You are on page 1of 31

Direct selling during economic

downturn

T. Span, M.Sc.
dr. J.M.P. de Kok
Zoetermeer, July 2013

This research has been financed by Herbalife.

Panteia BV

Panteia BV

Bredewater 26

P.o. box 7001

2715 CA Zoetermeer

2701 AA Zoetermeer

079 322 22 00

The Netherlands

www.panteia.nl

+31 79 322 22 00

The responsibility for the contents of this report lies with Panteia. Quoting numbers or text
in papers, essays and books is permitted only when the source is clearly mentioned. No part
of this publication may be copied and/or published in any form or by any means, or stored
in a retrieval system, without the prior written permission of Panteia. Panteia does not
accept responsibility for printing errors and/or other imperfections.

Contents

Executive summary

Introduction

Why direct selling may be related to unemployment

2.1

A closer look at direct selling

2.2

Determinants of direct selling

10

2.3

Model specification

12

Research methodology

13

3.1

Data

13

3.2

Estimation methodology

14

Results

15

4.1

Descriptive statistics

15

4.2

Multivariate analysis results

17

Discussion and conclusions

21

5.1

Discussion of main results

21

5.2

Limitations and future research

22

References

25

Annexes
I

Methodology of literature review

II

List of search terms used

27
29

III

Country descriptive statistics

31

Executive summary
L a c k of s c ie n t i f i c a tt e n t i o n f o r d ir e c t s e ll i n g
During the period 2002-2011, an ever increasing share of the labour force in
European countries participated in direct selling activities, by selling consumer
goods either door-to-door, at home or at organised parties. The economic relevance of direct selling is demonstrated by the total sales volume. In the European Union alone, the total sales volume has increased from 10.7 to 11.6 billion, representing a growth rate of 8.1% in 2010. Despite the potential economic
relevance of this sales channel, it has received little attention in academic literature. Hardly any scientific publications are available that examine determinants
of the choice of individuals to become active in direct selling (at micro level
and/or macro level), the income effects of direct selling for the direct sellers, or
macro economic effects of direct selling.
The objective of this study is to provide a first exploration of macro-economic
determinants of the direct seller rate. In particular, we have examined the role
of the unemployment rate as a determinant of the direct seller rate. In addition,
we tried to determine if the data supports the notion that direct selling can be
seen as a manifestation of entrepreneurship.
D i r e c t s e l l i n g a s a m an i f e s t a t i o n o f e n t r e p re n e u r s hi p
The point of departure for our study was the premise that direct selling can be
seen as a manifestation of entrepreneurship. In particular, in times of economic
downturn, it may be interpreted as a specific form of necessity-based entrepreneurship. This suggests a positive relationship between the unemployment rate
and the direct seller rate: a higher unemployment rate in a certain year is expected to have a positive effect on the direct seller rate in the following year.
Loss of income (either because the loss of a job or for other reasons) would drive
people to take matters in their own hands and generate income by entering a direct selling scheme.
A positive relationship between the unemployment rate and the direct seller rate
does, however, not necessarily imply that direct selling is indeed a manifestation
of entrepreneurship. To the direct sellers, direct selling may be seen primarily as
a secondary job rather than an entrepreneurial activity. These people may not
necessarily consider themselves to be entrepreneurs. We have therefore included
entrepreneurship indicators in our model, so that we can see to which extent the
data supports the notion that direct selling can be seen as a manifestation of entrepreneurship.
M e t h o do l o gy
We have collected data on direct selling characteristics in 28 countries for ten
consecutive years (2002-2011). In addition, we collected data on potential
macro-economic determinants of the direct seller rate such as the lagged unemployment rate, entrepreneurship rates, income levels, population density and the
second-job rate. Variables indicating the country and year of each observation
were included to control for additional factors that may affect the direct seller
rate and to allow for the use of panel data estimation techniques. We have used
these data to estimate a model where the direct seller rate is explained by the
unemployment rate and the other potential determinants.

U n e m pl oy m e n t h a s a l on g - t e r m c um u l a t i ve i mp a c t
In a set of increasingly expansive multivariate analyses, we examined the relation between unemployment and direct selling. A first finding was that these
multivariate analyses provided no support for a relationship between the direct
seller rate on the one hand, and the unemployment rate of the previous year on
the other hand (controlling for other variables and unobserved country differences).
It is however conceivable that the effect of changes in the unemployment rate
take a few years to materialize. We have therefore examined the possibility of a
long-run impact on the direct seller rate and found statistical support for the
presence of such a long-term relationship: our results suggest that, over a
three-year period, a one percentage point increase in the unemployment rate
has a cumulative effect of increasing the direct seller rate by 0.31 percentage
point. This effect is considerable in size given the 2.1% average direct seller rate
in our sample.
I m p a c t i s so m e w h a t g re a t e r f or c ou n t r ie s w it h l o w e r i n c o me le ve l s
The relationship between unemployment rates and direct seller rates could also
depend on the welfare level of a country. The data suggests that this is indeed
the case: the relationship between unemployment rate and direct seller rate becomes weaker when GDP per capita rises. This suggests that the positive impact
of a higher unemployment rate on the direct seller rate is stronger for countries
with lower per capita income. The differences between the countries are, however, small: the parameters range from 0.23 for the country with the highest
GDP/capita level in our sample (Luxembourg) to 0.25 for the country with the
lowest GDP/capita level in our sample (Romania).
D i r e c t s e l l i n g i s n o t a m a n i f e s t a t i o n o f e n t r e pr e n e u r s h i p
The point of departure for our study was to view direct selling as a form of entrepreneurship. While a positive correlation between necessity-based entrepreneurship and the direct seller rate does exist, none of our multivariate analyses
return a significant relationship between entrepreneurship indicators and the direct seller rate. Hence, our findings do not support this view. It seems that people engaged in direct selling do not see themselves as entrepreneurs and do not
register themselves as such, but rather see themselves as being engaged in a
secondary job to amend their income.

