On Fri I attended the NBER Asset Pricing coming together (program here) in Chicago, organized past times Adrien Verdelhan too Debby Lucas. The papers were unusually interesting, fifty-fifty past times the high standards of this meeting. Alas the NBER doesn't postal service slides thus I don't accept cracking visuals to exhibit you.
Lars Hansen started alongside the latest inwards the Hansen-Sargent ambiguity / robustness work,Sets of Models too Prices of Uncertainty. Stavros Panageas gave a beautiful discussion, complete alongside powerfulness indicate animations. He characterized the newspaper equally a major advance, for reducing the arrive at of models over which an ambiguous agent looks for the worst representative scenario, too for making that arrive at state-dependent.
In the application, the agent worries that the hateful increment charge per unit of measurement of consumption too the AR(1) coefficient mightiness survive wrong; a to a greater extent than persistent consumption increment procedure is hurtful, too that hurting is to a greater extent than inwards bad times.
I haven't followed this run closely enough. I notwithstanding wonder what the testable implcations are -- how dissimilar is the property pricing model from i inwards which the truthful consumption increment procedure is simply a fleck dissimilar from our estimate, inwards the worst possible way?
Still, it's prissy to see a Nobel Prize winner leading off a conference, too alongside easily the most technical newspaper at that conference, alongside simply about other i (Rob Engle) inwards the audience. That tells you lot something virtually the seriousness of this group. Also, this is serious behavioral finance past times whatever metric -- a disciplined model of probability misperceptions, which is prissy to see.
Robert Novy-Marx presented Testing Strategies Based on Multiple Signals, discussed past times Moto Yogo. We're all familiar alongside the phenomenon that if you lot seek 10 characteristics too choice the best few to forecast returns, t statistics are biased too functioning falls out of sample.
Robert pointed out that if you lot set those best 3 inwards a portfolio, they diversify each other, reducing the in-sample variance of the portfolio, too boosting Sharpe ratios too t-statistics fifty-fifty further.
Many ``smart beta'' funds are doing this, thus the fall-off inwards functioning from backtest to existent coin is relevant beyond academia.
The extent of this bias is impressive. Here is the distribution of t statistics that results when you lot choice the best iii of twenty completely useless signals, too set them inwards a portfolio. Critical values of iv too five exhibit upwards routinely inwards Robert's calculations.
Laura Veldkamp presented her run alongside Nina Boyarchenko, David Lucca, too Laura Veldkamp, Taking Orders too Taking Notes: Dealer Information Sharing inwards Financial Markets. Discussed ably (of course) past times Darrell Duffie. Is it a employment that the dealers who are the prime number bidders at treasury auctions accept been caught talking to each other ahead of the auction? Surprisingly, no: The Treasury tin arrive at the sack come upwards out ahead when dealers part information alongside each other, too investors tin arrive at the sack potentially come upwards out ahead too.
This warms my contrarian economist heart. We know thus piddling virtually how markets work, too regulators are thus quick to saltation on supposedly bad behavor, it's lovely to see a clear too convincing model, that explains the form of second-order too equilibrium effects that economists are proficient at.
Brian Weller presented Measuring Tail Risks at High Frequency, discussed nicely past times Mike Chernov. Brian's basic thought is to run cross-sectional regressions of bid/ask spreads, normalized past times book too depth, on the cross-section of factor betas. Since spreads are larger when dealers are to a greater extent than worried virtually large jumps, this produces a mensurate of time-varying probability x size of such jumps. The mensurate correlates good alongside the VIX.
Michael Bauer presented his newspaper alongside Jim Hamilton Robust Bond Risk Premia discussed really nicely past times Greg Duffee. (My discussion of a previous presentation). This newspaper is genuinely virtually whether macro variables care to forecast bond returns. We're used to "Stambaugh bias:'' if you lot forecast returns alongside a persistent regressor, too the conception inwards the regressor is strongly negatively correlated alongside the conception inwards the return, too thus the near-unit-root downward bias inwards the regressor autocorrelation seeps over into upward bias of provide predictability. But macro variables forecasting bond returns accept innovations nearly uncorrelated alongside the returns, thus that's non much of a problem. Michael too Jim exhibit simply about other problem: alongside overlappping returns, t statistics tin arrive at the sack survive biased downward too.
