----------------- HES POSTING ----------------- 
 
I just want to echo Lawrence Moss's excellent post. "Endogeneity" and  
"exogeneity" are rhetorical conceits -- they are concepts created to help  
understand the real world.  And the concepts ARE useful.  It is a step  
beyond making simple correlations, like noting that fat people don't  
make as much money as slim people.  When I say it that way, it sounds  
as if the "direction of causation" -- another formal term -- is going from  
being fat to not making money.  But ... people with low income tend to  
eat food products that are higher in sodium and fat, and tend not to get  
enough physical exercise (paradoxically, as in bygone centuries it used  
to be the other way around).  So now where is the direction of  
causation?  If you have a statistic that links obesity to low incomes, are  
you finding that people who are obese have trouble getting higher  
paying jobs, or are you finding that people with low incomes tend to  
become obese?   
 
That's when you go to the next step and try to set up a "model" of the  
situation -- to specify a line of causation.   
 
But it becomes a human choice to decide what is "endogenous" to the  
model and what is "exogenous".  And when making that decision -- if  
you are not careful -- you have DECIDED the line of causation already.  
 If I set the model up so that obesity is an EXOGENOUS variable and  
income is the ENDOGENOUS variable, then if I am not very careful in  
how I set the model up (with other variables that contribute to income  
variation) I have committed the same tautology that I would have just  
using a simple correlation.  I will have CREATED the "proof" of  
causation simply by assuming from the beginning what is THERE in  
the first place, and what is then the PRODUCT -- I have DECIDED  
what is the cause and what is the effect.    
 
The whole concept of exogenous and endogenous variables, everything  
that is trying to be done in this sort of effort, is associated with  
Marshall because he was really the first (I think) to incorporate INTO a  
best-selling textbook, and INTO a highly-regarded teaching program,  
the "new" mathematics of physics -- using calculus (and algebra) in a  
particular way. (Not really new in terms of years, but "new" in the sense  
of its use in the history of economic ideas).    
 
If you are going to explain economic behavior in the real world using  
two-dimensional graphs drawn onto a blackboard or printed in a  
textbook -- two of the four major methods of dissiminating new  
information in the university system (the others being verbal lectures  
and laboratory experiments) -- then the concept of exogeneity and  
endogeneity become more than a "rhetorical conceit" -- an idea that  
helps explain things -- it also becomes a "heuristic device" -- a teaching  
tool.    
 
I say that because although economic research has gotten way beyond  
two dimensions, students are still introduced into economists via two- 
dimensional diagrams and two-dimensional thinking:  via the heuristic  
device of algebra, and then calculus.  (The use of "vectors" has had its  
own results -- some of which I find most unfortunate, enabling the  
researcher to "pretend" that clumsy variables don't exist because it all  
folds into the matrix, so reality can look a lot cleaner than we OUGHT  
to know it really is when we are using such complex models -- hope  
that made sense!).   
 
Both algebra and calculus require a left and a right side of the equation.   
And this is where you get the need for ONE type of variable (on the  
right side of the equation) and a DIFFERENT type of a variable (on the  
left side of the equation).   
 
Mathematicians tend to go ballistic when they see the basic two- 
dimensional price and quantity equations that economists draw on  
blackboards to teach freshmen -- because we have our X's and Y's  
reversed, you know.  But that's a story for a different time.   
 
At any rate, it seems to me that there is just not enough self-awareness  
of the degree to which this narrow conceptualization of causation --  
while useful -- easily creates the same types of tautologies that we laugh  
about when sociologists are using "simple correlation."  How do you  
know when you are playing with tautology?  I think it's when the results  
start to diverge GREATLY from the evidence coming in from different  
TYPES Of research.  Any model that comes out with the result that  
"there was no discrimination against women in the nineteenth century  
Anglo-American job market" is deliciously counter-intuitive to many  
economics -- and ridiculously out-to-lunch to everybody else who does  
research in this area. Rather by definition, if the people doing the hiring  
are quite up-front about the discrimination, then how on earth can a  
simple model prove that they're wrong?  ???   I'm waiting for the  
person to derive a model that says slavery never happened ...   
 
So where am I headed with this?  Well, it seems to me that today  
economic THEORY no longer requires such a strict separation of  
exogeneity and endogeneity -- which is a good thing, because in the real  
world you get feedbacks and iterative processes and codetermination  
and all sorts of complications that make it impossible to designate  
something completely exogenous or endogenous, ever.   
 
Our use of math to conceptualize reality has gone past algebra and  
calculus and nineteenth century physics.    
 
BUT -- to a large degree, the economist's use of modeling HASN't, and  
the economist's teaching methods haven't either.   
 
And this is where Marshall comes in again.  (I am leaning HEAVILY  
here on the British historians of economic thought who have written  
about this, BTW -- I didn't come up with this all by myself!) -- it is not  
that Marshall instituted the practice, or popularized the notion, of  
models that have one exogenous and one or many endogenous  
variables -- what he popularized was the concept of "catallectics" (the  
spelling is certainly wrong here) -- not that this was a POSSIBLE way  
to analyze economic problems, but that it was the ONLY way -- that  
we had to stick with models that could use observable variables (like  
price and quantity) so we could "prove" or "disprove" this or that  
theory about economics.    
 
Everybody learned their Popper pretty well, so we all know we are  
supposed to set the model up to "disprove" whatever it is we want to  
disprove to prove whatever we want to prove.  Then Samuelson (I  
believe) pretty much nailed down the whole requirement that only  
observable phenomena counted (in his first textbook) -- and combining  
all that with statistics and you get econometrics, which is where you  
will find the currently used definitions of exogenous and endogenous --  
again, rhetorical conceits designed to "fit" a problem into something  
that can be "modeled" using econometrics, and then we can  
comfortably say we have an "answer."   
 
The problem with all this (I believe) is that our conceptualization of  
economics is mathematically too advanced to fit the requirements of  
equations with a left and a right side.   
 
For the past 40 years (my timing could be off, but this is what it looks  
like to me) there's been an increasing divergence WITHIN economics  
between what economists SAY they are doing with models and what  
they are actually doing, what they SAY they are doing with mathematics  
and what they are actually doing, what they SAY they are doing with  
theory and how theory and reality intersects.   
 
And it seems to me that THIS is a problem for historians of economic  
thought.   
 
So ... going beyond the initial valid query, (and if you have made it to  
the bottom of this overlong message, for which I apologize) -- what  
does the history of economic thought have to say about the statement I  
just made?  When Marshall wrote, there was a unity among the level of  
mathematics he used, the type of models he created, the sorts of  
problems he addressed, and the state of the scholarly literature at the  
time.  (Catallectics was definitely an improvement over economic  
theories based on "national character" or genetic variations in "race" . .  
. )   
 
But that's just not so today.  And the intellectual dilemma that has  
created has been a problem within economics as a profession -- I am  
suggesting -- since about the end of the 1950s, when economists began  
writing seriously about microeconomic topics that could not be solved  
with reference only to "prices" and "quantities" -- but have continued  
to use tools created to serve a type of economics BASED on the  
esssential significance of only those two types of variables.   
 
Mary Schweitzer 
 
------------ FOOTER TO HES POSTING ------------ 
For information, send the message "info HES" to [log in to unmask]