝑆𝑆) = 𝜂𝜂 + 𝐷𝐷𝑆𝑆𝑐𝑐𝑑𝑑𝑖𝑖+ 𝑈𝑈𝑐𝑐𝑆𝑆𝐴𝐴𝑃𝑃𝑗𝑗+ 𝑐𝑐𝑆𝑆𝑐𝑐𝑐𝑐𝑆𝑆𝑆𝑆𝑐𝑐 𝐴𝐴𝑐𝑐𝐴𝐴𝐴𝐴𝑆𝑆𝑆𝑆𝑑𝑑
Because the variable Init is binary, being one when a region is known to have at least one alternative currency organization, and zero when there is none, we perform a logistical regression and express the regression results into odds ratios for a more straightforward interpretation (table 3). An ordinary least squares model can be used to fit a linear probability model. However, since the linear probability model is heteroskedastic and may predict probability values beyond the (0,1) range, the logistic regression model is used. With a regression Chi-squared probability of 0, the model seems to be correctly specified, indicating that the coefficients of the model are mutually significantly different from zero.
Because the specification of the statistical areas among the countries is so diverse, the odd-ratios of the country dummies have no straightforward meaningful interpretation. We are only interested in whether an odd-ratio is between 0 and 1, meaning that odds that the dependent variable will occur if the dependent variable increases in value, will decrease, whether it is larger than one, meaning that the dependent variable’s occurrence becomes more likely if the independent variable’s value increases, or finally, when it is equal to one, meaning that occurrence of the dependent variable is unaffected by changes in the independent variable. The Netherlands has a relatively high ratio of initiatives to municipalities (one to four), the other countries have ratios lower than one. We should note here, again, that the differences in initiative-to-municipality ratios between the countries are going to affect the magnitude of the effects of any of the explanatory variables on the predicted chance of finding an initiative in a municipality.
The odd-ratio of the country dummy of any of the countries tells us that when we add a typical observation in our sample from one of these countries (the dummy value increases by one), the likelihood that this observation will also include an occurrence of an alternative currency organization decreases (odds lower than one, therefore
chances lower than 50/50). The odds will decrease the least when we add a Belgian observation, which makes sense as Belgium has the second highest initiative to municipality ratio of one to six.
Following social movement theory and the hypotheses by Collom (2005), we would expect that initiatives occur more often in more densely populated areas as well as in areas with higher unemployment levels. Our evidence is in line with Collom’s findings that economic marginality (high unemployment levels) have explanatory power in predicting alternative organization occurrence. While the odd ratios in table 3 suggest that population density is completely neutral (an odd-ratio of one indicates that an increase in population density would cause an increase the chance of an occurrence by 50-50 i.e., would not have predictive power). However, we should note here that the odd-ratios are significant larger than one in the cases of Spain, France and The Netherlands and insignificant negative in the cases of Belgium and Australia.
In table 4 we present the marginal effects resulting from the regressions, which provide an alternative interpretation of the same thing. The marginal effects are the predicted chances of finding an alternative payment system for different values of population density (1st quartile, median and 3rd quartile) as well as for unemployment rates, with the differences in percentage points in parentheses. We have calculated the marginal effects for all the performed logistic regressions, by changing the variables population density and unemployment separately, and to their first quartile, mean and third quartile values. The percentages shown show represent the chances of finding an alternative payment system given a certain values of population density or unemployment, keeping all other factors at their means.
We note that the unemployment rate has an unambiguous positive effect on the chance that an alternative system occurs, and that it is generally larger than the effect population density has. We should note here that unemployment and population density are only minorly correlated with each other by six percent. The changes in magnitudes of the marginal effects are again explained by the observation to municipality/statistical region ratios. The larger this ratio, the larger the marginal effect.
4.7.2 Estimation on country level
When we estimate the same model for each country separately, similar results emerge except for Australia. The estimated coefficients, p-values and the pseudo-R-squared, as shown in table 3 indicate that both unemployment and population density have no significant explanatory value in explaining alternative payment system occurrence. We should remind ourselves however, that the pseudo-R-squared is not equivalent to the R-squared reported in OLS regressions, and its values are usually lower. This could either mean that the hypothesis does not fit for Australia, or that something else is going on which could be caused by the type of statistics used for this estimation. As earlier noted, the statistical partitioning of Australia is different from the other countries as, for instance, a city is divided up into its city center and its surroundings, both being different statistical areas. This has also created a significant drop in the number of initiative occurrences included in this regression compared to the actual number of initiatives found. Furthermore, this specific statistical geography of Australia results in rather high mean and median density figures for the country, which are higher than one would expect them to be. This is caused by the fact that the statistical areas are aimed to have a more or less equivalent number of inhabitants, meaning that there will be a small number of very large areas with low densities, and numerous smaller areas with high densities. While we do not see signals of an upward bias caused by the rules used to connect the statistical areas to the postal code / city areas, as expected, we note that the coefficients for Australia are as expected but that the data probably suffers from inconsistencies for which should be corrected before any claims are made.
