1 / 77

Introduction to Macro Economic Forecasting

Introduction to Macro Economic Forecasting. Macroeconomic forecasting models with econometric design considerations. Dan Hamilton August 9, 2016. Outline Economic Forecasting Estimating Correlation Inertial Models Structural Models Comparison: Inertial vs Structural Uncertainty

garygarcia
Download Presentation

Introduction to Macro Economic Forecasting

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Introduction to Macro Economic Forecasting Macroeconomic forecasting models with econometric design considerations. Dan HamiltonAugust 9, 2016

  2. Outline • Economic Forecasting • Estimating Correlation Inertial Models Structural Models • Comparison: Inertial vs Structural • Uncertainty Confidence Intervals Scenarios / Example • Hybrid Estimation Techniques Error-Correction ARIMAX • A Little Bit of History • Econometrics vs Data Mining/Machine Learning

  3. Who Am I? • Director: MS in Quantitative Economics • One – Year and you are Done! • Learn How to Forecast and Analyze Data • Research: Forecasting Methods, Applied Macroeconomics • Forecasting: Center for Economic Research • and Forecasting • United States • California • Ventura County • An Invitation: Ventura County Economic Forecast conference, includes lunch, on Nov. 10, 10-2, at the Serra Center in Camarillo.

  4. What is our current U.S. forecast?

  5. How do Economists Make Forecasts? • Computer Programs • Data • Crystal Ball • Statistics • Theory • The Story

  6. Computer Programs • Subroutines • More subroutines • The more subroutines, the more crashing • You get the idea… • Why cannot computers read our minds?

  7. Data • Lots of Data • Not the Detail we need • Timeliness • Local Data?

  8. Crystal Ball Get to know the Area or Phenomena you are Forecasting

  9. Statistics • Lies, Damned Lies, and Statistics • Fancy techniques • But we are measuring people who are making decisions!

  10. Statistics “Econometrics recognizes that social behavior is exceedingly complex and that a limited number of variables related together in fairly simple and elegant equations cannot explain the whole of such behavior.” Lawrence Klein, 1953

  11. Theory • Correlations without meaning • How do different sectors & markets relate to each other? • Can you find a proxy for missing data?

  12. The Story • What else is going on in the economy? • Is there greater or less risk at this time? • Is the risk asymmetric? • How can we provide information to the statistical model?

  13. The Key Statistical Building Block of Forecasting is Correlation

  14. Note: Economists Love Running Regressions! 

  15. Sir Francis Galton 1822 to 1911 Papers in the 1880s on correlation, regression, and other related topics

  16. A Fork in the Road There are 2 main ways to proceed with using correlation analysis as a building block for forecasting. We can use “Inertia” or “Structure”

  17. MacroEconomic Characteristics - Production can span multiple quarters - Investment can span many years • Household savings and consumption behavior tends to evolve slowly over time • Trading partners tend to “do business” with each other for many years at a time before changing partners. - Activity is often “path dependent” - Supply-chains have “inertia” - Technology shocks can last many years

  18. Question Given the characteristics on prev. page – do you think that GDP and GDP(-1) are correlated? Levels i.e. “Yt” 99.6% Changes i.e. “Yt – Yt-1“39.6% We can exploit this to build an “inertial”regression model of GDP. This style of regression model does not have X’s, i.e. independent variables.

  19. Inertial Models

  20. What Do Inertial Models Look Like? Three typical styles: AR(2) Yt= c + αYt-1 + φYt-2 + εt MA(2) Yt= d + θεt-1+ βεt-2+ εt ARMA(2,2) Yt = c + αYt-1 + φYt-2 + θεt-1 + βεt-2 + εt … etc

  21. There is an issue of filtering prior to estimating the regressions on the previous page. Why? Macroeconomic data tend to be consistent with a mathematically unstable stochastic difference equation. What to do? Take the first difference: ∆yt = yt - yt-1

  22. That was Univariate

  23. Multi-Variate Inertial Models Also: “Vector Auto-Regressions” or “VAR” In general, a system with N variables and P lags. This example is a system with 3 variables and P lags. Y1t = a10 + a11Y1t-1 + … + a1(NxP)Y3t-P + ε1t Y2t = a20 + a21Y1t-1 + … + a2(NxP)Y3t-P + ε2t Y3t = a30 + a31Y1t-1 + … + a3(NxP)Y3t-P + ε3t Again, if the data is unstable, difference the Y’s.

