Conference PaperPDF Available

Explaining Differences in Compliance with Food Standards: Evidence from the International Featured Standard

Authors:

Abstract and Figures

We use a new database of food producing firms to explain different outcomes of third party certification audits across firms. The database from the retailer-initiated certification scheme of the International Featured Standard (IFS) includes auditing reports from more than 13,000 firms producing in 96 countries over the period 2011 to 2013. Information on each firms' conformance with the private standard require-ments as well as company specific characteristics such as branch, region, past experience with the standard and firm size is provided. While the previous literature emphasizes the role of auditor independence for differing audit results, we suggest that company specific variables play an important role. In our empirical analysis, we find a U-shaped relationship between firm size and performance: microenter-prises and large companies are more likely to conform to the requirements of the standard than small to medium sized companies – confirming our predictions from organizational theory. We conclude our research with suggestions for further re-search.
Content may be subject to copyright.
Explaining Differences in Compliance with Food
Standards: Evidence from the International Fea-
tured Standard
Axel Mangelsdorf , Michel Tolksdorf
Berlin Institute of Technology
&
BAM Federal Institute for Materials Research and Testing, Germany
Abstract
We use a new database of food producing firms to explain different outcomes of
third party certification audits across firms. The database from the retailer-initiated
certification scheme of the International Featured Standard (IFS) includes auditing
reports from more than 13,000 firms producing in 96 countries over the period 2011
to 2013. Information on each firms’ conformance with the private standard require-
ments as well as company specific characteristics such as branch, region, past
experience with the standard and firm size is provided. While the previous literature
emphasizes the role of auditor independence for differing audit results, we suggest
that company specific variables play an important role. In our empirical analysis,
we find a U-shaped relationship between firm size and performance: microenter-
prises and large companies are more likely to conform to the requirements of the
standard than small to medium sized companies confirming our predictions from
organizational theory. We conclude our research with suggestions for further re-
search.
Key Words: Firm size, Food safety, Food Manufacturing, Private Standards, Stand-
ard compliance, Tobit model, Ordinary least squares, Generalized linear model
1. Introduction
Private food safety standards increasingly gained importance within the last three decades. Sev-
eral food safety scandals in the European Union lead to drawbacks in consumers trust in re-
tailers’ safe food. Changes in governmental product liabilities have driven food retailers and
retailer associations to develop new measures. Private standards have emerged and partly sub-
stituted management standards like the ISO 9001. Auditors from third party certification bodies
are responsible to confirm firms’ compliance with private standards. Private standards are de-
signed in a more specific and suitable way than management standards and contain harsher
requirements than governmental food safety regulations. They are used for legal reasons and
also as a tool for improvement of reputation and business relations (Schulze et al. 2008).
The International Featured Standard (IFS) is a private food safety standard. It was introduced
by the German retailer association Handelsverband Deutschland (HDE) in 2002. The French
retailer association Fédération des Enterprises du Commerce et de la Distribution (FCD) to-
gether with Italian and Swiss retailers joined the Technical Committee to develop new versions
of the standard. With increasing global scope, North and South American as well as Asian
stakeholders are now cooperating in standardization (IFS 2012).
The German retailer associations provide us full access to the global auditing database.
1
With
the database, we can use information of more than 30,000 auditing reports from more than
13,000 companies over the period January 2011 to July 2013. The purpose of our paper is to
explain differing audit results of certified companies. In contrast to previous literature on au-
diting (including financial auditing), where the focus was on the auditor independence, we ex-
amine the impact of producer features. Specifically, we focus on two company characteristics.
First, we examine the relationship between standard conformance and the size of the food pro-
ducing companies. Second, we examine the relationship between the companies’ past experi-
ence with the standard and the companies’ ability to satisfy the standard requirements.
The remainder of our paper is structured as follows. In Section 2, we discuss the theoretical
background, including producers’ necessity of quality disclosure and how this affects a certifi-
cation scheme and its actors. In Section 2, we also briefly review the existing literature on the
influence of auditor independence and producer features in the outcome of audits. We borrow
ideas from organizational theory to formulate our hypotheses in Section 3. In Section 4, we give
information on the database we use and provide descriptive statistics. In Section 5, we then
conduct the empirical investigation. In our econometric models, we can confirm the predictions
of our hypotheses regarding the impact of producer features on the outcome of the auditing
process. In the last section, we summarize the findings of our paper and give recommendations
for future research.
1
Acknowledgement: We would like to thank Stephan Tromp from HDE and Wendy Ehsemann from convivo
GmbH for providing us full access to the IFS database.
2. Theoretical Background and Related Literature
Classification of Goods Attributes
Why is it necessary for food producers to become certified? To answer that question, we first
look at different classifications of goods in terms of consumers’ ability to assess the quality of
the product. In Figure 1, we show the classification of different types of goods with increasing
information asymmetry. Nelson (1970) distinguishes between goods that have search and ex-
perience attributes. For search attributes, consumers know the quality of the good before pur-
chasing. This applies, for instance, to the freshness of food products. Experience goods, in con-
trast, require that the consumers learn the quality of the product after consumption. For exam-
ple, consumers only know the taste of food products after consumption. A third category, cre-
dence attributes, was introduced by Darby and Karni (1973). Credence goods have attributes
which can be observed by consumers only to prohibitive costs. One example for a credence
attribute is the nutrition content of food products. Finally, Tietzel and Weber (1991) revealed
another type of product attribute, so called Potemkin attributes. Like credence attributes the
quality cannot effectively be observed after consumption, but may also not be observed with
additional effort, as it is the case for credence attributes. Only process accompanying observa-
tion offers genuine information about Potemkin attributes.
