Once we have indicated or we have highlight some

issues rise to the quality of composite indicators.

We are going now, just as a help for you,

a sort of overall quality of what we would like to have from a composite indicator.

These are basically five issues that are described in the slide we are showing you now.

Basically, these issues are the following: first of all, a composite indicator,

the mathematical model that this represent in it,

must be a good representation of reality.

Second, the indicators that are present in the model,

they must take large or small values.

It is much more interesting to have

indicators that are very large and very small values than

indicators that are close to zero because basically this variability is going to

help them to understand the final evolution of the composite indicator.

It's important to have a good direction with enough strength.

Third point, which is also important,

is the relative importance of the different partial indicators. What do I mean by this?

This is basically referred to the system of weights.

Weights need really to indicate which is the importance of

these partial indicators in the final result of the composite indicators.

This is really a crucial issue in terms of the choice of

composite indicators and we will take rather quite a lot of time

in the next sections of the course just to describe how

to elaborate the weights and what is

the importance of defining the weights in a proper way.

There is also an important issue related to

the mathematical aggregation we are having in mind.

Probably with the discussion we had before

about the ordinality and the cardinality questions,

you realize already how important are the correct definition

of the mathematical expressions in the composite index and

the different possibilities we have by assuming

cardinality or ordinality and the consequences of that.

Finally, we have not discussed yet about possibly one of the most important issues

within the choice of or the elaboration of composite indicator which are the data.

Data, if we go now to the next slide,

are always difficult and always hard to work with.

Why? There is, first of all,

a first choice we need to make between quality and accuracy.

Sometimes when building up composite indicators,

we would like to have the best data in the world. What does it mean?

We want data that are available soon.

We want data of high quality and we

want to understand really what is going on in the production of this data.

Unfortunately, this does not happen many times and

usually we have to choose or we have to select data not

by its accuracy or its quality but only because

this type of data are available at that time we are

going to have to work the composite indicator.

Let me give you a very simple example that can clarify this.

Imagine we are designing a composite indicator of economic activity,

sort of leading economic indicator,

and we want to have available result for next month.

Unfortunately, of the composite of

the partial indicators we have already designed to be a part of this composite indicator,

some of them are not available for next month.

What can we do then? Do we need to abandon the idea of creating a composite indicator?

Maybe this is not just the case,

maybe what we have to do is to select another data issue

or another data that are not of such best quality we would like to have,

but at some point it's useful for our purpose.

Where do the data come from?

Basically, we have two sources of data in our analysis.

First, we could get data from experiments.

Usually, composite indicators in social sciences are not

that lucky as in experimental sciences so we cannot,

in social science, construct or design experiments.

There is, of course you know,

a branch of economics,

the number experimental economics that does this,

but it's not generalized.

And of course we cannot hope that we are going to have the source of data

from many different dimensions that we need to.

If this is, unfortunately the case,

then we need to go to a second source of data which is what we call the survey data.

Survey data are coming either from private institutions or from public institutions.

Of course, some private institutions have got a lot of reputation and we

definitely can take advantage of their work and use

these data for our elaboration of the composite estimators.

But, in general, our recommendation is to take data from public institutions.

Why? Basically because, usually public institutions when they produce data,

they are sealed, they are stamped with the name of official statistics.

An official statistics has some sort of properties or

conditions that make them much interesting and much reliable to work with.

Just to have a look to the point I'm talking to you about,

please go to the next slide.

We have taken the job to summarize

what we call the European Statistics Code of Practice.

We have tried to summarize the European Statistics Code of Practice. What is this?

This is a code that was approved by the European Commission and is

actually a sort of benchmark of quality of public statistics.

We basically distinguish among

three different dimension on these public statistics on this code.

What has been called the institutional environment,

what has been called the

statistical processes and the dimension finally of statistical output.

In the next slide, you will find what are the main issues related to

institutional environment but its not really of our interest here in this course.

I would like to focus your attention basically on the two other dimensions.

For this, if you please go to the next slide,

we would like to point out that for statistical processes,

the European Commission points out the attention on four different issues.

First of all, sound methodology.

The methodology that it has been designed to produce public statistics,

official statistics must be well-detailed and must be technical enough.

Second, there must be

the corresponding appropriate statistical procedures in terms of sampling design,

in terms of inference.

Third, there must not be excessive burden for the representatives.

That is, the official statistical office

cannot ask you 20 times in differing ways for the same data.

If the official institution has already one data from you in this direction,

he should take this, not take the others.

Finally, there must be an appropriate relationship with the cost of the operation.

Now, let us go to the next slide where we show you

what are these European Code of Practice for the statistical output.

This is mainly the most interesting part in terms of

elaborating the composite indicator because there

is where we will see what are

the appropriate characteristics of data we need to use for these composite indicators.

So basically, the data must be relevant.

Data must be of interest of the public opinion of the different European Union states.

Second, the data must be accurate and must be reliable.

These two words probably they are already familiar to you and I mentioned

them at the beginning of this section of

the course related with the composite indicators.

I will repeat them again many times because they

are crucial when we are talking about production of data.

There is a third issue that is also very important,

which is time and punctuality of the statistic product.

If I have a compromise to obtain a leading economic indicator

for next month and the national bureau of

statistics of my country is going to produce

some indicator I need for this composite indicator,

for this leading economic indicator,

I need to have it on time and this is the compromise the statistical offices is made.

Finally, an important issue is also the transparency and the accessibility

of the data and this is already something you understand why.

So then as a summary of all the section,

I would like to point out basically three items that you should keep in mind.

What do we want from a good composite indicator?

First, we want to have good data to feed it.

Second, we want to have appropriate indicators.

And finally, we want to have an economic framework,

statistical economic framework, that makes somehow indicators fit in the whole frame.