Introduction
B a c k g r ou n d
An ever increasing number of people earn income by selling consumer goods
door-to-door, at home and at organised parties. The emergence of direct selling,
as this distributional form is called, has been stirring up the entire retail sector.
The economic relevance of direct selling is demonstrated by the total sales volume. In the European Union the total sales volume has increased from 10.7 to
11.6 billion, representing a growth rate of 8.1% in 2010 (Seldia, 2011). This
shows that direct selling continues to perform strongly during the global economic crisis. Nevertheless, despite the fact that direct selling can be considered
an emerging economic phenomenon, it has received little attention so far in the
academic world (Stanworth et al., 2004). Hardly any scientific publications are
available that examine determinants of the choice of individuals to become active
in direct selling (at micro level and/or macro level), the income effects of direct
selling for the direct sellers, or macro economic effects of direct selling.
O b j e c t i ve of t h i s s t u d y
Herbalife, a global nutrition company whose products are sold through direct
sellers, commissioned Panteia to analyse data on the direct selling industry in 28
European countries over ten consecutive years (2002-2011). Panteia was tasked
with modelling the potential macro-economic determinants of the direct seller
rate, such as the lagged unemployment rate, entrepreneurship rates, income
levels, population density and the second-job rate. The outcomes of this study
have been used to inform a new research paper from Herbalife, which concentrates on public attitudes to a changing and more flexible European labour market1.
R e s e a r c h me t h o d o l o g y
An extensive literature search has been conducted to identify existing scientific
literature on the determinants of direct selling (see annexes I and II for the
methodology and search terms used). The results of this literature search confirmed the lack of attention: no empirical scientific studies have been identified.
For the empirical part of our study, we compiled a dataset of direct selling characteristics in 28 countries for a ten-year period and included data on macroeconomic variables that may affect the direct seller rate. The longitudinal nature
of our data allows us to use panel data estimators in determining the relationship
between lagged unemployment and direct selling.
Paper outline
In the next chapter we introduce the concept of direct selling, discuss the theoretical reasoning behind the determinants included in our analyses and present
the model we wish to estimate. The third chapter discusses the data and our estimation methodology. The fourth chapter contains the results of both descriptive
and multivariate analyses. Finally, the fifth chapter concludes our study with the
most important findings and a discussion of both our results and the limitations
of our paper.

This report is available at http://company.herbalife.co.uk/second-income-society-report

Why direct selling may be related to unemployment

2.1

A closer look at direct selling


Direct selling can be defined as the marketing of consumer goods and services
directly to consumers on a person-to-person basis, generally in their homes, or
the homes of others, at their workplace and other places away from permanent
retail location. (the Federation of European Direct Selling Associations, 2003).
The direct sales sector mainly consists of the vending of cosmetics, cleaning
products, kitchenware and health products. There are different methods for the
direct sales of goods. Person-to-person selling involves the promotion of products in private homes, at work or on an individual basis. Alternatively, salespersons can sell their goods through organised parties, at which they demonstrate
the products to a group of people (Brodie, 1999). In the majority of cases, direct
selling is a part-time activity and mostly performed by women (Abilio, 2011).
More data on direct sellers will be discussed in the next chapter.
Different organisational structures can be distinguished among the direct selling
companies. Direct sales firms that use a single-level distribution and compensation system are characterised by a horizontal structure. These companies are in
direct contact with their salespersons and pay them for the performance delivered. Multi-level direct selling companies, however, make use of a hierarchical
employment structure. In addition to the sales of goods, independent salespersons are responsible for the recruitment and management of other direct sellers
and receive a commission on their work (Brodie, 1999).
T h e p o s i t i o n o f t h e d i r e c t s e l le rs
The growth of the direct sales industry is related to an underlying trend: the outsourcing of services by large companies (Stanworth et al., 2004). Likewise, direct selling companies do not hire their salespersons as staff, but instead offer
contractual relationships that individuals can enter in order to gain access to an
established business format, in return for performance-related remuneration
(Stanworth et al., 2004:51). Formally speaking, this makes the direct sellers
self-employed. From a labour market perspective, direct sellers can be considered a modern type of sales representatives (in paid employment), whose income is completely determined by performance and no formal employeremployee contract is signed. Alternatively, direct selling can be regarded as a
manifestation of entrepreneurship. Entrepreneurship is not a dichotomous variable 1, but instead many different types of entrepreneurship exist, including intrapreneurship (where employees introduce and develop new and innovative
ideas) and self-employment. Since direct selling allows people to run and manage their own business and further develop their entrepreneurial abilities, a case
can be made that direct selling is also a form of entrepreneurship.

Implying that one is either a business owner or not.

M o t iv e s t o s t a r t w it h d i r e c t s e l l i n g
Due to the dualistic nature of the direct selling business, people reveal different
motives to enter the sector. Overall, those employed in a multi-level scheme can
be considered career-minded entrepreneurs, as they commit to a greater
amount of responsibility and time and have to exert higher managerial and administrative efforts. The multi-level direct sellers more often report a positive
reason for entering the business (Stanworth et al., 2004). They regard it as an
opportunity to further develop themselves and their career. By contrast, salespersons employed in the single-level direct selling companies will more frequently report push-related factors as their motivation (Stanworth et al., 2004).
They are forced into the business from a situation of necessity (i.e. unemployment). Furthermore, single-level sellers more often work on a part-time basis
and are more likely to view it as a secondary source of income. Other motivations reported to enter the direct selling business are the desire for an increased
flexibility of work and change in direction of career (Stanworth et al., 2004).