Figure 1: Typology of Attributes
Search Attribute
Experience Attribute
Credence Attribute
Potemkin Attribute
Qualities which are known
before purchase
Qualities, which are
known only after con-
sumption
Qualities, which can be ob-
served by a single cus-
tomer only to prohibitive
costs, but buyers can rely
on third party judgments
Process-oriented qualities,
which are hidden for third
parties as well as for cus-
tomers at the end product
level
Freshness, appearance
Taste, shelf life
Nutrition, contamination
Animal welfare, fair trade
Increasing information asymmetry
Source: Based on Jahn, Schramm, and Spiller 2005
Food safety certifications are mainly directed at credence and partly at Potemkin attributes. For
example, consumers fail to recognize harmful bacteria or salmonella. Instead, laboratories have
to test food safety and verify that it is safe for consumption. Retailers of food products have an
incentive to sell only safe food to their consumers and force their suppliers to become certified.
Requirements of food safety standards minimize that the food products are contaminated. By
forcing their suppliers to become certified, retailers minimize the risk of extensive costs from
recall campaigns and reputation losses. Food producers have the incentives to become certified
additionally due to the ability to access larger markets. For both producers and retailers, a third
party certification acts as a quality signal. The quality signal reduces the information asymme-
tries between producers and buyers. Food safety certifications aim to increase the trust of con-
sumers that the food is safe.
Third party assessments are better quality signaling instruments than second party assessments
(Tanner 2000), i.e. quality assessment activities by the producers themselves (first party) or by
the retailer (second party). Moreover, Holleran, Bredahl, and Zaibet (1999) show that due to
the dominance of retailers on the market, the requirement of a third party certificate effectively
shifts the quality assurance costs from retailers to suppliers. The suppliers are willing to bear
these costs for an increase in market access, due to the more competitive environment of this
stage in the supply chain. However, third party assessments can also have a cost saving effects
since they prevent multiple audits by different retailers and are thus more efficient.
A number of additional factors further increases the trustworthiness of the IFS standard, next
to its third party role. First, the certification body must be accredited by an official accreditation
agency. Second, besides the accreditation, the standard owner additionally assesses the services
of the certification body. Third, auditors require a special training before receiving the official
IFS auditor status.
Related Literature Review on Differences in Auditing Results
A number of studies try to explain differences in auditing results. In the financial auditing lit-
erature, the independence of auditors is in question (DeAngelo 1981). The main argument is
that certification bodies are economically dependent from their customers which leads to biased
outcomes of the certification process. Duflo et al. (2012) find that the payment structure has an
influence on the truth-telling of auditors. The authors show in a field experiment that a pooled
payment benefits auditor independence compared to payments by the audited firms.
In a series of papers Albersmeier et al. (2009, 2010) find that auditor independence and auditor
quality also plays a role for food certifications. The authors use the database based on the Qual-
ität und Sicherheit GmbH (QS) and argue that differences in auditing quality occur because
some auditors lack training and knowledge. With regard to auditor independence, the authors
assume that low audit quality can be a result of auditors’ incentives to minimize their costs and
increase their chances of future contracts with the food supplier.
Albersmeier et al. (2009) argue that their study is limited to the analysis of the QS database
which does not provide sufficient information to explain all reasons for variations among au-
diting quality. Schulze et al. (2008) questioned IFS certified companies about how they per-
ceived the implementation and certification process. They find higher intrinsic motivation to
implement the standard for smaller companies than for medium sized companies, who per-
ceived the certification as a costly and enforced requirement. Large companies had no problems
with the implementation, but also lacked intrinsic motivation, partly because of the similarities
to other management standards, that are already implemented in their companies. Since our
database on companies performance with the IFS standard includes more company specific
information than the QS database, we add to the research by focusing on company specific
information. Therefore, instead of focusing on the auditors to explain differences in the outcome
of the certification process, we argue that specific features of the audited firms are responsible.
In the next section, we develop two Hypotheses: (1) The impact of company size and (2) the
impact of companies’ experiences with the standard.
3. Hypotheses Concerning Impact Factors on Standard
Conformance
Beginning with company size, we borrow a model from organizational theory to explain the
relationship with variances in certification performance. A classical model developed by
Greiner (1972) shows the age and size of a company as the determining factors that lead com-
panies to find different solutions for management problems, or ‘crises’.
According to the model, companies go through different phases shown in Figure 2. Small and
young companies pass through the creativity stage, where leadership problems might occur.
This is solved by ‘direction’, i.e. solving managerial problems, introducing business techniques
and determining business activities. With increasing company size an ‘autonomy’ crisis will
eventually occur. The manager or owner of the company is overburdened with the increasing
load of tasks and he needs to delegate responsibilities. As delegation is newly introduced to the
company, ‘control’ issues may arise. This happens if the executives of the delegated responsi-
bilities begin making inconsiderate and uncoordinated decisions. Therefore coordination solves
the ‘control’ crisis. For the food sector we assume that especially small to medium companies
that just began dealing with their growth by delegating responsibilities might be overburdened
with a simultaneous introduction of a food safety standard like the IFS, whereas companies that
are ‘small enough’, like microenterprises, do not yet face the ‘autonomy’ and ‘control’ crises
and therefore struggle less with the introduction. Larger companies, in contrast, already over-
came the delegation and coordination crises and are less likely to have problems with conform-
ing to requirements of private standards. Moreover, large enterprises have sufficient financial
resources to hire a quality manager responsible for quality assurance tasks.
We conclude from the model that the relationship between company size and the ability to meet
the requirements of the standard is non-linear. That is, we assume that very small companies
with only few employees have relatively small problems to deal with hygiene and safety
measures and quality management tasks. Large companies which employ a quality manager
and have experience with other quality management standards are also likely to confirm to the
requirements of the standard. Small to medium sized companies are more likely to face auton-
omy and control problems and are therefore more likely to have worse certification outcomes
than microenterprises and large companies. Therefore, we formulate Hypotheses 1 as follows:
Hypothesis 1 “Firm Size”: The relationship between firm size and standard conform-
ance is likely to follow a U-shape. The ability to conform to the standard requirements
decreases with firm size to a turning point and increases after.
