WEBVTT

00:00:00.590 --> 00:00:02.830
To do research, we need some data.

00:00:02.830 --> 00:00:06.290
The data comes from what we know as measurement.

00:00:06.290 --> 00:00:10.850
Singleton & Straits define measurement
as a process of assigning numbers.

00:00:10.850 --> 00:00:15.389
And we could also say that it's a process
of assigning numbers according to a known

00:00:15.389 --> 00:00:21.460
rule to be more precise, and measurement can
have different properties.

00:00:21.460 --> 00:00:26.439
The idea of measurement is that we typically
start with some kind of theoretical concepts.

00:00:26.439 --> 00:00:30.300
And then we need to come up with empirical
measures of those theoretical concepts.

00:00:30.300 --> 00:00:34.290
And then we do calculations using those empirical
measures.

00:00:34.290 --> 00:00:36.720
How it actually works, let's take a look.

00:00:36.720 --> 00:00:40.040
I'm going to use this diagram from Mikko Ketokivi's
book.

00:00:40.040 --> 00:00:44.649
And it's based on Bacharach's article
in Academic Management Review.

00:00:44.649 --> 00:00:48.840
To analyze this diagram, or use this diagram,
we need a claim.

00:00:48.840 --> 00:00:56.280
And our claim or our proposition is that naming
a woman as a CEO causes profitability to increase.

00:00:56.280 --> 00:01:01.250
And this is a proposition, it's a level, it's
a claim on the level of theory.

00:01:01.250 --> 00:01:03.829
And it contains two theoretical concepts.

00:01:03.829 --> 00:01:06.350
We have first concept is CEO-gender.

00:01:06.350 --> 00:01:08.899
The second concept is profitability.

00:01:08.899 --> 00:01:15.279
And this proposition here is basically a causal
claim, that female CEO causes profitability

00:01:15.279 --> 00:01:17.079
to go up.

00:01:17.079 --> 00:01:18.849
How do we then actually test this?

00:01:18.849 --> 00:01:22.000
So the idea is that we have a proposition.

00:01:22.000 --> 00:01:24.810
Then we need to have a way of assigning

00:01:24.810 --> 00:01:29.919
numbers to represent profitability and represent
CEO gender.

00:01:29.919 --> 00:01:35.560
Then we make a hypothesis that these representations
or empirical concepts are associated.

00:01:35.560 --> 00:01:38.950
Because we can't really observe causality,
we can only

00:01:38.950 --> 00:01:40.369
observe associations,

00:01:40.369 --> 00:01:44.409
Then we collect some data and we test for
statistical association.

00:01:44.409 --> 00:01:51.130
If we find a statistical association, then
we conclude that we could not reject the

00:01:51.130 --> 00:01:56.009
hypothesis, therefore we found
some evidence for the proposition.

00:01:56.009 --> 00:01:58.889
So how do we actually measure CEO-gender.

00:01:58.889 --> 00:02:01.549
We could have every CEO to go through a medical

00:02:01.549 --> 00:02:05.850
examination and the doctor would determine
their gender.

00:02:05.850 --> 00:02:10.110
Or we could send them surveys, ask them
to report the gender.

00:02:10.110 --> 00:02:12.489
And but that's, that's not practical.

00:02:12.489 --> 00:02:18.650
So how do we go in practice of determining
which CEOs are men which are women.

00:02:18.650 --> 00:02:21.099
One easy way would be looking at their names.

00:02:21.099 --> 00:02:27.500
So the names of CEOs that is public information,
and then we can check whether the name is

00:02:27.500 --> 00:02:32.480
a man's name, or a woman's name, and assign
the gender variable according to that rule.

00:02:32.480 --> 00:02:36.590
So measurement is about assigning numbers
according to the rule.

00:02:36.590 --> 00:02:42.129
Whether the rule is entirely, reliable and
entirely valid is

00:02:42.129 --> 00:02:45.599
probably not, because some names can be used

00:02:45.599 --> 00:02:46.709
for both genders.

00:02:46.709 --> 00:02:52.680
So there's a bit of unreliability,
there's randomness in how a person

00:02:52.680 --> 00:02:53.890
evaluates them.