2.2

Determinants of direct selling


An extensive literature search reveals that hardly any academic literature exists
on the subject of direct selling and the factors that determine the direct seller
rate in a country. Our literature review has identified only three papers that discuss a relationship between the number of persons engaged in direct selling and
unemployment rates (Welsh and Pendleton, 2006; Abilio, 2011; De Casanova,
2011). None of these papers investigate this hypothesis empirically, nor did we
find any other papers that do so. The same holds for the influence of any other
factors on the direct seller rate. Our study contributes to the scarce existing literature by empirically investigating the influence of potential determinants of the
direct seller rate, in particular the unemployment rate. Our study can therefore
be considered an explorative one. This section discusses the potential determinants of the direct seller rate we choose to consider.
U n e m pl oy m e n t r a t e a n d e n tr e pr e n e u r s h i p
The point of departure for our study is the premise that direct selling can be
seen as a manifestation of entrepreneurship. In particular, in times of economic
downturn, it may be interpreted as a specific form of necessity-based entrepreneurship. This is underlined by the fact that many single-level direct sellers report push-related reasons for entering the direct selling business. An increase in
the unemployment rate, and loss of job at the individual level, will induce people
to search for alternative sources of income. Direct selling can provide people
with such a source of livelihood. Necessity-based entrepreneurship (in contrast
to opportunity-based entrepreneurship) may therefore act as a transmission
channel from unemployment to direct selling.
The entrepreneurial attitude of a countrys workforce is liable to affect the operation of this transmission channel. An indicator of the overall entrepreneurship
rate in a country is therefore also included, as both a proxy of entrepreneurial
attitude and as an alternative measure of entrepreneurial activity.
D i r e c t s e l lin g a s s o ur c e o f in c o me
Direct selling provides a source of income, either primary or secondary. In both
of these cases, increased unemployment rates may increase direct seller rates:
for those that are (recently) unemployed, direct selling may become a primary
source of income, for those that are not unemployed the wish for additional in-

10

come may increase, e.g. because of reduced income levels in their primary income source, unemployment of partner, etc.
The income generated through direct selling is likely to be affected by the position of a direct seller in a direct selling scheme. Income levels will be highest for
those employed at higher levels in a multi-level scheme; for those employed at
the lowest level of such a scheme and for single-level direct sellers, the additional income generated will be relatively low 1. Due to a lack of data, we cannot
examine this assumption.
In this argument, we assume that increases in the unemployment rate may increase the number of people active in direct selling out of necessity. This does
not necessarily translate into greater rates of (necessity-based) entrepreneurship. Depending on how entrepreneurship is measured, entrepreneurship levels
only increase if the people that start with direct selling activities register themselves as self-employed worker or business owner2, or if they consider themselves as being involved in setting up a new business 3. If the people involved
consider their direct selling activities as a secondary job to supplement their income, they may not register themselves as entrepreneurs4 and may not report
themselves as being involved in entrepreneurial activities. In this case, the link
between the direct sellers rate and the share of entrepreneurs/business owners
may be weak. We therefore need to differentiate between necessity-based activities and necessity-based entrepreneurship.
The role of direct selling as secondary source of income suggests that the direct
seller rate in countries may be correlated with the second-job rate. We therefore
include the second-job rate in our model as a proxy for the willingness of a
population to take on a secondary job.
G D P p e r c ap i t a
Generally speaking, direct selling is often treated as a mediocre source of income
that is mainly seen as a means for additional income (either a secondary source
of income or a primary source of income if other sources are not available). If
this is indeed the general opinion of a countrys labour force, a negative relationship between a countrys welfare level and the direct seller rate may be expected: increasing welfare levels may indicate an increase in the number of wellpaid jobs that are more highly valued than direct selling activities, in which case
less people may take up direct selling. In addition, to the extent that higher welfare levels are associated with higher unemployment benefits and social welfare
benefits, this also implies a negative correlation between a countrys welfare and

Even if income levels are low, they may still count as the primary source of income for the direct
seller, as long as it is the only source of income. This may be the case for the majority of direct
sellers (women working part-time).

The entrepreneurship measure provided by Eurostat is based on the number of self-employed


workers and/or the number of established business owners (see also chapter Three).

The Global Entrepreneurship Monitor (GEM) uses surveys to measure nascent entrepreneurship
and early-stage entrepreneurial activities. Basically, these measures are based on questions
whether the respondent is currently involved in setting up a business or has recently set up a
business (see also chapter Three).

Formal requirements are likely to differ between countries. We have not examined whether direct
sellers are required to register as independent entrepreneur in all of the countries involved in
this study.

11

the direct sellers rate1. In this study we use GDP/capita as a measure of a countrys welfare level.
Population density
Finally, we expect the population density to affect direct selling, specifically in its
door-to-door form. In more densely populated areas (i.e. with a greater degree
of urbanisation), it is possible to reach a greater potential clientele in a shorter
amount of time. On the other hand, direct selling may prove to be an instrument
in bringing certain specific goods to the more scarcely populated areas that are
not sold there through traditional retail channels. The direction of the relationship thus remains an empirical question.

2.3

Model specification
The main relation of interest in our study is that between the unemployment rate
and the direct seller rate. In addition, our model includes multiple control variables to control for other factors influencing the direct seller rate in a country.
The model we estimate takes the following form:

Direct seller rateit X it Unemployment rateit 1 vt u i it


In this specification, the subscript i denotes the countries and t indexes the
years; X is a set of explanatory variables; the one-year lag of the unemployment
rate is used to allow for temporal shifts in the direct seller rate; v t is a year-fixed
effect that consists of a set of dummies for each but the basis year 2002; u i is a
country-fixed effect that consists of dummies for all but one country; and it is a
random idiosyncratic error term, consisting of other unobserved factors that influence the direct seller rate. The set of control factors is comprised of the determinants of direct selling discussed earlier.

E f f e c t o f u n e m p l o y me n t r a t e me d i a t e d b y e n t r e p re n e u r s h i p ?
In case the unemployment rate proves to be a primary determinant of the direct
seller rate, we will examine to which extent the impact of unemployment on direct selling runs via necessity-based entrepreneurship: increased unemployment
levels increase the level of necessity-based entrepreneurship, which in turn affects the direct seller rate. Necessity-based entrepreneurship is then said to mediate the relationship between unemployment rate and the direct seller rate.
For necessity-based entrepreneurship to qualify as a mediating factor, it must
adhere to two prerequisites: first, the relationship between the unemployment
rate and necessity-based entrepreneurship must be statistically different from
zero and second, the relationship between necessity-based entrepreneurship and
the direct seller rate must be statistically different from zero.

12

We expect that this relationship may be stronger between countries than within countries over
time.

Research methodology

In this chapter we discuss the data that we use and the applied estimation techniques.