The second hypothesis is related to the relationship between companies’ experience with the
standard and the ability to conform to the standard requirements. When a standard scheme is
established, firms will have no experience on how to meet the standard. With every audit, firms
gain experience. A company that fails an audit in the first round will eventually improve its
conformity and re-apply for the certificate and is likely to achieve better results in the re-audit.
Moreover, the IFS standard we consider in our paper includes the formulation of improvement
suggestions, when requirements were not or not completely met in the audit. Therefore, com-
panies have the incentive to take such suggestions into account and improve their performance
in the next audit. Against this background, we formulate our second Hypothesis as follows:
Hypothesis 2 “Experience with the Standard”: The ability of firms to meet the re-
quirements of the standard increases with each audit.
Besides firm size and experience with the standard, we include a number of control variables
in our model. One variable is the region and country where the food producing company is
situated. This control variable captures effects such as weather and climate conditions, i.e. hot
climate conditions may have a negative effect for producers of perishable food products. We
also include dummy variables for producers of specific food items in order to capture sector
specific effects. For example, we expect that producers of perishable goods, such as meat and
vegetables are more likely to have lower certification results than producers of non- perishable
goods.
4. The Database of the International Featured Standard
The data that is used for the analysis was provided from the IFS database and contains infor-
mation of all conducted IFS audits from January 2011 to July 2013. The complete inventory
count of the database enables us to produce more robust result than survey data (which may
suffer from survey biases or unanswered questions) or case study approaches which may fail to
provide universal results.
The aim of our paper is to explain differences in outcomes of the certification process across
firms. Therefore, we first look at the standard conformance - the dependent variable in our
empirical model. The IFS food standard includes hygiene and safety requirements as well as
management practices similar to the ISO 9001 management standard. In contrast to IS0 9001,
where the certification decision is bivariate, i.e. either the company complies with the standard
and becomes certified or not, the IFS food standard is designed so that food manufacturers can
achieve different results.
Depending on how the company meets the different requirements, auditors can give a producer
a score from 0 to 100 percent while 0 percent means a very bad and 100 percent a very good
result. In order to become an IFS certified producer, the company has to score at least 75 per-
cent. The 75 percent level represents the so called ‘base level’. All firms which fall below the
‘base level’ receive no certificate and can apply for a re-audit. In order to become a ‘high level’
producer, food producers must at least score 95 percent. The frequencies and shares of each
grade are presented in Table 1. More than 80 percent of the certificates are granted on the high
level’ and less than 2 percent of the certificates are refused.
Table 1: Frequencies and shares of dependent variable (certificate score)
Frequency
Share (in %)
Certificate refused (<75%)
602
1.89
Base level (>75%)
5,391
16.91
High level (>95%)
25,895
81.21
Total
31,888
100
In Table 2 we present descriptive statistics of our Model Variables. The first variable ‘Audit
results’ represents the outcome variable shown in Table 1. The mean value of 95.81 percent
shows that overall companies receive very good results. However, as the minimum value is 0
percent, some companies are completely failing the audit. The second variable ‘Audit duration’
measured in hours, measures the time the auditor was actively auditing the food company. On
average, the auditor is about 16 hours in the company. There is a wide span between minimum
(1 hour) and maximum (71 hours). With regard to company size, the average food producing
company which is certified with the IFS food standards has 120 employees. The largest com-
pany in our dataset has 9,000 employees. We use the variable ‘Number of employees per com-
pany’ to test the predictions of our first variable. The IFS database shows that 62 certification
bodies are responsible to conduct audits: On average, a certification body audited about 500
companies in the respective period.
Further we include a dummy variable for each certification body, as we suggest that our pre-
dictions apply independently of auditor specific characteristics.
Table 2: Descriptive Statistic of Model Variables
Variable
Mean
Standard Deviation
Min
Max
Audit result (in %)
95.81
7.25
0
100
Audit duration (in hrs.)
16.43
5.88
1
71
Num. of employees
121.32
251.29
0
9,000
Num. of audits (per CB)
499.94
634.10
1
3,151
Table 3 shows the number of audits in the years 2011 to July 2013. The number of audits in-
creased from 2011 to 2012. Since a certificate is granted for one year only, a company can be
audited up to three times in the reviewed period. In total, 13,354 companies have been audited
from 2011 to 2013. In our multivariate empirical investigation in the next Section, we include
the year as a variable with the year 2011 as the base year, normalized to 0. Since each addi-
tional year represents further experience with the IFS food standard, we use the year variable
to test the predictions of our second Hypothesis.
Table 3: Audits per Year
Year
Frequency
2011
11,583
2012
12,097
2013 (until July)
8,208
The IFS food standard is a product of German and French retailer associations but audited sup-
pliers of food products come from 96 countries across the globe. As shown in Table 4, the
majority of audits (64 percent) take place in four countries: Germany, France, Italy and Spain.
Together with the remaining countries (‘OECD Rest’ in Table 4), OECD countries account for
more than 95 percent of the conducted audits. In order to control for the impact of countries and
regions on the outcome of the certification process, we include regional dummy variables in
our empirical investigation.
Region
Frequency
Share (in %)
Germany
7,324
22.97
France
4,035
12.65
Italy
5,047
15.83
Spain
4,029
12.63
OECD (Rest*)
8,019
25.15
Non-OECD High income countries
231
0.72
East Asia and Pacific
1,037
3.25
Europe and Central Asia
1,671
5.24
Latin America and the Caribbean
176
0.55
Middle East and North Africa
171
0.54
South Asia
73
0.23
Sub-Saharan Africa
75
0.24
Table 5 shows the different branches (product groups). The IFS food standard has been updated
from version 5 to version 6 in the period under investigation. Version 5 distinguished 19 differ-
ent branches and version 6 11 different branches. For our empirical investigation, we create a
dummy variable for each product group. In addition, we also create a dummy variable for the
Version 6 of the standard. Given the predictions of Hypothesis 2, we expect a negative sign for
the Version 6 dummy variable. Companies have a lack of experience with a new standard ver-
sion which will likely translate into a lower audit result. Additionally with Version 6 of the IFS
a new chapter of requirements concerning ‘food defense’ was introduced, therefore all compa-
nies will at least lack experience with these new requirements.