00:02:53.890 --> 00:02:59.040
And also, there are names that are from different
countries, they might be difficult for us

00:02:59.040 --> 00:03:00.360
to evaluate.

00:03:00.360 --> 00:03:03.250
So we get some data about specific companies.

00:03:03.250 --> 00:03:05.459
So that's the measurement process.

00:03:05.459 --> 00:03:07.569
And then we evaluate.

00:03:07.569 --> 00:03:13.469
How do we evaluate profitability,
we need to define what is our measure, and

00:03:13.469 --> 00:03:16.330
in this case, our measures is return on assets.

00:03:16.330 --> 00:03:20.469
How do we claim that this is a valid measure
of profitability?

00:03:20.469 --> 00:03:24.780
We can claim validity because
ROA is actually, something that investors

00:03:24.780 --> 00:03:29.950
look at when they look at profitability differences
between companies.

00:03:29.950 --> 00:03:35.069
So it is valid in the sense that
that is what people actually use.

00:03:35.069 --> 00:03:39.629
We can also argue validity of ROA, based on
understanding

00:03:39.629 --> 00:03:43.319
that profitability or performance of a company

00:03:43.319 --> 00:03:47.879
is related to how much money or how much profits
that makes.

00:03:47.879 --> 00:03:50.739
And in ROA, the returns is the profits.

00:03:50.739 --> 00:03:56.629
And then assets simply scales, those profits
to be comparable across companies of different

00:03:56.629 --> 00:03:57.629
sizes.

00:03:57.629 --> 00:04:00.370
So we can, we can do two different arguments.

00:04:00.370 --> 00:04:02.239
We can say that this is what people actually

00:04:02.239 --> 00:04:07.670
use for profitability, or we can see that
ROA can be derived from profitability,

00:04:07.670 --> 00:04:09.730
and therefore it's a valid measure.

00:04:09.730 --> 00:04:16.420
Then we collect some data for companies
and specific firms, we calculate some kind

00:04:16.420 --> 00:04:17.769
of associations.

00:04:17.769 --> 00:04:20.019
In this diagram.

00:04:20.019 --> 00:04:21.049
Reliability is here.

00:04:21.049 --> 00:04:23.130
It's a very low level thing.

00:04:23.130 --> 00:04:29.020
Would we get the same data if we collected
the ROAs and CEO-genders again, the next day?

00:04:29.020 --> 00:04:30.020
Probably yes.

00:04:30.020 --> 00:04:31.950
So this is probably highly reliable.

00:04:31.950 --> 00:04:34.110
Reliability also concerns the rating here.

00:04:34.110 --> 00:04:39.950
So if a person rates some companies as
men-led or women-led based on name, would

00:04:39.950 --> 00:04:42.810
that be perfectly reliable?

00:04:42.810 --> 00:04:43.810
Maybe not.

00:04:43.810 --> 00:04:51.949
Maybe 99,95% reliable, but certainly not 100%
reliable, because some names are gender ambiguous.

00:04:51.949 --> 00:04:57.880
And then validity concerns whether we can
claim that actually name is valid information

00:04:57.880 --> 00:04:58.979
of gender.

00:04:58.979 --> 00:05:01.979
We can say that that is because of the tradition
of naming boys in

00:05:01.979 --> 00:05:04.050
one way, girls in another way.

00:05:04.050 --> 00:05:08.759
And then whether ROA is a valid measure
of profitability, which we can claim based

00:05:08.759 --> 00:05:13.530
on the definition of ROA or based
on the fact that ROA is actually used as a

00:05:13.530 --> 00:05:16.810
measure of profitability by for example, investors.

00:05:16.810 --> 00:05:22.669
So, this is in a nutshell the idea of measurement,
we have concepts, and then we derive empirical

00:05:22.669 --> 00:05:25.240
representations for those concepts.

00:05:25.240 --> 00:05:31.440
We collect some data, we analyze association,
if association exists, we conclude that we

00:05:31.440 --> 00:05:33.080
got some support for the theoretical purposes.