3.1

Data
D i re c t s e l le r r a t e
The direct seller rate is defined as the share of the labour force active in direct
selling. It is calculated by dividing the number of direct sellers by the labour
force (which consists of the sum of employed and unemployed persons). Labour
force statistics are collected in the quarterly Labour Force Survey (LFS) and obtained from Eurostat.
The required direct selling data are collected and published by Seldia, the European direct selling association. The most recent annual data are presented on
their website 1. Seldia reports on the number of people engaged in direct selling
in a country, as well as on the methods of direct selling (person-to-person or
party plan) and the shares of female and part-time direct sellers. Seldia also reports sales figures of the direct selling force, but these values are highly inconsistent over time and therefore not included. Data on single- and multilevel direct selling schemes are not reported. For this study, we have obtained data for
the ten-year period from 2002 to 2011.
The Seldia data mostly refer to the entire direct selling industry in a country, including both members of the national Direct Selling Association (DSA) and nonmembers. For some countries, observations are presented on DSA members
only. For yet other countries, the reported data are research estimates by the
World Federation of Direct Selling Associations (WFDSA).
The countries included in our dataset have been selected based on the availability of direct selling data for the period of interest. This resulted in a sample consisting of 28 countries. Data is available for every single year in 25 of these
countries. For the remainder, data were available for nine out of the ten years.
The missing years for these cases are either 2002 or 2011.
U n e m pl oy m e n t r a t e s
National unemployment rates are obtained from the Labour Force Survey (LFS)
from Eurostat. The unemployed population is defined as consisting of those persons 15 to 74 years of age who where not employed during the reference week,
but did actively seek for employment and were available immediately to begin
working. The unemployment rate relates these persons to the labour force. We
incorporate a one-year lag of the unemployment rate to allow for temporal adjustments in the direct seller rate and control for the issue of endogeneity (direct
selling also affects the unemployment rate).

http://www.seldia.eu/statistics

13

C o n t r o l v a r ia b l e s
An additional set of control variables is collected. The following are also obtained
from Eurostat. The population density is defined as the absolute value of the average population per square kilometre in a country. The entrepreneurship rate
originates in the LFS and relates the number of self-employed in a country to the
size of its labour force1. The second-job rate relates the number of people with a
second job to the labour force. Furthermore, data on GDP per capita is obtained
from UNECE. Population density and GDP per capita are included in logarithmic
form to account for their non-normal distributions.
Finally, entrepreneurship variables are obtained from the Global Entrepreneurship Monitor (GEM) database. The GEM is an annual international survey on entrepreneurship in over 85 countries, taken from 1999 onward. It includes various
entrepreneurship indicators, including indicators that differentiate between necessity-based and opportunity-based new entrepreneurship (new entrepreneurship includes nascent entrepreneurs and early-stage entrepreneurial activity, the
owning or managing of a business no older than 3.5 years). In contrast to other
ratios, this figure is related to the total population (i.e. not only to the labour
force) and to a different age group, namely those aged 18 to 64.
All variables have been checked and controlled for outliers.

3.2

Estimation methodology
Our dataset follows the same set of countries over a ten-year period. The longitudinal characteristics of the data allow for the use of panel data estimators of
the relationship between unemployment and the share of direct sellers in the labour force. The panel data structure permits us to include lagged terms to inspect the effects of temporal shifts in the direct seller rate and also increases the
number of observations, and thus the precision of our estimates.
We analyse the specified model by applying fixed effects estimation. This approach allows us to specifically control for those variables omitted in our model
specification that remain constant over time, by accounting for the timeinvariant part of the error term: the fixed effect. We estimate a pooled ordinary
least squares regression and include country and year dummies.
The models are estimated with standard errors clustered by country to account
for autocorrelation in the errors and for heteroskedasticity. Clustered standard
errors account for the possibility that the standard errors are independent between countries, but may be correlated within countries. Clustering increases the
standard errors as it allows for less variation. Hence, it also suppresses statistical significance of the results.

14

The number of self-employed workers (and/or the number of established business owners) is a
common indicator of entrepreneurship, but it represents a static view on entrepreneurship. More
and more, researchers use a more dynamic perspective and use indicators on the start-up rate of
new enterprises and/or the share of young enterprises to measure entrepreneurship (as is the
case with the GEM entrepreneurship indicators that are also included).

Results

This chapter presents the results from our analyses. It is broken down into two
distinct sections. The first part presents descriptive statistics on the collected
data, including simple summary statistics and a correlation matrix which provides a first indication of the relations between the variables in our model. The
second part presents the actual results of different specifications of our fixed effects estimation.

4.1

Descriptive statistics
Table 1 presents summary statistics on the direct selling data and main explanatory variables in our model for the total sample of 28 countries in the year 2010.
Mean and median values are given, accompanied by standard deviations to allow
for a more complete view of the distribution of the variables.
Table 1 Summary statistics for 2010
Mean

Median

Standard
Deviation

Direct selling characteristics


Direct seller rate

Share of female direct sellers

Share of part-time direct sellers


Second-job rate

2.1

1.7

1.5

79%

80%

0.1

78%

86%

0.2

3.8

3.4

2.1

Macro-economic characteristics
Unemployment rate
Population density

b c

GDP per capita in $US

9.6

8.4

4.2

128.8

102.2

105.1

26.4

26.4

15.0

13.3

12.4

5.1

1.1

1.1

0.7

3.9

4.0

1.2

Entrepreneurship characteristics
Entrepreneurship rate

Necessity-based entrepreneurship rate

Opportunity-based entrepreneurship rate


Source: a Seldia

Eurostat c ILO

UNECE

GEM

The figures presented in the table show that on average, 2% of the labour force
in a country is engaged in direct selling. The majority of direct sellers is female
and works on a part-time basis.
There exists a great range in the values when comparing the countries in our
dataset. The direct seller rate for instance, is at its lowest at only 0.2% in Switzerland versus 5.8% of the labour force in Russia. The characteristics of the direct selling labour force also greatly differ from country to country, with some
strongly deviating from the mean values of the full sample presented earlier. The
appendix to this study contains a more detailed table with summary statistics by
country (Table 5 in Annex 3)

15

A first indication of the relations that exist between variables is given in Table 2
which presents Pearsons correlation coefficients for all of the variables relevant
to our model. These correlation coefficients provide an indication of the strength
of the relation between two variables. No relationship exists when the correlation
is equal to zero and a perfect linear relationship exists when it is equal to +/-1.
The greater the (absolute value of the) coefficient, the stronger the relationship.

