Press "Enter" to skip to content

Class 03

Econ 398 Class 03 of 16 Sept 2015
Empirical issues: leads/lags, exponential-like growth and  regressions
• leads/lags
= PPI versus CPI: does PPI help us know what CPI will do?
› this is a question about leads / lags: we want to know for example whether past PPI predictions inflation today PPIt-6 CPIt which is the same as PPIt CPIt+6
› I don’t know whether you can graph this using FRED
» this is easy to graph in excel, easy to do in stata [not demonstrated in class, but you use for example L1.dlogus instead of dlogus for a 1-period lag]
= basic macro theory (IS-LM) reflects a static equilibrium world. for policy questions we typically care a lot about dynamics: theory alone is insufficient, even if it indicates that there may eventually be an effect in the direction we’d like the economy to move.
= discussion of monetary policy
› FOMC process; historic links to W&L – no national money market in 1900
– not the only issue: also the Panic of 1907
› current monetary policy directly affects the Fed Funds rate, an overnight interest rate
» businesses may have already made most decisions that affect the economy over the next 6 months, cf. example of retail inventory given the time needed from ordering (say) a toy to be made to when it ships, and the time required for shipping consumer goods from China to when they make it onto store shelves
» so monetary policy must be forward looking
• empirical issues: variables that grow
= conceptually, do we expect a [causal] relationship of Japanese consumption on US consumption?
› discussed magnitude of US exports in US GDP, relative size of Japan in global trade and hence likely very weak link between greater Japanese “C”, higher Japanese imports “M”, larger US exports “X”, higher US GDP “Y” and thence higher US consumption
= when we regressed US consumption on Japanese consumption, we got a very large t-statistic and a large R2
= so look at the graphs: both grow a lot
› in general growth correlates with growth so (i) coefficient biased towards 1.0 and (ii) R2 biased towards .99
› so regress growth rate on growth rate
» expected result: numerically small and statistically insignificant coefficient with R2 near zero