Table 5: Audits per Product Group
Branch IFS Version 5
Frequency
Branch IFS Version 6
Frequency
Fruits and vegetables
2,942
Fruit and vegetables
2,815
Meat products and prepa-
rations
1,984
Grain products, cereals, industrial
bakery and pastry, confectionery,
snack
2,551
Beverages
1,952
Red and white meat, poultry and
meat products
1,942
Bakery and baked products
1,791
Beverages
1,410
Ambient stable hermeti-
cally sealed products
1,494
Dairy products
1,266
Dairy
1,481
Combined products
1,195
Ready to eat
1,385
Fish and fish products
809
Dried goods
1,003
Dry products, other ingredients and
supplements
696
Confectionery
977
Oils and fats
402
Red meat - Chilled and Fro-
zen
778
Egg and egg products
251
Fish products and prepara-
tions
676
Pet food
77
Food ingredients
577
Fish - Chilled and Frozen
551
Poultry - Chilled and Frozen
494
Snacks and breakfast cere-
als
447
Oils and fats
437
Egg
366
Co-packers (Co-Packing
and Handling)
345
Wholesale
1
Total number of IFS 5 audits
20,698
Total number of IFS 6 audits
11,190
The last variable that we consider in our multivariate analysis captures whether the audit for the
IFS food standard was a combined audit. A combined audit means that multiple standards are
certified simultaneously within one audit. According to our database, additional standards are
other private standards, such as the British Retail Consortium (BRC) or Marine Stewardship
Council (MSC) or formal international management standards such as the ISO 9001. We in-
clude a dummy variable for combined audits that takes the value one when requirements for
another standard have been audited simultaneously and zero otherwise. About 5,000 of the
30,000 total audits are combined audits.
5. Statistical Investigation
In this Section we conduct the empirical analysis and present our results. The aim of our anal-
ysis is to find factors at the firm level that are likely to influence the outcome of the certification
process. In Section 3 we developed two hypotheses. First, based on a model that we borrow
from organizational theory, we argue that there is a non-linear relationship between firm size
and certification outcome. That is, microenterprises and large firms are likely to perform better
than small to medium sized firms. Second, we assume that there is a positive relationship be-
tween experience with the standard and the certification score. We use two variables to capture
the experience effect: year dummies and a dummy variable that captures the switch from IFS
food version 5 to version 6. We expect a positive effect of the former and a negative effect of
the latter variable.
As shown in Figures A1 and A2 in the appendix, both the average audit results and the share of
successful certifications have a non-linear pattern. Still this pattern might be biased by unob-
served factors like region, branch, year of certification etc. Decoupling the different influential
factors with descriptive statistics would be very extensive, as any combination of regions and
branches must be considered. Further the number of observations must be sufficient for each
case. Therefore we use an econometric approach to control for different influential factors and
to test whether the predictions are stable.
We use multivariate approaches, which, compared to bivariate models, have the advantage to
simultaneously regard the impact of several variables. Our initial dependent variable is the
standard conformance, expressed in percent of compliance with the requirements of the stand-
ard. Therefore we chose our models with respect to the fact that the dependent variable can only
take values between 0 and 100. For example an ordinary linear regression might have mislead-
ing results, as it would not consider the boundaries of the dependent variable.
There are two traditional approaches to deal with this problem. First the Tobit model, which is
restricted by an upper and a lower limit. The results are therefore attuned to the fact that there
are no observations above or below the threshold. Though the model still assumes that there are
potential observations above or below the threshold, but as long as there are not too many ob-
servations at the borders, the results are superior to those of a linear regression. Second is the
logit transformation of the dependent variable as in (1) where p remarks the audit score in per-
cent and t the transformation. After the transformation an ordinary least squares (OLS) regres-
sion can be applied. Advantages are the reduced bias of the distance between observations when
there is conglomeration of observations at the borders and that no potential results below or
above the threshold are assumed. A drawback is that the transformation is not defined for the
values 0 and 100, therefore such observations become missing values, still for the remaining
observations the predictions are more accurate.
(1)
  
  
Due to the differing advantages and drawbacks of both of these approaches we apply both of
them. Further we verify our findings with the application of a generalized linear model (GLM)
proposed by Papke and Wooldrige (1996). We consider these findings highly relevant that are
stable throughout all of the 3 models and questionable if they were discovered by 1 or 2 of the
models. In the Tobit model the dependent variable is the audit score with values between 0 and
100. In the OLS regression it is the audit score, transformed as described in (1) and in the GLM
model it is a proportional value between 0 and 1, which is the audit score divided by 100. The
independent variables represent the company specific factors and control variables that we de-
scribed in the last section. We estimate ‘Audit score’ as a function of firm level variables and
control variables:
(2)
  
Substituting the independent variables into equation (2) delivers the following equation (3):
(3)
         
    
Where Audit score is the standard compliance, as defined above; emp is the number of employ-
ees per firm; aud_n is the number of audits per certification body; aud_h is the length of the
auditing process in hours; emp, aud_n and aud_h have an underlying functional form, which
allows for transformations, including the squared form of the variables and including interaction
effects; year is a variable that takes the value 0 for the year 2011, 1 for 2012 and 2 for 2013;
ifs6 is a dummy variable that takes the value 1 when the IFS food standard switches from ver-
sion 5 to version 6; combi is a dummy variable that takes the value 1 when a combination audit
was conducted and 0 otherwise; branchi are dummy variables for the branches described in
Table 5 with i=1,…,30 for the different branches of both versions; regionj captures dummy
variables for the regions presented in Table 4 with j=1,…,12 for the different regions and cbk
with k=1,…,62 represents dummies for the 62 certification bodies.