Unemployment rate

0.23

**

Population density

-0.38

**

-0.21

**

GDP per capita

-0.41

**

-0.52

**

0.24

Second-job rate

-0.11

-0.28

**

-0.00

0.23

Entrepreneurship
rate

0.02

0.21

**

-0.02

-0.36

**

-0.26

**

Necessity-based
entrepreneurship

0.30

0.41

**

-0.25

**

-0.43

**

-0.39

**

-0.21

**

0.48

Opportunity-based
entrepreneurship

Necessity-based
entrepreneurship

Entrepreneurship
rate

Second-job rate

GDP per capita

Population density

Unemployment rate

Pooled correlation matrix

Direct seller rate

Table 2

**

-0.04

-0.33

**

**

**

**

0.21

**

0.40

**

-0.07

0.13

Source: Panteia, 2013.


**

correlation is significant at the 1% level (2-tailed)

The signs on the coefficients mark the directions of the relations. When the correlation between variables is positive, a higher value for one variable is associated with a higher value for another variable. For a negative correlation, a higher
value for one variable is associated with a lower value for another variable. The
signs of the correlations are therefore of first interest. Notice that a correlation
represents a relationship, but not necessarily a causal relationship (let alone that
it shows the direction of the causality).
The coefficients presented in the leftmost column provide an indication of the relation between the direct seller rate and the explanatory variables. Most relations
are as expected: there is a positive relation between the unemployment rate and
the direct seller rate. The same holds for necessity-based entrepreneurship,
which is significantly positive correlated with the direct seller rate. Furthermore,
a lower income seems to correspond to a higher direct seller rate. The other relations do not agree with our intuition. Higher population density was expected
to make direct selling easier, but the table indicates a negative relation. This
also holds for the second-job rate.
These results must be interpreted carefully. For instance, none of the correlations between the direct seller rate and the explanatory variables are particularly
strong. A cut-off point of 0.5 is often used to distinguish between weak and
strong correlation. Not all of the coefficients are statistically significant. The co-

16

efficient on the second-job rate may seem surprising, but its statistically small
effect provides a first indication that a (strong) relation between second jobs and
direct selling does not exist. Furthermore, since the presented coefficients apply
to the full sample, it is impossible to discern the effects of correlation within and
correlation between countries.
The matrix also shows that the correlation amongst the explanatory variables is
not particularly strong in most cases. Multicollinearity (high correlation between
the explanatory variables) is therefore not likely to be an issue (multicollinearity
leads to great variance in the model estimators, making the estimated effects
less accurate).

4.2

Multivariate analysis results


In this section, we will present the outcomes of our multivariate regressions. The
multivariate nature of these estimations may lead to different findings than those
suggested by the correlation matrix. Table 3 presents the outcomes of six increasingly expansive specifications of our model. The t-statistics are presented in
parentheses underneath the coefficient estimates.
The first column of the table presents model (1), which is the most straightforward of our estimations. It is a bivariate linear regression model that relates the
direct seller rate to the one-year lagged unemployment rate. The coefficient estimates confirm the relation suggested by the correlation matrix. Lagged unemployment is positively related to the direct seller rate: a one percentage point
increase in the unemployment rate in the previous year is associated with an increase in the direct seller rate by 0.11 percentage point. This effect is statistically significant. The explanatory power of the model as exemplified by its R 2,
however, is very low. Only 6% of the variation in the direct seller rate is explained by variation in the unemployment rate. Finally, a non-linear relationship
has been tested for by introducing a quadratic function of the unemployment
rate, but no such relation was found.
The second column adds country- and year-fixed effects to model (1) in the form
of dummies. These drastically raise the explanatory power of the model which
now explains 83% of the variation in the direct seller rate. Most of this increase
can be attributed to the country dummies, suggesting that there is an important
country-specific factor that influences the direct seller rate more strongly than
unemployment. The significantly positive relationship between unemployment
and the direct seller rate, however, has now completely dissipated. Its sign has
become negative, but this effect does not differ from zero significantly.

17

Table 3

Results of different pooled OLS regression analysis specifications

Variable

Direct seller rate


(1)

Constant

Unemployment rate t-1

(2)

(3)

(4)

(5)

(6)

1.12 *

3.78 **

1.52

14.38

4.02

12.66

(2.38)

(11.68)

(0.33)

(1.04)

(0.58)

(1.62)

0.11 *

-0.01

0.01

0.08

-0.06

0.27 *

(2.45)

(-0.30)

(0.10)

(1.95)

(-0.73)

(2.20)

Population density log

GDP per capita log

Second-job rate

Entrepreneurship rate

-0.46

0.12

0.24

0.14

(-0.81)

(0.23)

(0.64)

(0.30)

1.76

-3.66

-1.01

-3.38

(1.16)

(-1.19)

(-0.40)

(-1.08)

-0.16

-0.17

-0.36 *

-0.20

(-1.28)

(-1.02)

(-2.34)

(-1.27)

-0.03
(-0.53)

Necessity-based
entrepreneurship rate
Opportunity-based
entrepreneurship rate

0.10

0.38

0.12

(1.15)

(1.88)

(1.30)

-0.02

-0.03

-0.03

(-0.19)

(-0.49)

(-0.45)

Unemployment rate t-2

0.19
(1.13)

Unemployment rate t-3

0.19
(1.22)

Unemployment rate t-1

-0.01 *

* GDP per capita

(-2.32)

Country-fixed effects

No

Yes

Yes

Yes

Yes

Yes

Year-fixed effects

No

Yes

Yes

Yes

Yes

Yes

0.06

0.83

0.87

0.88

0.92

0.88

269

269

236

159

123

159

Observations
Source: Panteia, 2013.
*

coefficient is significant at the 5% level

**

coefficient is significant at the 1% level

The next two columns present specifications (3) and (4). Both expand upon
model (1) by including additional explanatory variables. Both models include the
logarithm of population density, the logarithm of GDP per capita (in $1,000) and
the second-job rate. In addition, model (3) includes the entrepreneurship rate
while model (4) includes indicators for necessity-based and opportunity-based
entrepreneurship. The results of model (3) show some surprising results for the
additional explanatory variables that are counterintuitive. Their signs have
changed in model (4), but each of the effects is not statistically significant. Of
specific interest are the entrepreneurship indicators. While none of their effects
is statistically different from zero, model (4) does hint that necessity-based and
opportunity-based entrepreneurship may have opposing effects on the direct
seller rate.

