Data collection and analysis tools for food security and nutrition

 et al., 
2017, 2022). 

---

8 
]
Chapter 1
SETTING
THE STAGE
Italy, 30 March 2022, Italian National Coastguard officer monitors vessels fishing.
© FAO/Cristiano Minichiello

---

[
9
1 
SETTING THE STAGE
B
efore discussing challenges in data 
systems for FSN, it is important to lay 
out the key definitions and conceptual 
framework that will guide this analysis.
DEFINING KEY TERMS
In the title of this report and in the following 
sections we use the concepts of 
data, analysis 
tools and data governance, which may mean 
different things to different readers. A clear 
definition of the way we define and use the terms 
is thus critical to avoid confusion on the intended 
meaning of some of the statements, implicit or 
explicit value judgements, and recommendations 
we present in the rest of the report.
DEFINING DATA
A variety of definitions of data can be found in the 
popular and scientific literature, many of which 
include 
facts, statistics or knowledge, among 
a variety of related terms. Several definitions 
emphasize the numeric aspects, while others 
recognize that data may also take other forms. For 
this report, we adopt a definition inspired by Kitchin 
(2021, p. 2), which states that data are:
 any set of codified symbols representing 
units of information regarding specific aspects 
of the world that can be captured or generated, 
recorded, stored, and transmitted in analogue or 
digital form. 
At initial glance, this phrasing may seem overly 
complex, yet it represents a substantive difference 
from many other existing definitions for at least 
four reasons.
First, the expression 
codified symbols allows a 
meaningful description of data without use of the 
terms 
fact or knowledge. Knowledge and facts are 
indeed inferences that can be gleaned following 
consolidation, analysis and interpretation of data, 
in relation to a specific question in context (Zins, 
2007), but are not, in themselves, data. It is only 
once such inferences are codified, recorded, stored 
and transmitted that they become 
new data, thus 
closing the circle and justifying the image of a data 
cycle that evolves into an ascending spiral where, 
at the completion of each cycle, the amount of data 
and information available for use and re-use grows.
Second, the use of 
codified symbols is 
appropriately inclusive language, as it makes it 
clear that
data do not necessarily need to be 
numeric
. While in many cases data represent 
measured quantities or proportions, thanks to 
the increased digitalization of the information, we 
often deal with datasets consisting of essentially 
qualitative information, stored in the form of 
texts, images, sounds and other forms.
Third, referring to data as 
codified symbols 
has the further advantage of making the 
importance of codifying explicit: 
symbols used 
to record and store data must be chosen 
carefully and their meaning must be properly 

---

10 
]
DATA COLLECTION AND ANALYSIS TOOLS FOR FOOD SECURITY AND NUTRITION
communicated
. One problem that is too often 
encountered in human and social science 
contexts is treating data that are of an essentially 
qualitative nature as if they represent measured 
quantities. The problem arises with indexes or 
scores corresponding to counts of binary (
yes/
no) events, and that are therefore codified with 
integer numbers, which only contain ordinal 
information on the involved cases. In those 
cases, the numeric representation encourages 
an incorrect treatment of such indexes or 
scores, with analysts computing averages or 
other statistics that are only meaningful and 
appropriate for 
interval measures. Such scores 
or indexes should instead be properly treated as 
ordinal measures.
6
More generally, qualitative 
data must be coded following standardized 
coding procedures, which inevitably begins with 
the adoption of 
clear operational definitions 
of the concepts, constructs or attributes 
captured by the data
. This is crucial to avoid 
ambiguity in interpreting data, but not always 
easy to achieve. Contrary to quantitative variables 
reflecting unambiguously defined attributes 
of the physical world (e.g. length, mass, etc.) 
that can be directly observed and measured, 
most qualitative data in social science consist 
of variables and indicators intended to reflect 
concepts or constructs that are not always 
defined unambiguously and understood in the 
same way by everyone. Think, for example, of the 
concepts of 
gender or ethnicity, or constructs 
such as 
poverty or food insecurity. This poses 
several philosophical and practical challenges, 
as even the apparently simple process of just 
recording data, for example, might entail active 
decision-making regarding 
which value to 
record, which may even have moral implications 
(e.g. deciding on a person’s ethnicity simply 
by observing them walking down the street or 
looking at a photograph of them, or on the basis 
of their name, or by asking the respondent’s 
opinion in a survey; or identifying poverty with 
monetary levels of disposable income; or food 
insecurity with inadequate dietary energy intake). 
These considerations point to the importance of 
always accompanying data with clear
metadata
which provides sufficient information on the 
assumptions made in producing them, and of 
ensuring that sufficient competence exists to 
correctly interpret them at all levels of the data 
cycle when the data are used to inform decisions.
The continuing development of sophisticated 
analytic methods, both in statistics and data 
science, necessary for proper treatment of 
non-traditional data, creates a growing need 
for human resources skilled in the use of such 
methods
.
As we shall discuss in more detail in Chapter 
3, and stresses the importance of investing in 
training and education, especially in the current 
era of 
big data
and the new emerging data 
science (see, for example, Oliver, 2021).
Fourth, an important part of the definition of data 
is that data are 
generated, recorded, stored and 
transmitted so that – unless artificial barriers are 
put in place to prevent it – they can be accessed 
repeatedly and by different users at the same 
time at little or no additional cost to the owner 
of the data. This is because,
when data are 
used, they continue to exist and to be available 
and useful
. They are neither appropriated, nor 
consumed. Hence, if we want to ensure their 
efficient use, there are strong arguments for 
promoting as open access as possible to any 
set of existing relevant data. As the issue of 
open access to data may be controversial, and 
in light of the ever increasing amount of data 
being generated and held by private entities 
and the growth of markets for data, we devote a 
specific section to discuss this topic in Chapter 
5, where, we note how the generation of data 
has outpaced the consolidation of relevant moral 
and ethical considerations and their reflections 
in appropriate national and international legal 
arrangements.
DEFINING ANALYSIS TOOLS
Another potentially ambiguous expression used 
throughout the report is 
analysis tool. In the context 
of this report, it is interpreted quite generally as:
6 For an enlightening discussion on the incorrect interpretations 
of counts, indexes and scores as measures in human and social 
sciences, see Wright, 1999.