Table 6 presents the results of the regressions. In Column 1 we present the results of the Tobit
model. In Column 2 we show the results of the OLS regression on the transformed dependent
variable. The GLM regression is presented in Column 3. We include the squared number of
employees in order to test the non-linear predictions of Hypothesis 1 in all models. Moreover,
we include a number of interaction variables in order to study the simultaneous influence of the
interacted variables and to control for possible dependencies.
Starting with the number of employees, we see that the variable has a negative and significant
effect on the audit score, while the squared form of the number of employees is significant and
positive in all three models. This suggest a U-shaped relationship between certification perfor-
mance and company size, as we argue in Hypothesis 1. A rise of the number of employees has
an initially negative impact, it will then reach a minimum, and from there on a further increase
has a positive effect. Figure A3 is a graphical representation of the estimated parameters for the
functional form of the number of employees of the Tobit model. Still this does not account for
the several interaction effects. In the graphic the minimum is at 137 employees, but actually it
might be lower, considering the positive level effect of the number of employees and audit
duration, as the audit duration is always a positive value and is higher for larger companies.
The number of audits per certification body has a positive, but insignificant, effect in the Tobit
model and in the GLM regression. However, the OLS regression found a significant negative
impact, thus the effect of different certification bodies activity remains unclear.
Audit duration has a significant negative effect in the Tobit model, as well as a negative but not
significant effect in the GLM regression. The OLS regression found a positive, though very
small effect. Still all models share the negative and significant effect on the squared form of the
audit duration. Therefore a long audit duration will eventually lead to a lower audit score. We
explain this result as follows: auditors need more time to audit low performing firms because
they have to explain in detail for which standard requirement the firm got a low score and why.
Additionally, auditors give improvement suggestions and formulate corrective action plans.
The variable Year that captures the time effects has a positive and significant effect in all three
models. The results suggests that an additional year of experience with the standard version
leads to a better audit score. The result is confirmed by the Version variable. The variable enters
with a negative sign and is significant in all models. Obviously, a switch to a newer version has
a negative impact on the auditing results because companies - at least partly - lose their experi-
ence of how to efficiently implement the standard or are unfamiliar with newly introduced re-
quirements. Both variables (Year and Version) are indicators that companies need experience
with a standard to achieve better audit results.
The effect of combination audits is positive in all models, but only significant in the Tobit
regression, whereas the interaction variable of the combination audit dummy with the number
of employees is negative in all models and also significant, except for the GLM regression. This
could possibly imply that combination audits lead to better results, while this effect becomes
smaller with increasing company size and eventually negative, though the evidence for that is
not stable. The interaction of combination audits with the number of audits and the audit dura-
tion has no clear results when comparing the three models.
With regard to the other interaction variables, we see that the interaction between the number
of audits per certification body and audit duration, as well as the number of employees and audit
duration have significant positive effects on the audit performance. The result suggests that for
long audit durations which are conducted by certification bodies with a high activity, companies
are better evaluated and also larger companies are more likely to receive better results on long
audits. This effect is opposing to the negative effect of the squared form of the audit duration.
Therefore it depends on the company, the certification body and the audit duration respectively,
whether the ultimate effect of a longer audit duration is positive or negative.
Table 6: Results of the Regressions
Tobit
OLS
GLM
Number of Employees (Log)
-0.738
-0.167
-0.154
(0.075)
***
(0.047)
***
(0.082)
*
Number of Employees (Log)²
0.075
0.012
0.022
(0.019)
***
(0.005)
**
(0.010)
**
Number of Audits per CB (Log)
0.416
-0.160
0.036
(1.094)
(0.065)
**
(0.168)
Audit duration in hrs. (Log)
-1.620
(0.855)
*
0.007
(0.181)
-0.282
(0.263)
Audit duration in hrs. (Log)²
-0.518
(0.096)
***
-0.178
(0.022)
***
-0.122
(0.029)
***
Number of Employees (Log) ×
Number of Audits (Log)
0.000
0.000
0.000
(0.000)
(0.000)
(0.000)
Number of Employees (Log) ×
Audit duration in hrs. (Log)
0.289
0.065
0.054
(0.000)
***
(0.017)
***
(0.026)
**
Number of Audits per CB (Log) ×
Audit duration in hrs. (Log)
0.345
0.066
0.075
(0.109)
***
(0.020)
***
(0.033)
**
Year
0.347
0.053
0.089
(0.061)
***
(0.167)
***
(0.023)
***
Version (Dummy)
-0.609
-0.167
-0.149
(0.185)
***
(0.037)
***
(0.085)
*
Combination Audit (Dummy)
3.183
0.231
0.494
(0.946)
***
(0.206)
(0.416)
Combination Audit (Dummy) ×
Number of Employees (Log)
-0.128
-0.035
-0.029
(0.062)
**
(0.016)
**
(0.032)
Combination Audit (Dummy) ×
Number of Audits per CB (Log)
-0.220
(0.120)
*
0.001
(0.022)
-0.021
(0.043)
Combination Audit CB (Dummy) ×
Audit duration in hrs. (Log)
-0.104
0.050
-0.018
(0.218)
(0.051)
(0.105)
Intercept
94.619
4.641
3.104
(6.898)
***
(0.468)
***
(1.095)
***
Pseudo R-squared
0.0107
R-squared
0.1595
Number of observations
24,473
24,347
24,473
Robust standard errors are in parentheses (adjusted for 12,572 clusters according to company
ID for Tobit and GLM and 12,542 clusters for OLS); 57 left-censored observations at percent=0;
69 right-censored observations at percent=100 in the Tobit model; *** significant at 1%; **
significant at 5%; * significant at 10%; 12 region dummies, 30 branch dummies and 62 certifica-
tion body dummies considered
With regard to regional and branch dummy variables (presented in Table A in the Appendix),
we reveal that Italian companies achieve significantly better results than other OECD countries,
while Spanish producers have significantly lower results. This may be caused by the fast diffu-
sion of the standard, which occurred in a rather short time in Spain. For example the number of
audits conducted in Spain and France are nearly equal, with both countries relatively similar in
size, though Spanish retailer associations joined the standard much later. Therefore this implies
that Spanish companies have less experience with the standard than French companies. Coun-
tries from the developing regions of the world (Latin America, East Asia, Middle East and
North Africa) achieve significantly lower audit scores. With regard to branch dummies, we find
that companies producing perishable goods - such as meat or fish products - achieve signifi-
cantly lower scores than non-perishable goods such as beverages or dry products.
