0
$\begingroup$

I have seen in time series models that GDP is considered a random variable. At first, I found this troubling because I could not see any random process underlying the measurement. However, the argument for this claim is both sound and odd at the same time: in theory, GDP can take any value depending on political, natural and economic conditions (among other factors that affect this measure). Therefore, every measure of GDP is just a realization of something that could have occurred in a counterfactual world.

If this is so, then almost every measurable thing in social sciences, no matter if the source is a survey or a census, must be a random variable too as those things that affect GDP and make it follow an stochastic process also affect every aspect of human life. Is this correct?

$\endgroup$
2
  • $\begingroup$ Yeah, that's pretty much correct. 🤷 $\endgroup$
    – Galen
    Commented Mar 5 at 3:27
  • $\begingroup$ "Is this correct?" type questions are a poor fit for the SE format. $\endgroup$
    – Giskard
    Commented Mar 5 at 6:46

2 Answers 2

2
$\begingroup$

By the same argument, a die roll is also not random,1 because when I measure the outcome, i.e., look at it, that is much more reliably deterministic than the measuring of GDP.

The models probably do not treat the measurement as random, but the future value of the variable itself. In which case GDP, a component of which is strongly weather dependent agricultural output, is very clearly random from our point of view. (Without magic supercomputers + apriori.)


1Yes yes, a physics super computer might be able to predict all die rolls depending on the exact nature of the universe. This is beside the point.

$\endgroup$
7
  • $\begingroup$ “The models probably do not treat the measurement as random” - this is false in applied economics, you can have look at any empirical work. Even past observations of GDP are considered to be random variable in applied work. You can literally open any work using let’s say regression analysis and GDP as dependent or even independent variable and it will be treated as random variable during the whole process $\endgroup$
    – 1muflon1
    Commented Mar 5 at 18:12
  • $\begingroup$ Disclaimer: I am 100% sure you have more insight into this area than I do, I was a game theorist. However, I am not sure we disagree here. I meant that they do not consider the act of measuring random. In any point of time the variables ahead of that point of time will be treated as random variables. $\endgroup$
    – Giskard
    Commented Mar 5 at 21:05
  • $\begingroup$ what do you mean by act of measuring? Act of measuring does not need to be random, but any empirically observed quantity is a random variable prima facie because it is generated by some unknown data generating process (DGP), What has this to do with game theory? In a theoretical game theory models you probably worked with there arent any DGPs $\endgroup$
    – 1muflon1
    Commented Mar 5 at 21:39
  • 1
    $\begingroup$ Did you read my comment or just skim it? "I am 100% sure you have more insight into this area than I do, I was a game theorist." $\endgroup$
    – Giskard
    Commented Mar 6 at 6:42
  • $\begingroup$ "Act of measuring does not need to be random, but any empirically observed quantity is a random variable prima facie because it is generated by some unknown data generating process (DGP)" I mean, we can juggle with what we mean by the word "act", but I think we are saying pretty much the same thing. $\endgroup$
    – Giskard
    Commented Mar 6 at 6:43
1
$\begingroup$

Formally, mathematically, random variable is not actually a variable but a function that assigns possible values from the probability space to the measurable space.

Or in short variable $Y$ is a random variable if:

$$Y: \Omega \to M$$

Where $\Omega$ is the probability space, and $M$ measurable space.

In applied economics, i.e. statistically speaking, GDP is clearly random variable, as the observed GDP is simply the mapping from the probability space of all GDP values that could exist to some measurable outcome that we measure.

By the way this has nothing to do with social sciences, statistically speaking, even a measured mass of proton is a random variable, since when you try to estimate the mass of proton there is a probability space of various possible masses, and every taken measurement (which will necessarily have some measurement error) can be thought of as a result of mapping from the probability space to measurement space.

$\endgroup$
11
  • $\begingroup$ Hi! Sorry, I don't get your argument. You give a formal definition of random variables and then argue that the randomness of GDP in time series models does actually come only from measurement? $\endgroup$
    – Giskard
    Commented Mar 5 at 8:57
  • $\begingroup$ @Giskard my argument, if its not clear from the answer is as follows; 1. Random variables have certain definition, 2. Recorded GDP satisfy the definition, this is among other things because of measurement error. For example if it helps, Output (which is equivalent to GDP since the output is given by sum of all consumption, investment, government spending and net exports) in undergraduate IS-LM is not a random variable $\endgroup$
    – 1muflon1
    Commented Mar 5 at 10:56
  • 1
    $\begingroup$ Okay, but 1. why is a mathematical definition necessary here? You never use the notions again. Are you padding the answer like one would an academic submission to make it look more sciency? 2. Yeah, undergrad IS-LM model treats both demand and supply as deterministic, so no surprise there. $\endgroup$
    – Giskard
    Commented Mar 5 at 11:30
  • 1
    $\begingroup$ I just don't see how your answer answers the original question I guess? It is "randomness can also follow from measurement errors", just very very long? Do you think this is the most relevant source of randomness in time series models? $\endgroup$
    – Giskard
    Commented Mar 5 at 11:31
  • 1
    $\begingroup$ @Giskard what are you talking about? I usually move comments to chat when there is a lot of back and forth, even if some of it is useful. That is what I was told by other older moderators even outside E.SE that is the right thing to do, even when some comments are useful. I was told that multiple times by senior mods in teachers lounge. Some mods automatically do that for any case where there is automatic flag from a bot. I am also happy to respond to community so I brought your comments back even though a regular mod would probably just kept them in chat anyway. $\endgroup$
    – 1muflon1
    Commented Mar 5 at 21:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.