  24. Structural Models

  25. Structural Models Also Exploit Correlation Do you think that Consumption and Wealth are correlated? Consumption’s correlation with Wealth: Levels 98.6% Changes 47.5%

  26. Structural Modeling: Small Example Y1= α + β1X+ β2Y2+ ε1 Y2= φ + β3X+ ε2 Y3= Y1+ Y2 + X where: Y1is endogenous (stochastic) Y2is endogenous (stochastic) X is exogenous (outside the model) Y3is endogenous (identity) ε1 and ε2 are error terms Your model will compute the forecasts of Y1, Y2, and Y3. But, how is the forecast of X determined? You (!?) can specify its forecast.

  27. Exogenous Forecast Specified by the Forecaster Wow, that’s un-scientific?! Actually, its ok. Why? [a] guidance might be available, e.g. monetary policy: these days the Fed wants the public to know their plans for the next year or so. [b] If you are an expert on what you are forecasting, you might have relevant information that the model cannot see. [c] Make it a “most likely”, which could be thought of statistically as a moving average of history. [d] Use a professional forecast for it, e.g. the price of oil is a common exogenous variable, and the US-EIA provides both short and long term forecasts.

  28. Exogenous Forecast Specified by the Forecaster Are Assumptions the Only Option? No. Use inertial methods to forecast the “x”s.

  29. How Structural Models Work Exogenous Variables Forecast Model Endogenous Variables Forecast

  30. You Can Compute More Than One Forecast Why? Answer the question: what happens if X takes a different future path than what is likely? A common approach: Baseline path for X Baseline forecast Optimistic path for XOptimistic forecast Pessimistic path for XPessimistic forecast

  31. A Scenario is a Coordinated Change to a Set of Exogenous Variables

  32. The Baseline Macro Scenario Baseline Exogenous Forecast (many variables) Model Baseline Endogenous Forecast (many variables)

  33. How Scenarios Work Baseline Exogenous Path Alternate Exogenous Path Forecast Model Baseline Forecast Alternate Forecast

  34. What is a Black Swan?

  35. The Great Recession 2008-Q4 Dow Jones 1.0 - σ GDP 5.6 - σ Ted Spread (norm) 315 – σ !!!! Black Swan!

  36. What is a Black Swan? Baseline Exogenous Path Financial Crises / Deep Recession Force Forecast Model Baseline Forecast Deep Recession Forecast

  37. Scenario Example

  38. Macroeconomic Example What if Spain endured a financial crises, and also, left the Euro Area in a sudden and disorderly manner? There would be a number of immediate implications: [a] there would be a financial crises, with a congruent lack of credit [b] Spain would default on its bonds [c] there would be bank failures, bankruptcies, etc. [d] business investment would plummet [e] household spending would retrench

  39. Brexit Britain is leaving the EU in pre-negotiated manner. This is an orderly exit, not a disorderly exit. What has happened so far? [a] Financial market volatility (but not a crises) [i] Pound fell [ii] FTSE fell [b] Impact of [a]: [i] UK exports are up, [ii] Less investment in inventories/buildings/equipment /intellectual property. [c] UK GDP appears to be headed lower, however, it could have been headed lower anyway, as the EU’s economy is not doing well.

  40. More on Brexit [a] Britain does not actually exit the EU for 2 more years. Key point: this is an orderly exit. [b] Britain has not and will not default on their bonds [c] Germany, for example, needs Britain … many German exports go to Britain … it is not the case that Britain has no bargaining power in this situation. [d] Britain’s trade arrangements pre-date the formation of the EU, by decades, if not hundreds of years. [e] In the long run Britain may benefit from exiting the EU (i.e. better trade arrangements, less red tape).

  41. Brexit: the punchline There may be an impact, driven mostly by jittery financial markets feeding over into the real economy. The macro impact does not appear at this time like it will be very big. Will there be certain companies and industries adversely impacted? Yes.

  42. A Key Point The global economy was already weak and likely to get weaker despite Brexit.

  43. Macro-Model Structure Key Identity: Y = C + I + G + EX – IM C = α + β1(Y – T) + β2(W) + β3(R) + ε T = taxes, W=wealth, R=real interest rate I = α + β1(ΔY(-1)) + β2(Q) + β3(R) + ε Q = market value/cost of capital stock, R=real interest rate EX = α + β1(WGDP) + β2(ER) + ε WGDP = world GDP, ER = exchange rate IM = α + β1(Y – T) + β2(W) + β3(ER) + β4(CP) + β5(WIP) + ε CP = commodity price index, WIP = world intermediate goods price index

  44. What Happened to G?

  45. What are some other common Assumption Variables in Macro forecasting models? Oil Price Federal Funds Target Rate / Other Fed Activities World GDP Commercial Banking Spread Other Commodity Prices

More Related