6. Summary and Conclusion
Private standards are more and more important for food producing companies. Retailer associ-
ations are developing such private standards and third party certification bodies are evaluating
firms’ performance with the standards requirements. Why are some firms performing better
than others when it comes to complying with private standards? In order to answer our research
question, we use information from the IFS database that includes information from more than
13,000 companies in 96 countries.
While past research has focused extensively on the role of auditors’ abilities to evaluate the
conformance with standards, we focus on firm specific factors to explain variances in certifica-
tion outcomes. Specifically, we developed two hypotheses. For our first Hypothesis, we borrow
a model from organizational theory and we argue that there is a non-linear relationship between
firm size and the ability to conform to the standard. Microenterprises and large firms are likely
to perform better than small to medium sized firms. In our second Hypothesis we assume that
there is a positive relationship between experience with the standard and the certification score.
In our empirical analysis, we use a linear multivariate Tobit model, an OLS regression on a
transformed dependent variable and a GLM regression to test the predictions of our two Hy-
potheses. For the first Hypotheses, we include the number of employees per firms to capture
the firm size effect. For the second Hypotheses, we use two variables to capture the experience
effect: a year variable and a dummy variable that captures the switch from IFS food version 5
to version 6.
We can confirm both Hypotheses. We find that firm size has a negative impact on the audit
score and the square of firm size has a positive sign. Thus, we find a U-shaped relationship
between audit score and firm size. With regard to the second Hypotheses, we find that the year
variable has a positive influence and the switch to a new standard version has a negative influ-
ence. Therefore, we can confirm that companies need more experience in order to fulfill the
requirements set out in the food standards. Besides the variables to check the predictions of our
Hypotheses, we include a number of control variables such as sectors and regional dummies.
The inclusion of control variables means that our results are stabile across countries and sectors.
Regarding the influence of countries or regions on firms’ ability to conform to the standard, we
find that firms from developing countries have significantly lower audit scores than firms from
high-income OECD countries. Moreover, as expected, we find that firms which produce per-
ishable goods achieve significantly lower scores than non-perishable goods.
While our research for the first time provides evidence on a number of firm level variables on
the influence of variances in certification results, we argue that future research should consider
additional factors. For instance, it would make sense to test whether firms which employ a
dedicated quality manager are better able to deal with the requirements of food standards than
firms with a provisional quality manager, who also has other responsibilities in the company.
Also we could not clearly determine the effect of combination audits and therefore whether it
is beneficial for a firm, in terms of the audit result, to have multiple standards certified at once.
Literature
Albersmeier, F., H. Schulze, G. Jahn, and A. Spiller. 2009. "The reliability of third-party
certification in the food chain: From checklists to risk-oriented auditing." Food Control
20 (10):927-935. doi: 10.1016/j.foodcont.2009.01.010.
Albersmeier, Friederike, Holger Schulze, and Achim Spiller. 2010. "System dynamics in food
quality certifications: Development of an audit integrity system." International Journal
on Food System Dynamics 1 (1):69-81.
Darby, Michael R, and Edi Karni. 1973. "Free competition and the optimal amount of fraud."
JL & econ. 16:67.
DeAngelo, L. 1981. "Auditor size and audit quality." Journal of Accounting and Economics
3:183-199.
Duflo, Esther, Michael Greenstone, Rohini Pande, and Nicholas Ryan. 2012. "Truth-telling by
third-party auditors: Evidence from a randomized field experiment in India." 21st
BREAD Conference.
Greiner, Larry E. 1972. "Evolution and revolution as organizations grow."
Holleran, Erin, Maury E Bredahl, and Lokman Zaibet. 1999. "Private incentives for adopting
food safety and quality assurance." Food policy 24 (6):669-683.
IFS. 2012. IFS Food - Standard for auditing quality and food safety of food products, Version
6.: Internation Featured Standards.
Nelson, Phillip. 1970. "Information and consumer behavior." The Journal of Political
Economy:311-329.
Papke, L. E. and J. Wooldridge. 1996. Econometric methods for fractional response
variables with an application to 401(k) plan participation rates. Journal of Applied
Econometrics 11: 619-632
Schulze, Holger, Friederike Albersmeier, C Gawron, Achim Spiller, and Ludwig Theuvsen.
2008. "Heterogeneity in the evaluation of quality assurance systems: the International
Food Standard (IFS) in European agribusiness." International Food and Agribusiness
Management Review 11 (3):99-139.
Tanner, Bob. 2000. "Independent assessment by third-party certification bodies." Food
control 11 (5):415-417.
Tietzel, Manfred, and Marion Weber. 1991. "Von Betrügern, Blendern und Opportunisten :
eine ökonomische Analyse." Zeitschrift für Wirtschaftspolitik 40 (2):28.
