so what is being done to make statistical models more accurate in predicting the impact of changes in different policies? is the relatively new branch of economics, behavioural economics, somehow incorporated into the statistical models?
Tarig, Good question. The general answer is a lot is being to understand how various monetary policy actions impact the economy. Models and data have gotten a lot better in the last decade or so. That doesn't mean they are perfect, however. There's a relatively new field, behavioral macroeconomics, that is working to incorporate some of the cool micro level behavioral stuff in macro-level models. See this for example: voxeu.org/article/behavioural-economics-also-useful-macroeconomics
Kevo, An excellent question. There are a number of ways to approach this problem. The easiest way is to allow some time to pass so that data from the policy change become available. Then, the model can be re-estimated and the new regression coefficients observed. But, this can take a long time for enough data to use in the analysis. The next way is to use statistical techniques that allow for regression coefficients to change at some point in time--like when the policy changes. This will allow you to see how the policy changes impact the coefficients more quickly than the first method. Still, this method still needs at least a little post-policy change data to be used. Lucas (and others) suggest that the best way (but it's the most complicated way) to handle a policy change is to model what are sometimes referred to as deep parameters. These parameters arise from decisions by people at the micro level. People's tastes, preferences (for saving vs consumption, for example), etc are modeled to see how a policy change impacts behavior at the micro level. These models are sometimes referred to as micro-foundation models--they build macro level models based on the behavior at the micro level. These are relatively complicated to create. The bottom line is that it's quite hard sometimes to predict how policy changes impact the economy.
@@bentleyuniversityec391mone6 I don't understand why the micro-founded models resolve Lucas' critique, doesn't Lucas' critique also apply at the micro level? At the micro level, would not the policy change also change the behavior of the agents? therefore also the microeconometric model used? Or how is it that microfounding the model manages to capture the change in the behavior of an agent from policy changes? Is it perhaps because with a microfounded model we make explicit the policy instrument within the optimization problem, which captures its behavior, of an economic agent?
@@dskevinperezgarcia Kevo, Your last sentence is the key: "Is it perhaps because with a microfounded model we make explicit the policy instrument within the optimization problem, which captures its behavior, of an economic agent?" Consider this example: suppose marginal tax rates are changed. As per the example in the video, a macro-level model may not properly capture how the MTR changes impact behavior at the individual level. But..with a micro foundation based model (if it's done well, of course) one of the parameters among many is one for how people make the consumption/saving decision. Lots of factors will go into this decision, including how the MTR change impacts the utility of saving vs spending the marginal $1 of income. Such models can be really complex. The first wave of such models were known as representative agent models. Here, the decisions of one person--the agent--are modelled. If all people have the same (or at least similar) utility functions, the micro-based models that use the behavior of the agent in response to policy changes might work well. A more recent class of models uses heterogenous agents. For example, there might be one group of people with a high MPC and another group with low MPC. Tax changes for example might impact the behavior of each group differently. These models are even more complicated.
Mark, this logic is one of the reasons why it was hard to see 2008. For example, when ratings agencies, such as Moodys and S&P were estimating the risk of mortgage backed securities, they used models to simulate the impact of changes in housing prices on the default risk of MBS. However, the models were not able to incorporate the dramatic decrease in housing prices in many parts of the country because the magnitude of the declines had no historical precedent, and so were not incorporated in these models.
beautifully explained! thank you very much
YMJ: You are very welcome! Glad to have helped.
Very clear explanation. Appreciate!
Roman,
Glad to have helped!
so what is being done to make statistical models more accurate in predicting the impact of changes in different policies? is the relatively new branch of economics, behavioural economics, somehow incorporated into the statistical models?
Tarig,
Good question. The general answer is a lot is being to understand how various monetary policy actions impact the economy. Models and data have gotten a lot better in the last decade or so. That doesn't mean they are perfect, however. There's a relatively new field, behavioral macroeconomics, that is working to incorporate some of the cool micro level behavioral stuff in macro-level models. See this for example: voxeu.org/article/behavioural-economics-also-useful-macroeconomics
Nice videa and a good explanation.
Marcelo,
Thanks! Glad you liked it.
So how are macroeconomic models estimated from Lucas's critique?
Kevo,
An excellent question. There are a number of ways to approach this problem. The easiest way is to allow some time to pass so that data from the policy change become available. Then, the model can be re-estimated and the new regression coefficients observed. But, this can take a long time for enough data to use in the analysis. The next way is to use statistical techniques that allow for regression coefficients to change at some point in time--like when the policy changes. This will allow you to see how the policy changes impact the coefficients more quickly than the first method. Still, this method still needs at least a little post-policy change data to be used. Lucas (and others) suggest that the best way (but it's the most complicated way) to handle a policy change is to model what are sometimes referred to as deep parameters. These parameters arise from decisions by people at the micro level. People's tastes, preferences (for saving vs consumption, for example), etc are modeled to see how a policy change impacts behavior at the micro level. These models are sometimes referred to as micro-foundation models--they build macro level models based on the behavior at the micro level. These are relatively complicated to create. The bottom line is that it's quite hard sometimes to predict how policy changes impact the economy.
@@bentleyuniversityec391mone6
I don't understand why the micro-founded models resolve Lucas' critique, doesn't Lucas' critique also apply at the micro level? At the micro level, would not the policy change also change the behavior of the agents? therefore also the microeconometric model used? Or how is it that microfounding the model manages to capture the change in the behavior of an agent from policy changes? Is it perhaps because with a microfounded model we make explicit the policy instrument within the optimization problem, which captures its behavior, of an economic agent?
@@dskevinperezgarcia
Kevo,
Your last sentence is the key:
"Is it perhaps because with a microfounded model we make explicit the policy instrument within the optimization problem, which captures its behavior, of an economic agent?"
Consider this example: suppose marginal tax rates are changed. As per the example in the video, a macro-level model may not properly capture how the MTR changes impact behavior at the individual level. But..with a micro foundation based model (if it's done well, of course) one of the parameters among many is one for how people make the consumption/saving decision. Lots of factors will go into this decision, including how the MTR change impacts the utility of saving vs spending the marginal $1 of income. Such models can be really complex. The first wave of such models were known as representative agent models. Here, the decisions of one person--the agent--are modelled. If all people have the same (or at least similar) utility functions, the micro-based models that use the behavior of the agent in response to policy changes might work well. A more recent class of models uses heterogenous agents. For example, there might be one group of people with a high MPC and another group with low MPC. Tax changes for example might impact the behavior of each group differently. These models are even more complicated.
@@bentleyuniversityec391mone6 thank you very much for taking the time, you helped me to elucidate some issues.
@@dskevinperezgarcia You're more than welcome. Glad to help.
well understood......thankyou 🙏
Dreikanter,
You're very welcome! Glad to have been able to help.
Suddenly the failure to see 2008 is getting clearer.
Mark, this logic is one of the reasons why it was hard to see 2008. For example, when ratings agencies, such as Moodys and S&P were estimating the risk of mortgage backed securities, they used models to simulate the impact of changes in housing prices on the default risk of MBS. However, the models were not able to incorporate the dramatic decrease in housing prices in many parts of the country because the magnitude of the declines had no historical precedent, and so were not incorporated in these models.