Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Economic significance

From   Maarten Buis <>
Subject   Re: st: Economic significance
Date   Tue, 19 Jun 2012 09:28:48 +0200

--- On Tue, Jun 19, 2012 at 1:06 AM, FredVer wrote:
> I have a question about the economic significance of a variety of variables
> on a variety on dependent variables. I thought that the economic
> significance is calculated as:
> (standard deviation indepent variable*coefficient indep variable)/standard
> deviation dependent variable. The standard deviations can be interpreted
> from the summary statistics right?
> The thing i think interpretation differs when the variable of itself is
> ratio or logged, right? How to interpret it than? It is really hard to find
> on the net (also only one thread on statlist) how economic significance must
> be calculated. Some also say that you must not use a one standard deviation
> but a one percent?!:S

Please read the Statalist FAQ before posting on Statalist. Here is one
relevant quote:

"It is long-standing practice on Statalist that most members,
especially the most active members who supply a large fraction of the
answers, post using their real names. This is one of the ways in which
we show respect to others. So we discourage you from posting from
behind fake names or identifiers. [...] Please note that many members
are less inclined to answer anonymous emails, sometimes to the point
of ignoring them on principle. These choices are personal, but your
chances of eliciting a good reply are greatly diminished if you write
and conceal your identity."

What you describe is standardization, which may or may not help with
determining economic significance. Economic significance just means
that you interpreted the coefficient and made the subjective decision
that it is so big that it has a meaningful effect.  Standardization
will in some cases help you with that, but in many cases it will make
it harder. The key is that you (and your readers/audience) can
substantively understand the size of the coefficient.

Standardization can be done in various ways, e.g. by standard
deviations, percentages, but also ranks. These are all ways that are
sometimes useful to assign a scale to a variable. If a variable
already has a natural scale (e.g euros, liters, headloads(*), etc. )
than using these methods will just make it harder to interpret your
coefficient. Sometimes a variable has no natural scale, typically this
occurs when the variable is a composite index. In those cases any of
these methods might make sense. Sometimes standardization is used to
compare effects of variables. Now you need to decide if a standard
deviation increase in income is really comparable to a standard
deviation increase in education. In most cases I would argue that they
are just different, and such comparisons make no sense.

-- Maarten

(*) How much one can carry on his or (more common) her head. In some
regions this a useful metric for things like firewood, as this is the
relevant metric for the respondent.

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index