18

Model (5) adds two additional time-lags on the unemployment rate to account for
the possibility that changes in unemployment do not affect the direct seller rate
immediately in the following year. The signs on these lags are as expected, but
their inclusion changes the sign on the one-year lag. Their individual effects do
not differ from zero significantly. However, when the three lags are taken together, the estimated finite distributed lag model can be written as follows:
n

Direct seller rateit X it Unemployment rateit 1 vt u i it


s 3

The depth of the lags is denoted by s. This model can be interpreted as before,
but with one key difference. The estimated coefficients of the three unemployment lags will be summed to provide one cumulative estimate. This long-run effect is equal to 0.31. This is called the long-run propensity or multiplier and
gives the effect of changes in the unemployment rate on the long-term equilibrium level of the direct seller rate. The combined t-statistic is 2.37, so that this
cumulative effect is statistically significant.
The cumulative coefficient on the lagged unemployment terms indicates that a
one percentage point increase in the unemployment rate, has a long-run effect
of raising the direct seller rate by 0.31 percentage point. This is a considerable
effect, given that the average direct seller rate for our sample is 2.1%.
The final column presents model (6), which drops the additional lag terms and
introduces an interaction term between the one-year lagged unemployment rate
and GDP per capita. The idea is that unemployment might have a different impact on the direct seller rate for countries with higher levels of per capita income. Including the interaction term changes the interpretation of the existing
variables: the coefficient estimate on the lagged unemployment rate, for instance, now measures its influence on the direct seller rate when GDP per capita
equals zero. The negative sign on the interaction term indicates than an increase
in the unemployment rate leads to a lower increase in the direct seller rate when
GDP per capita is higher. In other words, the relation between unemployment
and the direct seller rate is weaker in richer countries.
M e d i a t i n g e f f e c t o f n e c e s si ty - b a s e d e n tr e pr e n e u r s h i p
In Chapter Two we discussed the possibility that necessity-based entrepreneurship might mediate the relationship between unemployment rate and direct seller
rate. This would be the case if the impact of unemployment on direct selling runs
via necessity-based entrepreneurship. For necessity-based entrepreneurship to
qualify as a mediating factor, it must adhere to two prerequisites: first, the impact of unemployment on necessity-based entrepreneurship must be statistically
different from zero and second, the impact of necessity-based entrepreneurship
on the direct seller rate must be statistically different from zero.
According to the results of models 4 to 6, the second condition is not met. In an
unreported analysis we test for the first condition by repeating the models in
Table 3, this time with necessity-based entrepreneurship as the dependent variable. In all of the models in which we include country- and year-fixed effects, the
impact of the unemployment rate on necessity-based entrepreneurship is not
statistically different from zero. Hence, we do not find evidence to support our

19

premise that necessity-based entrepreneurship mediates the relation between


unemployment and direct selling.

20

Discussion and conclusions

We finalise our study with a discussion of our most important findings, followed
by a discussion of the main limitations and suggestions for future research.

5.1

Discussion of main results


C or re l a t i on b e t w e e n u n e m p lo y me n t r a te a n d di r e c t s e l l e r r a t e r e f l e c t s
c o u n t r y d if fe re n c e s r a t h e r t h an c au s a l r e l a t io n s h i p
A significantly positive correlation is found between unemployment rate and direct seller rate. The most basic of our multivariate models confirms this relationship. However, once we control for country and year effects, the relationship disappears (model 2). This implies that the correlation between unemployment rate
and direct seller rate does not indicate a causal effect of unemployment on the
direct seller rate (that is identical for all countries and years involved). Instead,
the correlation is due to omitted variable bias: there are other (countryspecific) variables that affect both unemployment rate and direct seller rate,
which in turn results in the correlation that is present. This shows the importance
of multivariate analysis: correlations cannot reveal the presence of multivariate
relations.

E f f e c t o f u n e m p lo y me n t r a te o n s e l l e r r a te t ak e s a fe w ye a r s t o m a n i fest itself
Additional analysis shows that there is a relationship between the unemployment
rate and the direct seller rate, which however takes (at least) three years to
manifest itself. Our estimations show that over a three-year period, a 1%point
increase in unemployment has a long-term effect of 0.31% higher direct selling
rate (Model 5).
An example can illustrate what this means. Suppose that the unemployment rate
in a certain country in, say, 2008, was 8.4% and that the direct seller rate was
1.7% (the median values). Next, suppose that the unemployment rate increases
with 2%point (from 8.4% to 10.4%). Our results suggest that this will decrease
the direct seller rate in 2009 with 0.12%point, followed by an increase with
0.38%point in 2010 and 0.38%point in 2011. The overall effect in 2011 is an increase of the direct seller rate with 0.62%point (from the median value of 1.7%
to 2.3% - an increase of 35%).
The parameters for the individual lags do not differ significantly from zero, which
makes it difficult to determine how the cumulative effect is distributed over time.
Nevertheless, the results suggest that the main impact of an increase in the unemployment rate occurs after two and three years rather than in the first year.