---

[
11
1 
SETTING THE STAGE
A set of formal rules
7
used to guide the 
processing of available data, aimed at 
obtaining analytic results for a specific 
purpose or research question
.
Several aspects in this definition of analysis 
tool warrant discussion. First, by stressing 
that analysis is conducted on existing data, 
we implicitly distinguish 
data analysis from 
data generation in a conceptual data cycle. We 
recognize that the results of an analysis are 
often, and usefully, stored and remain available 
in the form of new data, so that they can be 
used for further and different analyses. We 
also explicitly recognize that, in some cases, 
existing data may be perceived insufficient 
to address the problem at hand, and may 
therefore lead to a call for generating new 
data. Nevertheless, it is useful to distinguish 
the two steps from a conceptual point of view, 
as – especially in the era of 
big data
– roles 
and responsibilities for data collection, curation 
and dissemination are very often distinct from 
roles and responsibilities in the use of data 
for evidence-based action. The latter entails 
decisions regarding which data to use to inform 
actions aimed at addressing a specific problem 
and how to analyse such data. These decisions 
can be made by agents who have had no direct 
involvement in the collection of 
primary data
.
This leads to another aspect highlighted in 
the definition above, namely that effective 
analysis tools are 
specific, in the sense that 
they must be properly designed to respond to 
well-defined questions. While general analytic 
methods and specific techniques for data 
treatment exist (say, for example, ordinary 
least square methods to estimate parameters 
of a linear regression, used in the context of 
an econometric analysis, or pile-sort methods 
to collect and highlight associations in data 
collected in the context of an anthropological 
study) and are necessary components of any 
7 Rules encompass procedures and techniques belonging to 
different methods of inquiry, both quantitative and qualitative, as 
appropriate, depending on the nature of the data and the objective of 
the analysis.
analytic tool, these should never be confused 
with the analytic tool itself. Insisting on the 
need for specificity of the analytic tool should 
encourage analysts to carefully consider the 
problem at hand and select the kind of data 
needed to answer the question, choosing 
the most appropriate combination of analytic 
methods and techniques for data treatment, 
and – very importantly – present and discuss 
the various assumptions made in setting up the 
analytic model. Unfortunately, we have found 
there to be a discouraging paucity of examples 
of good analysis tools specific to food security 
and nutrition, despite a relative abundance of 
data and of qualitative and quantitative analytic 
methods and techniques.
The final aspect that the above definition 
emphasizes is that the rules that define the 
analysis tool 
must be formalized. That is, they 
should be explicitly and clearly described in 
a way that makes application of the analysis 
tool replicable, consistent and susceptible to 
scrutiny by reviewers.