Skip to main content
Search form
  • 00:11

    BEN LAMBERT: My name is Ben Lambert.I'm a researcher at the University of Oxford.I use mathematical, computational,and statistical tools to research into malaria.Specifically, malarial carrying mosquitoes.I also research into kidney development as well asempirical economics.

  • 00:32

    BEN LAMBERT [continued]: In this tutorial, I'm going to provide a short introductionto econometrics methods.I'm firstly going to talk about what ismeant by econometric methods.I'm going to talk about how we can use these methodsand why you might use these methodsover other alternatives.I'm then going to talk about what ismeant by a causal relationship.

  • 00:56

    BEN LAMBERT [continued]: Why it's difficult, often, to identifythese causal relationships.And finally, I'm going to in brief talkabout a method called instrumental variables, whichyou can use to try to identify causal relationships.So what is meant by econometrics?

  • 01:18

    BEN LAMBERT [continued]: It's essentially the application ofstatistical and other empirical techniquesto economic and social data.More recently, some researches havesaid that the important thing about econometricsis to try to identifying causal relationships of interest.The difficulty with real life is it's very difficultto organize proper experiments.

  • 01:44

    BEN LAMBERT [continued]: So what we have to do in the social sciencesis to try and to use data to isolatethe part of the system in which we are interested.Econometrics is a tool that allowsyou to focus in on the relationships with which you'reinterested, trying to forget about the various noisy partsof a system, which are uninteresting.

  • 02:12

    BEN LAMBERT [continued]: So how does one go about doing econometrics?Well, there are various statistical software packages,which people use, some of them including Stata or EViews.But the underlying methods are the sameacross all of these different software tools.And it's these statistical tools that I'm goingto talk about in this tutorial.

  • 02:33

    BEN LAMBERT [continued]: In real life, noise often obscuresthe signal and econometric methodsare a way of getting rid of that noiseand allowing you to focus in on the part of a system with whichyou're interested.So what are some examples of a noisy system?Well, for example, you might be interested in evaluatingthe impact of a particular policy program, whichaims at increasing the length of time which people spentin full-time education.

  • 03:08

    BEN LAMBERT [continued]: There are a number of different factorswhich might influence a person's ultimate wage that they end upcommanding.Parental education, parental income, wealth,and all we're really interested inis trying to evaluate whether if theyspend longer in full-time education,do they tend to earn more.

  • 03:31

    BEN LAMBERT [continued]: Econometrics is a way of removingall of these superfluous factors and allowingus to focus in on the relationship of interest, whichhere is just the relationship between the number of yearsof education and the wage that they earn.

  • 03:51

    BEN LAMBERT [continued]: So for policy analysis, it's very importantto understand if there is a causal relationshipbetween that say, a education policy,and the variable which you're looking to effect.So it might be the individual's wage, their health,or some other measure.In policy, if you can't understandwhether there is actually a causal link between variablesof interest, then it's very hard to evaluate the effectsof a potential policy.

  • 04:23

    BEN LAMBERT [continued]: Hence using econometrics is a really essential toolto allow one to understand whether previous policies havehad any sort of causal effect and henceto gauge what the effects of future policy will be.So what exactly do I mean by a causal relationship?Well firstly, it's important to understand that correlationdoes not imply causation.

  • 04:50

    BEN LAMBERT [continued]: So while there might be a correlationbetween the intelligence of a person and their shoe size,that is purely a correlative measure.The reason that there might be a correlationbetween intelligence and shoe sizeis because as people get older, they tend to wear larger shoesand they tend to get more intelligent.

  • 05:10

    BEN LAMBERT [continued]: So that's purely a correlative relationshipbetween those variables.An example of a causal relationshipmight be that between rain and flood damage.There's a clear causal pathway here,which is very easy to understand.And it's obvious the rainfall does ultimatelylead to a greater risk of flood damage.

  • 05:33

    BEN LAMBERT [continued]: So that's a causal relationship.Econometrics is a tool which allows you to differentiatebetween purely correlative relationships and thosethat are causal.In real life, it's often very difficult to prove causality.

  • 05:54

    BEN LAMBERT [continued]: An example might be the relationshipbetween foreign aids which a given country receivesand their economic growth.So the idea is that we would hopethat foreign aid to a country boosts, for example,their health indicators, perhaps their indicators of education,and also help to improve their institutions.

  • 06:17

    BEN LAMBERT [continued]: If we have a look at the graph of foreign aidagainst economic growth, we see a distinct negativerelationship between these variables.But does this actually mean that thereis a negative causal relationshipbetween foreign aid and economic growth?In other words, does giving foreign aid to a countryactually lead to lower economic growth,or is something else going on here?

  • 06:46

    BEN LAMBERT [continued]: So why does the negative relationshipbetween foreign aid and economic growth notindicate a negative causal relationshipbetween the variables?Well, the idea is that there is some sort of selection biaswhich is happening, which is countriesthat tend to group less and at a slowerrate are those countries which are more likely to receive aidin the first place.