E f f e c t o f u n e m p l o y me n t r a t e o n s e l l e r r a te d e c r e a se s w i th w e lf a re
l e ve l s
Another analysis confirms that there is no universal causal effect of unemployment on direct seller rate. The results of model 6 suggest that there may be a
causal effect, but that it decreases with levels of GDP/Capita. The relationship is

21

marginally higher in countries with the lowest levels of GDP/Capita, varying from
0.25 for Romania to 0.23 for Luxembourg1.
The variation between countries is not very high. The range (0.23 - 0.25) suggests a somewhat smaller effect than the cumulative effect according to model 5
(0.31%). On the one hand, this confirms the positive relationship (two different
models finding similar outcomes), but on the other hand it suggests that the
largest part of the effect occurs already in the first year which is inconsistent
with the findings from model 5. Ideally we would combine these two models, allowing the cumulative effect to vary over time and between countries, but we do
not have enough data to estimate such a complex model.

N o s u p p o r t f o r th e pr e m i s e t h a t di r e c t se l l i n g i s m a n i fe s t a t i o n o f e n t r e pr e n e u r s h ip
The point of departure for our study was the premise that direct selling can be
seen as a manifestation of entrepreneurship, in particular necessity-based entrepreneurship. Our data do not support this premise, however. Although we find a
significant positive correlation between necessity-based entrepreneurship rate
and direct seller rate, our multivariate analyses do not find any relationship between entrepreneurship and direct seller rate (neither for entrepreneurship in
general, nor for necessity-based or opportunity-based entrepreneurship).
In chapter 2 we differentiated between necessity-based activities and necessitybased entrepreneurship. The results of our analysis suggest that the majority of
the direct sellers do not see themselves as entrepreneurs and do not register
themselves as such: to them, direct selling is primarily a secondary job rather
than an entrepreneurial activity. Our data do not allow us to determine whether
this is indeed the case; this would require micro-level data on the attitudes of
individuals.

5.2

Limitations and future research


Our goal has been to conduct an explorative study into the macro-economic determinants of direct selling. Hence, our research is subject to some limitations
and room exists for future research into this topic.
L i mi t a t i o n s
Our literature search revealed that no academic empirical research into the determinants of direct selling currently exists. Our study should therefore be considered an explorative one, taking the first steps into empirically examining direct selling. This means that we work without a pre-existing theoretical framework to guide us in our choice of determinants and in the interpretation of our
results.
Our results suggest that much of the variation in direct seller rates can be explained by country differences. We have also found that the impact of unem-

22

The country-specific effect of the unemployment rate on the direct seller rate can be calculated
using the parameter estimates from model 6. The general equation is 0.27
0.01*Ln(GDP/Capita). For Romania, this becomes 0.27-0.01*ln(5.8)=0.25; for Luxembourg, this
becomes 0.27-0.01*ln(78.6)=0.23

ployment on direct selling is weaker for countries with higher welfare levels.
These differences call for country-level research (i.e. time-series analyses on a
country-by-country basis) to better examine how unemployment affects direct
selling. Unfortunately, our data includes too few observations to allow for such
analyses. Furthermore, our model seems to have missed certain key countrylevel variables that explain direct selling as can be seen by the importance of the
country-fixed effect. Examples of these factors could be the social security network in a country or the national attitude towards direct selling as a source of
income. Such factors should be considered in future research.
F u t ur e r e s e a r c h
The direct seller rate is likely to be endogenous: changes in the unemployment
rate affect the direct seller, but depending on the definition of unemployment
used, changes in the direct seller rate may also affect the unemployment rate.
We have focused on one part of this relation: the impact of changes in the unemployment rate on the direct seller rate. We have tried to account for the inherent endogeneity somewhat by including lags in our model. Future research
should also focus on the other part of the relation: the impact changes in the direct seller rate have on the unemployment rate. Such research may prove to be
more interesting from the perspective of economic policy.
Future research should also more explicitly consider the role of attitudes. This
could be the attitude direct sellers have towards their own position: do they consider themselves to be entrepreneurs or do they view direct selling as a secondary type of job merely meant to supplement their income? The attitude towards
direct selling as a source of income in a country could also be a valuable factor in
explaining the variation in the direct seller rate.

23

References

Abilio, Ludmila Costhek (2011), Labour makeup: a case study of 800,000 cosmetics resellers in Brazil, Work Organisation, Labour and Globalisation 5(1),
96-110
Brodie, Andrew Stewart (1999), Self-employment dynamics of the independent
contractor in the direct selling industry, PhD thesis, University of Westminster, London
De Casanova, Erynn Masi (2011), Making Up the Difference: Women, Beauty,
and Direct Selling in Ecuador, University of Texas, Austin, Texas.
Federation of European Direct Selling Associations (2003), available at:
www.fedsa.be (accessed December 2012).
Seldia (2011), Promoting the economic and consumer benefits of the direct selling channel Annual report 2010-2011, Seldia, Brussels.
Stanworth, John; Brodie, Stewart; Wotruba, Thomas; Purdy, David (2004), Outsourcing salesforces via self-employment: the case of direct selling in the
UK, Journal of Small Business and Enterprise Development 11(1), 50-59
Welsh, D.H.B.; Pendleton, Y. (2006), Direct selling worldwide: The Mary Kay
Cosmetics story, International Journal of Family Business 3(1), 61-68

25

ANNEX I

Methodology of literature review

In this annex we describe how the literature review has been performed.
Step 1: Identification of relevant search terms
The purpose of the literature search is to identify scientific publications that deal
with the following research questions:
What is the relationship between the amount of people involved in direct selling and the rate of unemployment?
Which factors determine the amount of direct sellers in a country?
Accordingly, appropriate search terms have been designed in order to capture all
relevant recent material on these subjects. The primary objective was to be as
complete as possible, i.e. to minimise the probability that relevant publications
would be missed due to a certain search term not being included. As a result, the
list of search terms includes many synonyms for certain terms (for example,
more than 10 different search terms were suggested that could refer to direct
selling, such as direct sales, independent selling, independent sales, person-toperson selling, person-to-person sales, etc.) The final list includes 26 different
search terms, which are listed in Annex II.
Step 2: Selection of relevant sources
The literature search has been conducted within the following on-line sources:
www.ideas.repec.org. The Research Papers in Economics (RePEc) database
currently contains over 1.2 million economic research pieces, including working paper series from universities across the world and international organisations such as OECD, and the ILO.
http://www.sciencedirect.com. Science Direct is an internationally renowned
database for research papers from various disciplines.
http://scholar.google.nl. Google Scholar contains scholarly literature across
many disciplines and sources, including theses, books, abstracts and articles.
Whenever the search options of the respective source permitted, our objective
was to apply a Boolean search methodology that would include all of the predetermined search terms (for example, by searching for ("direct selling" OR "direct
sales" OR "independent selling" OR "independent sales" OR "person-to-person
selling" OR "person-to-person sales" OR "peddling" OR "single level marketing"
OR "multi level marketing" OR "network marketing" OR "direct marketing") AND
(unemployment OR "unemployment rate")).