Appendix
Table A: Results of the Regressions (With Regions and Branches)
Tobit
OLS
GLM
Number of Employees (Log)
-0.738
-0.167
-0.154
(0.075)
***
(0.047)
***
(0.082)
*
Number of Employees (Log)²
0.075
0.012
0.022
(0.019)
***
(0.005)
**
(0.010)
**
Number of Audits per CB (Log)
0.416
-0.160
0.036
(1.094)
(0.065)
**
(0.168)
Audit duration in hrs. (Log)
-1.620
(0.855)
*
0.007
(0.181)
-0.282
(0.263)
Audit duration in hrs. (Log)²
-0.518
(0.096)
***
-0.178
(0.022)
***
-0.122
(0.029)
***
Number of Employees (Log) ×
Number of Audits (Log)
0.000
0.000
0.000
(0.000)
(0.000)
(0.000)
Number of Employees (Log) ×
Audit duration in hrs. (Log)
0.289
0.065
0.054
(0.000)
***
(0.017)
***
(0.026)
**
Number of Audits per CB (Log) ×
Audit duration in hrs. (Log)
0.345
0.066
0.075
(0.109)
***
(0.020)
***
(0.033)
**
Year
0.347
0.053
0.089
(0.061)
***
(0.167)
***
(0.023)
***
Version (Dummy)
-0.609
-0.167
-0.149
(0.185)
***
(0.037)
***
(0.085)
*
Combination Audit (Dummy)
3.183
0.231
0.494
(0.946)
***
(0.206)
(0.416)
Combination Audit (Dummy) ×
Number of Employees (Log)
-0.128
-0.035
-0.029
(0.062)
**
(0.016)
**
(0.032)
Combination Audit (Dummy) ×
Number of Audits per CB (Log)
-0.220
(0.120)
*
0.001
(0.022)
-0.021
(0.043)
Combination Audit CB (Dummy)×
Audit duration in hrs. (Log)
-0.104
0.050
-0.018
(0.218)
(0.051)
(0.105)
Intercept
94.619
4.641
3.104
(6.898)
***
(0.468)
***
(1.095)
***
Region Dummies
Germany
-0.213
(0.118)
*
-0.088
(0.027)
***
-0.060
(0.050)
France
-0.465
(0.168)
***
-0.155
(0.032)
***
-0.129
(0.057)
**
Italy
1.649
(0.113)
***
0.486
(0.026)
***
0.534
(0.043)
***
Spain
-1.030
(0.229)
***
-0.197
(0.038)
***
-0.237
(0.055)
***
East Asia and Pacific
-2.924
(0.271)
***
-0.585
(0.045)
***
-0.609
(0.055)
***
Europe and Central Asia
-0.187
(0.150)
-0.060
(0.033)
*
-0.053
(0.050)
Latin America and the Caribbean
-2.399
(0.863)
***
-0.317
(0.130)
**
-0.553
(0.159)
***
Middle East and North Africa
-1.915
(0.862)
**
-0.258
(0.107)
**
-0.444
(0.174)
**
Non-OECD High-income coun-
tries
-0.853
(0.559)
-0.173
(0.086)
**
-0.223
(0.140)
South Asia
-0.716
(0.721)
-0.119
(0.140)
-0.172
(0.162)
Sub-Saharan Africa
0.344
(0.543)
0.472
(0.185)
**
0.207
(0.241)
Branch Dummies (IFS5)
Egg
0.910
(0.299)
***
0.232
(0.059)
***
0.275
(0.099)
***
Red meat chilled and frozen
-1.011
(0.160)
***
-0.163
(0.037)
***
-0.251
(0.081)
***
Poultry chilled and frozen
-0.010
(0.284)
-0.278
(0.044)
0.024
(0.085)
Fish chilled and frozen
-0.117
(0.208)
-0.088
(0.053)
*
-0.028
(0.110)
Fruits and vegetables
0.301
(0.102)
***
0.020
(0.022)
0.086
(0.041)
**
Dairy
0.046
(0.120)
0.016
(0.026)
-0.031
(0.058)
Meat products and preparations
-0.049
(0.157)
-0.036
(0.025)
-0.003
(0.050)
Fish products and preparations
-0.885
(0.215)
***
-0.096
(0.050)
*
-0.207
(0.105)
**
Ambient stable and hermetically
sealed products
-0.606
(0.139)
***
-0.058
(0.030)
*
-0.159
(0.059)
***
Ready to eat
-0.192
(0.162)
-0.070
(0.027)
***
-0.050
(0.052)
Beverages
0.548
(0.121)
***
0.095
(0.025)
***
0.155
(0.051)
***
Bakery and baked products
-0.086
(0.138)
-0.080
(0.026)
***
-0.016
(0.044)
Dried goods
0.251
(0.146)
*
0.006
(0.033)
0.071
(0.057)
Confectionary
0.034
(0.163)
0.002
(0.034)
0.013
(0.059)
Snacks and breakfast cereals
0.633
(0.278)
**
0.120
(0.046)
***
0.173
(0.083)
**
Oils and fats
0.471
(0.207)
**
0.070
(0.049)
0.127
(0.090)
Food ingredients
0.441
(0.210)
**
0.156
(0.047)
***
0.148
(0.068)
**
Co-Packers (Co-Packing and Han-
dling)
-1.054
(0.301)
***
-0.138
(0.058)
**
-0.270
(0.116)
**
Wholesale
-1.570
(0.171)
***
-0.777
(0.039)
***
-0.490
(0.067)
***
Branch Dummies (IFS6)
Red and white meat, poultry and
meat products
-0.405
(0.196)
**
-0.030
(0.044)
-0.088
(0.107)
Fish and fish products
-1.185
(0.363)
***
-0.167
(0.060)
***
-0.211
(0.115)
*
Egg and egg products
0.956
(0.413)
**
0.137
(0.088)
0.225
(0.115)
**
Dairy products
0.400
(0.300)
0.071
(0.048)
0.094
(0.106)
Fruits and vegetables
-0.591
(0.180)
***
-0.053
(0.039)
-0.117
(0.084)
Grain prod. cereals, ind. bakery
and pastry, confectionary, snacks
-0.606
(0.197)
***
-0.076
(0.039)
*
-0.140
(0.089)
Combined products
-0.709
(0.214)
***
-0.068
(0.045)
-0.159
(0.099)
Beverages
0.279
(0.223)
0.024
(0.047)
0.073
(0.090)
Oils and fats
-0.004
(0.300)
0.068
(0.080)
-0.016
(0.174)
Dry products, other ingredients
and supplements
1.022
(0.298)
***
0.219
(0.056)
***
0.265
(0.094)
***
Pet food
1.248
(0.367)
***
0.096
(0.103)
0.302
(0.113)
***
Pseudo R-squared
0.0107
R-squared
0.1595
Number of observations
24,473
24,347
24,473
Robust standard errors are in parentheses (adjusted for 12,572 clusters according to company
ID for Tobit and GLM and 12,542 clusters for OLS); 57 left-censored observations at percent=0;
69 right-censored observations at percent=100 in the Tobit model; *** significant at 1%; **
significant at 5%; * significant at 10%; 62 certification body dummies considered; Base outcome
of Region is OECD (without Germany, France, Italy and Spain)
94,6
94,8
95
95,2
95,4
95,6
95,8
96
96,2
96,4
96,6
0-13 14-23 24-35 36-49 50-74 75-102 103-157 158-259 260+
average audit results in %
number of employees
Figure A1. Average audit results
96,5
97
97,5
98
98,5
99
99,5
0-13 14-23 24-35 36-49 50-74 75-102 103-157 158-259 260+
share of successful certifications
number of employees
Figure A2. Share of successful certifications
92
92,2
92,4
92,6
92,8
93
93,2
93,4
93,6
93,8
94
94,2
510 20 50 100 250 500 1000 3000 8000
estimated audit results in %
number of employees
Figure A3. Estimated audit results from the Tobit model
Preprint
Full-text available
Final Report on Research Project on the economic impact of BRCGS Food Safety Standards
Article
Full-text available