  • 07:14

    BEN LAMBERT [continued]: This selection bias causes there to bea strong negative relationship between foreign aidand economic growth, which is shown in this simple graphthat I've illustrated here.Econometrics is a way of discardingthe negative relationship and allowingus to look at the causal relationship of interest.

  • 07:37

    BEN LAMBERT [continued]: In other words, is there a causal impact of foreign aidon the economic growth of a particular country?An econometric method which is used oftento try to identify the causal relationshipsis instrumental variables.

  • 07:58

    BEN LAMBERT [continued]: I find it more helpful to teach the method of instrumentalvariables by means of a particularly good exampleof it's use, which is Angrist's 1990study of the effects of military participation on lifetimeearnings for Vietnam veterans.So there are various hypotheses whyan individual who's participated in the militarymight actually go on to earn less in a future career.

  • 08:28

    BEN LAMBERT [continued]: They might suffer psychological impactsof being in the military, or theymight have trouble reintegrating back into civilian life.Instrumental variables was used by Angristin this canonical study to try and to test these hypotheses.So why can't we simply draw a graph of whether or notan individual participated in the militaryagainst their earnings at some point in the future?

  • 08:58

    BEN LAMBERT [continued]: Well, the problem here is that a purely correlative relationshipbetween these variables doesn't indicate causation.And the reason for that is that thereis a degree of selection bias happening here.One could hypothesize that those individuals thathave lower lifetime earnings capabilitiesare those that are more likely to volunteerto participate in the military in the first place.

  • 09:26

    BEN LAMBERT [continued]: Hence a purely correlative measureis likely to overstate the strengthof a causal relationship between participatingin the military and future earnings.Instrumental variables is a techniquewhich Angrist used in order to try and to remove the selectionbias effect and figure out the causal effect of interest.

  • 09:50

    BEN LAMBERT [continued]: The fortunate thing for researchersis that the Vietnam War was actuallyquite an interesting social experiment.Males of fighting age were assigned a random numberon the basis of their birth date.And if that random number happenedto be greater than some threshold,then those people were then eligible to be drafted.

  • 10:15

    BEN LAMBERT [continued]: This wasn't actually designed for social scientists,but as it happens, it's a very helpful pseudo-experimentwhich has been carried out which allows us to removethe effect of selection bias.This in turn allows us to evaluatethe causal impact of military participationon lifetime earnings.

  • 10:41

    BEN LAMBERT [continued]: So looking at this graph which I have here,we can break down individuals into three separate categories.Firstly, there were those two groupsof individuals who did participate in the military.There were those who volunteered and those who were conscripted.The last group of individuals arethose who didn't participate in the military.

  • 11:07

    BEN LAMBERT [continued]: Either they didn't volunteer or they weren't conscripted.The problem with a simple comparisonof means between those that participatedin the military and those that didn't isthat the group of those the participatedis made up of both volunteers and conscripts.

  • 11:28

    BEN LAMBERT [continued]: This group are those individuals whoyou might hypothesize are less likely to earnas much as those individuals that were conscriptedbecause there was a reason that these people signed upto participate in the Vietnam War and, more generally,in the military in the first place.Instrumental variables allows us to removethis group of individuals who were volunteers.

  • 11:53

    BEN LAMBERT [continued]: Then what we can do is we can compare the lifetime earningsbetween those individuals who, on the basis of the draftlottery, were conscripted with thosewho were draft intelligible.And because of this randomized nature of the lottery,that particular comparison will indicate a causal relationship.

  • 12:22

    BEN LAMBERT [continued]: In this tutorial, I've spoken about what is exactlymeant by econometrics.I've spoken in brief about how it can be used.I've talked about why it's important to use econometricsto try to identify causal relationships.I then went on to discuss a particular methodwithin econometrics, which is that of instrumentalvariables, which can be used to try to identifycausal relationships in settings wherethere is some sort of semi-randomized naturalexperiment.

Video Info

Publisher: SAGE Publications Ltd.

Publication Year: 2017

Video Type:Tutorial

Methods: Econometrics

Keywords: conscription and volunteerism; economic growth; foreign aid; income; military service; noise; policy analysis; policy impacts; practices, strategies, and tools; veterans; Vietnam War ... Show More

Segment Info

Segment Num.: 1

Persons Discussed:

Events Discussed:



Dr. Ben Lambert explains econometrics methods and the type of research they are best suited for. One method, called instrumental variables, enables social scientists to isolate factors of interest and reduce noise in a system.

Looks like you do not have access to this content.

An Introduction to Econometrics

Dr. Ben Lambert explains econometrics methods and the type of research they are best suited for. One method, called instrumental variables, enables social scientists to isolate factors of interest and reduce noise in a system.