Figure 1: Alternative payment system occurrence across countries
Belgium Spain
Table 3: Regression results on initiative occurrence (Logit) (All
countries) (AUS) (BE) (FR) (SP) (NL) (NL-l) (NL-p)
Number of obs 47152 2086 589 36187 7928 362 361 361 Number of initiatives8 793 45 105 421 145 86 86 86 Pseudo R-squared 0.13 0.00 0.06 0.07 0.03 0.10 0.15 0.13 Constant 0.15*** (0.02) 0.02*** (0.01) 0.07*** (0.02) 0.00*** (0.00) 0.01*** (0.00) 0.03*** (0.01) 0.04*** (0.02) 0.03*** (0.02) Unemployment 1.09*** (0.01) 1.04 (0.04) 1.12*** (0.02) 1.09*** (0.01) 1.08*** (0.02) 1.43*** (0.11) 1.25*** (0.11) 1.27*** (0.11) Population Density 1.00*** (0.00) 1.00 (0.00) 1.00 (0.00) 1.00*** (0.00) 1.00*** (0.00) 1.00*** (0.00) 1.00*** (0.00) 1.00*** (0.00) Spain 0.06*** (0.01) France 0.03*** (0.00) Australia 0.06*** (0.01) Belgium 0.51*** (0.09) Left 2.56*** (0.80) Progressive 3.35*** (0.02) 8
These numbers represent the number of initiative occurrences that are used in the logit-regression. These are generally lower than the actual initiative count as reported in table 1 due to missing data for one of the independent variables, multiple initiatives in one region or incomplete data on the actual location of the initiative.
Table 4: Marginal effects of changes in density and unemployment values, percentage point changes in parentheses
(All countries) (AUS) (BE) (FR) (SP) (NL) (NL-p)
Mean density 219.23 1188.2 766.6 156.3 180.6 880.6 880.6 Median density 38.6 713.5 306 40 15.5 446.5 446.5 1st quartile 1.1% (-0.1%) 2.2% (+0.0%) 16.6% (+0.1%) 0.9% (-0.0%) 1.6% (-0.0%) 17.8% (-1.2%) 16.9% (-1.1%) Median 1.2% 2.2% 16.5% 0.9% 1.6% 19.0% 18.0% 3rd quartile 1.2% (+0.0%) 2.1% (-0.1%) 16.4% (-0.1%) 1.0% (+0.1%) 1.6% (+0.0%) 23.1% (+4.1%) 21.7% (+3.7%) Mean unemployment 9.0% 5.7% 9.0% 9.6% 7.6% 5.8% 5.8% Median unemployment 8.4% 5.1% 7.5% 8.9% 7.2% 5.6% 5.6% 1st quartile 0.9% (-0.2%) 2.0% (-0.1%) 11.2% (-3.1%) 0.8% (-0.1%) 1.4% (-0.2%) 15.0% (-5.3%) 16.0% (-3.4%) Median 1.1% 2.1% 14.3% 0.9% 1.6% 20.3% 19.4% 3rd quartile 1.5% (+0.4%) 2.2% (+0.1%) 20.5% (+6.2%) 1.2% (+0.3%) 2.0% (+0.4%) 28.5% (+8.2%) 24.3% (+4.9%)
4.7.3 Empirical estimation including political preference
Including the political dummies in the Dutch regression improves the fit of the model, judging from the pseudo R-squared figure. Furthermore, one should note that the odd-ratio of unemployment decreases, meaning that the political dummy takes over part of the effect. This indicates that the regression on the full set, as well as the other country-level regressions may suffer from some omitted variable bias. Because all-but-one political parties in The Netherlands are either right wing and conservative or left wing and progressive, our intuition says that including both dummies in one regression does not seem to make much sense. We note that the political categorization of progressive versus conservative seems to add the most value to the fit of the model. This basically tells us that voters for the Dutch “Partij voor de Vrijheid (PVV)”, which are considered to be left wing, do not fit our hypothesis well. If we look at table 4, our conclusions are similar as we see that the predictions of alternative system occurrences decrease both if population density and unemployment rates are high, compared to the estimation where the political dummy (Progressive) is excluded.
