That’s either a highly flawed or a highly dishonest understanding of Bastiat’s parable of the broken window. My money’s on the latter.
That’s right. One can only observe in cross sectional studies that countries that invested more in public infrastructure grew faster than those that didn’t. But the evidence is not looking very favorable to “privatize everything”. You’ll solve the debt problem but probably tank the economy.
Eh, let’s go back to talking about the argument.
Why should I, when you don’t seem to be playing fair?
What exactly are we arguing about anymore?
After reading over the whole thread it seems like the subject has changed drastically from the OP.
So clearly this thread no longer has a purpose.
Not sure. We did manage to create one of the longest threads on the front page, though.
Fuck this forum for eating my original post.
But you’ve demonstrated that you don’t understand what the parable of the broken window is. So, basically, every time someone has mentioned the broken window fallacy as a counter argument to you, and you then promptly dismiss it, you didn’t actually understand what you were dismissing. No wonder this thread is so long.
I suggest you take the time to read the parable. Perhaps you will then be on the same level as everyone else. Perhaps.
Yes, government activity implies the existence of negative consequences, by definition (so no proof is necessary). There exist only two realms of human interaction: (1) voluntary, free and (2) coercive, political. “Government activity” falls safely within the second realm, under any conceivable definition of the term. It implies the necessary existence of agents who suffer negative consequences from being coerced into predicaments they would have not chosen voluntarily.
This is going back to our earlier discussion. If you believe that taxes for public libraries and sidewalks constitute an immoral use of force, I understand what you are saying. But most of the people who vote to fund these programs apparently don’t consider themselves being robbed.
Autolykos, you asked me to explain some of the statistical techniques I mentioned. I’ll try to do that without involving too much math.
A lagged regression is a method for determining long run correlation. For instance, if one wants to determine the correlation between growth now, at the T, and spending five years from now, at time T-5, the data for growth at time T is simply regressed against the spending at time T-5.
GMM, or Generalized Method of Moments, equates population and sample moments to compute a point estimate for the population correlation coefficient. Moments are the integral of the distances from the mean for each sample data point. GMM requires the population distribution to be assumed to identify population moment parameters, however.
VAR (Vector Autoregression) models the time series autoregressions between variables. For instance, if the previous year’s GDP growth is correlated with the present year’s GDP growth, VAR will control for the effects of this autoregression on the computed result.
Cointegration tests determine whether or not the two variables both cause each other over time, as well as the direction of these causal effects. If cointegration is detected, the variables cause each other. Unidirectional correlation indicates that one variable causes another, but not vice-versa. No correlation indicates both variables follow a random walk relative to each other, with no shared variance.
Other approaches that may be used include linear and nonlinear regression, maxmimum likelihood methods, and rank-order methods, which I can go into more detail on if anyone is interested. They are all basically more involved ways of detecting correlation and time series relationships. Causality is not an enigma. It’s the residual correlation between events at different points in time.
How do they do that apart from measuring correlation over time?
Not to my knowledge. You left out the necessity of holding ceteris paribus.
They don’t. It’s all correlational. That doesn’t mean correlational models must exclude causality.
Residual correlation is correlation left over after controlling for confounding factors. Given a valid model, the independent variables which explain all subsequent variance of the dependent variables (DVs), or explain the variance of intermediate variables, could be said to have caused the DV’s variation (in the particular population under study). The residual correlation results from instrumenting, or holding constant, the intermediate relationships.
The model has to make certain assumptions, including completeness and the law of large numbers. If the actual causal variables are not included, then, of course, there’s a problem.
State spending as a percentage of GNP/GDP/whatever rises in tandem with productivity since the beginning of time. You can’t get blood out of a stone.
What is causality, a little badge to be pinned on the left hand corner of the page? Either causality is established or it isn’t. Causality is established in experiments where all variables but the one being tested are controls
After having parsed through this tangle of verbage, I note “after controlling for confounding factors”. What are they and how do your models control them?
Exactly. Economic history is treated as a natural experiment.
Typical confounding variables include demographics, initial GDP per capita, human capital, and fixed effects (a control for average growth rates within a country or region). It depends on the particular model. Partial correlations are established and the effects of these confounds are extracted from the coefficient of correlation.
The model results are usually walked through step-by-step, with one control variable applied at a time. A result can be considered more valid the more robust it is, or the degree to which it isn’t unduly affected by the application of additional controls.
That’s what I was going to ask. How many variables do these models typically incorporate? What are the types of variables?
Can you explain why you think the two assumptions you mentioned are justified?
How is that in any way sane? You’re taking something that in no way reflects an experiment, and then treating it as if it was such. The whole point of experiments is the ease at which you can replicate them. To take a bunch of snapshots of a constant stream of causal events and then relate particular variables to eachother seems to have no basis whatsoever. And what an insignificant number of variables you mention. What is the basis for this methedology?
http://www.freetheworld.com/regional.html reaches the opposite conclusion. Apparently, you can find data to fit any theory.
Illogically so.
So they set these variables for the model? Assuming their own assumptions on GDP and population growth? And not assuming relevant prices? Come to think of it, there aren’t irrelevant prices because each price determines the decisions of each consumer. How do these models account for that? One variable applied at a time? What is even the point in that case if they’re not going to assume all the controls? Are you describing to me an ‘experiment’ where no variables are controls except the chosen factor and GDP or population growth? And we’ve already discussed GDP’s problems…
What a mess.
Economic freedom, as measured by the Fraser index, is not robustly related to growth, except at low levels of EF. The government spending component in particular doesn’t exhibit significant relation, and includes only government consumption, not government capital expenditure.
Usually a handful, related to history and macroeconomic variables- initial GDP, capital, region, and so on. Although it is possible to correlate just about anything with economic growth.
The law of large numbers holds for a sufficiently large number of observations. It’s usually best to have at least twenty observations. Completeness can never be proven. Models can only be expanded.
Statistics is based on the belief that knowledge can be gained through observation. If one believes that this is an innapropriate approach to economics, then one must propose other methods.
Inflation measures can affect the result, but are not necessarily going to be systemically affected by the IV.
One control variable is not applied at a time; one control is added on at a time. In the final specificaiton, all controls are included. The measurement of relative prices and output are possible concerns. Microeconomic analysis can be conducted to look at physical inputs and outputs to the system.
If microeconomic assessments are more your thing, we can talk about those as well.