S t e p 3: A p p l i c a t i o n o f t h e me t h o d ol o g y
We have explored the scientific databases listed above. We aimed to make use of
all search terms that were determined up front, by searching for them in the abstracts of the publications included in the databases. However, due to limitations
within the search options of the respective databases, it was not feasible to include all possible synonyms in all sources at once. Therefore we had to test and
adapt search terms, in order to come up with smart combinations.
Whenever a specific search resulted in an acceptable number of publications (no
more than 250 titles), the titles and abstracts of these publications were read to

27

filter out any papers that turned out not be relevant after all. The remaining
publications were saved. In case a search yielded an unmanageably large number of articles (more than 250 titles), we adapted the search terms in order to
narrow down the results. Whenever narrowing down the number of results did
not work out (i.e. still a very large number of articles), we chose to examine the
first 250 results instead.
Step 4: Final selection of relevant papers
The final step in the search methodology consisted of reading the full text of the
identified papers. Based on their content, we established whether these papers
would provide an adequate answer to our research question. The main criterion
was whether the papers contained a solid empirical basis, implying the following:
usage of macro-level data;
description of the data collection process;
an empirically sound methodology.
In case the paper adhered to these criteria, we considered it a key piece and reported relevant findings in the literature review.

28

ANNEX II

List of search terms used

Table 4

List of search terms used

Direct selling

Sales persons

Unemployment

Determinants

Synonyms

Synonyms

Synonyms

Synonyms

direct selling

representatives

unemployment

determinants

direct sales

sellers

unemployment rate

factors

independent selling

salesman

independent sales

sales people

person-to-person selling

labour force

person-to-person sales

work force

peddling

sales force

variables

single-level marketing
multi-level marketing
network marketing
direct marketing
direct-to-consumer
merchandising
door-to-door selling

Source: Panteia, 2013.

29

ANNEX III

Country descriptive statistics

Population density

GDP per capita

Second-job rate

Entrepreneurship rate

x1

Austria

0,3

55

75

4.4

101.8

34,100

3.8

11.0

n.a.

n.a.

n.a.

Belgium

0,4

72

64

8.3

358.7

32,700

3.8

12.2

0.4

2.8

2.6

Croatia

0,4

60

95

11.8

78.1

10,200

2.2

15.3

1.8

3.5

4.1

Czech Republic

4,2

92

94

7.3

136.2

25,444

2.0

15.8

n.a.

n.a.

n.a.

Denmark

2,5

60

93

7.5

128.7

40,259

7.7

8.1

0.3

2.8

1.8

Estonia

3,1

95

*95

16.9

30.9

20,417

4.1

6.7

n.a.

n.a.

n.a.

Finland

3,0

80

92

8.4

17.6

36,357

4.1

11.8

1.0

4.3

2.4

France

1,1

78

80

9.7

102.5

29,900

3.0

9.8

1.5

4.2

3.8

Germany

1,6

73

90

7.1

229.0

30,500

3.4

10.1

1.1

2.8

2.5

Greece

3,1

75

12.6

86.4

27,836

2.7

26.3

1.5

4.0

2.1

Hungary

5,6

78

86

11.2

107.5

20,452

1.6

10.6

1.4

5.7

4.9

Ireland

1,0

75

85

13.9

65.4

40,637

1.8

13.9

2.1

4.5

4.4

Italy

1,6

75

94

8.4

200.7

31,946

1.4

21.4

0.3

2.0

1.3

Lithuania

1,9

94

58

18.0

52.4

8,400

4.2

7.6

n.a.

n.a.

n.a.

Luxembourg

0,3

100

100

4.6

196.0

78,600

2.9

7.3

n.a.

n.a.

n.a.

Netherlands

0,3

80

83

4.5

492.2

42,230

6.9

13.7

0.6

6.1

4.0

Norway

3,1

85

10

3.6

16.0

57,431

8.1

7.2

1.2

6.3

4.4

Poland

5,0

90

85

9.6

122.1

19,898

6.6

17.0

n.a.

n.a.

n.a.

Portugal

1,4

96

12.0

115.4

25,428

5.3

17.7

1.0

3.1

1.9

Romania

2,7

82

11

7.3

93.2

5,800

2.5

20.1

1.3

2.9

3.3

5,8

91

65

7.5

n.a.

15,768

n.a.

n.a.

1.3

2.5

2.2

85

14.5

110.7

12,100

1.0

13.6

n.a.

n.a.

n.a.

Russia

77

80

Nascent
entrepreneurship

Unemployment rate

Opportunity-based
entrepreneurship

Part-time
direct sellers

Necessity-based entrepreneurship

Female
direct sellers

Descriptive statistics by country for 2010

Direct seller rate

Table 5

Slovakia

3,7

Slovenia

2,9

65

80

7.3

101.7

17,400

3.4

11.2

0.8

3.7

2.2

Spain

0,9

86

86

20.1

91.8

22,800

1.7

12.7

1.1

3.1

2.2

Sweden

1,7

77

92

8.4

22.9

37,300

7.7

9.8

0.7

4.1

2.3

Switzerland

0.2

53

60

4.5

195.6

53,400

6.4

0.7

4.1

2.2

Turkey

2.7

90

65

10.7

95.0

7,500

2.5

22.4

3.2

4.7

3.7

United Kingdom

1.4

91

94

7.8

252.5

27,500

3.5

12.5

0.7

5.2

3.2

90

12.6

Source: Seldia; Eurostat; ILO; GEM; Panteia, 2013.


2008 data

31

You might also like