5 Conclusion and suggestion for further research
In this paper, we have made an attempt to add to the field of research to alternative currency organizations by gathering more in depth information for descriptive purposes, in order to make a step towards finding factors to determine organizational success, and to be able to expand on earlier work on environmental factors explaining the occurrence of these organizations. While we do not succeed in linking our rather rough measures of organizational success to any of the explanatory variables that follow from our survey, we were able to confirm a number of findings from Seyfang & Longhurst (2005) statistically, as we find clear cross-country differences in the responses to our survey. In addition to this, we also unexpectedly note subtle but important differences in the mindsets of participants in alternative currency organizations across countries.
We do succeed in linking alternative organization occurrence to unemployment figures, and find indication that areas with high unemployment are more likely to have such initiatives than areas with low unemployment. We also find population density to be a weaker predictor of the likelihood of the occurrence of such initiatives, which is likely to be due to data inconsistencies, especially for the case of Australia. Furthermore, we find that the regression results in the current setup may suffer from an omitted variable bias, as there may be multiple ways in which alternative payment system occurrence and unemployment data may be linked, for example political preferences.
While it is important to understand why alternative payment systems emerge and under which circumstances they persist, we have also made an effort in putting a number to the success of the initiatives by expressing the size of the currency in terms of a monetary equivalent. In its current definition, however, the measure appeared not fit as a measure of initiative success. Next from organizational success, however, putting a number on alternative currencies’ size is also interesting in a macro-economic context, where we are interested in unofficial macro-economic activity. While these systems may currently not be sizable enough for measurable economic effects, the uprise of these and other alternative payment systems such as crypto currencies may potentially change this in the future.
6 Reference list
Australian Bureau of Statistics. (2014, September). Data by Region. Retrieved from http://stat.abs.gov.au/itt/r.jsp?databyregion
Belastingdienst. (2008, July 12). Inkomstenbelasting. Resultaat uit overige werkzaamheden | Besluit | Rijksoverheid.nl. Retrieved from
http://www.rijksoverheid.nl/documenten-en- publicaties/besluiten/2008/12/17/inkomstenbelasting-resultaat-uit-overige-werkzaamheden.html
Belgian Federal Government (2013) Structuur van de bevolking volgens woonplaats - Statistieken & Analyses. Retrieved August 2014, from
http://statbel.fgov.be/nl/statistieken/cijfers/bevolking/structuur/woonplaats/
Brenes, E. (2011) Complementary currencies for sustainable local economies in Central America. International Journal of Community Currency Research. 15 Special Issue D32-38
CBS StatLine. (2014). Regionale Kerncijfers Nederland. Retrieved from http://statline.cbs.nl/
Collom, E. (2005). Community currency in the United States: the social environments in which it emerges and survives. Environment and Planning A. doi:10.1068/a37172
CES. (2014, July). Community Exchange System Australia - Online trading system for Australian LETS groups & Community Currencies. Retrieved from
http://communityexchange.net.au/
Dini, Paolo (2012) Community currencies and the quantification of social value in the digital economy. The London School of Economics and Political Science, London, UK. (Unpublished)
Fitzpatrick, T., Caldwell, C. (2001) Towards a Theory of Ecosocial Welfare: Radical Reformism and Local Exchanges and Trading Systems (LETS). Environmental Politics, 10-2 pp.43-67. Retrieved from: http://dx.doi.org/10.1080/714000532
Hirschman, A. O. (1970) Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States. Harvard University Press, Cambridge, MA
Institut national de la statistique et des études économiques (INSEE): Accueil. (2014). Conditions de vie – Société. Retrieved August 2014, from
http://www.insee.fr/fr/themes/theme.asp?theme=5
Letscontact Nederland. (2014, July). Letskringen gesorteerd op Alfabet. Retrieved from http://www.letscontact.nl
Lets Vlaanderen. (2011). Lets groepen in Vlaanderen. Retrieved July 2014, from http://www.letsvlaanderen.be/Groepen/tabid/147/Default.aspx
Liesch, P. W., Birch, D. (2000) Community-based LETSystems in Australia: localised barter in a sophisticated Western economy. International Journal of Community Currency Research. 4 Retrieved from:
http://www.geog.le.ac.uk/ijccr/vol4-6/4toc.htm
Mascornick, J. (2005) Local Currency Loans and Grants: Comparative Case Studies of Ithaca HOURS and Calgary Dollars. International Journal of Community Currency Research. 11 pp.1-22
Non-Profit Data. (2014). Werkloosheid naar leeftijd en geslacht in elke gemeente. Retrieved August 2014, from
http://www.npdata.be/BuG/212-Werkloosheid/Werkloosheid-per-gemeente.htm
North, P. (2005) Scaling Alternative Economic Practices? Some Lessons from Alternative Currencies. Transactions of the Institute of British Geographers. 30-2, pp.221-233.