In recent years, certification has become increasingly relevant for agribusiness. In Europe, substantial parts of the value chain are already being certified by standards such as the International Food Standard (IFS) or GLOBALGAP (the former EurepGap). It is not known, however, whether these approaches can de facto ensure high quality control. This article is based on a database analysis of the German certification system Quality and Safety (QS) and a workshop with the QS-certification bodies conducting 85% of all agricultural audits. It seeks to deduce the first empirical hypotheses concerned with the connection between the reliability of third-party certification and the institutional framing of standards.
Article
Full-text available
Due to the complex structure of certification schemes the risk of flaws and scandals is generally high. It has further increased by several developments during the last years. With regard to their potential effects, it is questionable whether the certification approaches are actually able to detect deficiencies within the system and thus prevent crises which may lead to its breakdown. Hence, the ability of a standard to meet its objectives of food quality and safety needs to be enforced. In this contribution we launch the implementation of a controlling tool which automatically monitors audit quality based on information of the respective data bases. By analysing possible negative influences, opportunistic behaviour can thus be detected.
Article
Full-text available
Due to the growing demands of customers and several food crises, quality assurance schemes have become increasingly popular in agribusiness. With this trend in mind, it seems worthwhile to take a closer look at the satisfaction of participating European companies. The study focuses on the IFS, which has gained much relevance in the food industry. A questionnaire concerned with perceptions of the advantages and disadvantages of the IFS was answered by 389 companies. The results indicate that the overall evaluation of the IFS is primarily affected by the perceived usefulness of the catalogue of requirements and its evaluation. Furthermore, a cluster analysis was conducted and three clusters were identified, representing heterogeneous evaluations of the IFS.
Article
Independent, third-party, involvement can make a significant contribution towards achievement of improved food safety, and food law compliance. Accredited certification organisations, with wide experience and expertise, such as NSF international, are increasingly supporting food industry and regulators by providing alternative, cost-effective services. The key is true independence: Consultants cannot certify, and an effective “due diligence” defence can be based on a certificate of compliance from an organisation without a conflict of interest in consultancy services. Regulators, too, are increasingly turning for help to accredited third-party organisations to achieve their objectives.
Article
The competitiveness of food companies in national and international markets depends upon their ability to adopt production processes which meet food safety and quality requirements. Food safety and quality assurance affect the cost of carrying out transactions, and therein lies the private incentive for adopting voluntary quality assurance systems. While quality assurance systems have the potential to reduce transaction costs by serving as the seller's guarantee of safety or quality, they may also serve as trade barriers.
Article
Regulators and small audit firms allege that audit firm size does not affect audit quality and therefore should be irrelevant in the selection of an auditor. Contrary to this view, the current paper argues that audit quality is not independent of audit firm size, even when auditors initially possesses identical technological capabilities. In particular, when incumbent auditors earn client-specific quasi-rents, auditors with a greater number of clients have ‘more to lose’ by failing to report a discovered breach in a particular client's records. This collateral aspect increases the audit quality supplied by larger audit firms. The implications for some recent recommendations of the AICPA Special Committee on Small and Medium Sized Firms are developed.
Article
The influence of history on an organization is a powerful but often overlooked force. Managers, in their haste to build companies, frequently fail to ask such critical developmental questions as, Where has our organization been? Where is it newt and What do the answers to these questions mean for where it is going? Instead, when confronted with problems, managers fix their gaze outward on the environment and toward the future, as if more precise market projections will provide the organization with a new identity. In this HER Classic, Larry Greiner identifies a series of developmental phases that companies tend to pass through as they grow. He distinguishes the phases by their dominant themes: creativity, direction, delegation, coordination, and collaboration. Each phase begins with a period of evolution, steady growth, and stability, and ends with a revolutionary period of organizational turmoil and change. The critical task for management in each revolutionary period is to find a new set of organizational practices that will become the basis for managing the next period of evolutionary growth. Those new practices eventually outlast their usefulness and lead to another period of revolution. Managers therefore experience the irony of seeing a major solution in one period become a major problem in a later period. Originally published in 1972, the article's argument and insights remain relevant to managers today. Accompanying the original article is a commentary by the author updating his earlier observations.