This led to a pleasant reassessment of bond provide forecasts. Some points that came up: econometrics aside, many provide forecasters don't make good out of sample. Many of the issues are specification issues orthogonal to this econometric point. For example, evaluating the huge forecastability of bond returns from a combination of grade too inflation documented past times Anna Cieslak too Pavol Povala, where the forecasters aspect a lot similar a trend, is genuinely virtually specification too interpretation, non econometrics. I held out the sentiment that the of import usage of my newspaper alongside Monika Piazzesi is the single-factor construction of expected returns, non whether modest main components care to forecast returns. We had a pleasant interchange on whether it's a proficient or terrible thought to run one-year horizon forecasting regressions. I similar them, because they attenuate mensuration error. Raising a weekly autoregression to the 52nd powerfulness yields junk. Greg likes them, too gave a stirring reminder of Bob Hodrick's indicate that you lot tin arrive at the sack include lags of the forecasting variables instead.
Nick Roussanov presented his newspaper alongside Erik Gilje too Robert Ready, Fracking, Drilling, too Asset Pricing: Estimating the Economic Benefits of the Shale Revolution with Wei Xiong discussing. They rails the reaction of stock prices to the shale fossil oil boom. In particular, they showed that stocks which rose on a huge shale annunciation afterwards rose fifty-fifty to a greater extent than equally to a greater extent than proficient shale intelligence came in. Until, equally Wei pointed out, prices collapsed.
Nick also used stock marketplace position value to seek to larn at an justice of the economic science benefits of fracking. It's a worthy effort, but let's recall the difficulties. In a competitive no-adjustment cost world, profits are null too at that topographic point are no abnormal stock returns. Stock capitalization may rise, equally firms number stock to invest. But that measures the value of uppercase invested, non the consumer surplus of shale. Still, the full general thought of mixing property pricing, unloose energy economics, too making economical measurements from stock prices is intriguing.
Jonathan Sokobin, Chief Economist, FINRA presented "An Overview of FINRA Data" which I alas had to miss. I'm delighted anyone from the regime wants us to usage their data!
The AP coming together has a prissy tradition. Usually the most tedious usage of a conference is the author's response to discussant. The AP meetings make away alongside this -- or rather, the writer tin arrive at the sack reply if somebody inwards the audience raises his or her paw too says "I'd similar to listen your response to x." That genuinely happened! But past times too large the AP meetings save fourth dimension too a tradition of really active participation too discussion, too this i was no different.
Lars Hansen started alongside the latest inwards the Hansen-Sargent ambiguity / robustness work,Sets of Models too Prices of Uncertainty. Stavros Panageas gave a beautiful discussion, complete alongside powerfulness indicate animations. He characterized the newspaper equally a major advance, for reducing the arrive at of models over which an ambiguous agent looks for the worst representative scenario, too for making that arrive at state-dependent.
In the application, the agent worries that the hateful increment charge per unit of measurement of consumption too the AR(1) coefficient mightiness survive wrong; a to a greater extent than persistent consumption increment procedure is hurtful, too that hurting is to a greater extent than inwards bad times.
I haven't followed this run closely enough. I notwithstanding wonder what the testable implcations are -- how dissimilar is the property pricing model from i inwards which the truthful consumption increment procedure is simply a fleck dissimilar from our estimate, inwards the worst possible way?
Still, it's prissy to see a Nobel Prize winner leading off a conference, too alongside easily the most technical newspaper at that conference, alongside simply about other i (Rob Engle) inwards the audience. That tells you lot something virtually the seriousness of this group. Also, this is serious behavioral finance past times whatever metric -- a disciplined model of probability misperceptions, which is prissy to see.
Robert Novy-Marx presented Testing Strategies Based on Multiple Signals, discussed past times Moto Yogo. We're all familiar alongside the phenomenon that if you lot seek 10 characteristics too choice the best few to forecast returns, t statistics are biased too functioning falls out of sample.
Robert pointed out that if you lot set those best 3 inwards a portfolio, they diversify each other, reducing the in-sample variance of the portfolio, too boosting Sharpe ratios too t-statistics fifty-fifty further.
Many ``smart beta'' funds are doing this, thus the fall-off inwards functioning from backtest to existent coin is relevant beyond academia.
The extent of this bias is impressive. Here is the distribution of t statistics that results when you lot choice the best iii of twenty completely useless signals, too set them inwards a portfolio. Critical values of iv too five exhibit upwards routinely inwards Robert's calculations.
Laura Veldkamp presented her run alongside Nina Boyarchenko, David Lucca, too Laura Veldkamp, Taking Orders too Taking Notes: Dealer Information Sharing inwards Financial Markets. Discussed ably (of course) past times Darrell Duffie. Is it a employment that the dealers who are the prime number bidders at treasury auctions accept been caught talking to each other ahead of the auction? Surprisingly, no: The Treasury tin arrive at the sack come upwards out ahead when dealers part information alongside each other, too investors tin arrive at the sack potentially come upwards out ahead too.
This warms my contrarian economist heart. We know thus piddling virtually how markets work, too regulators are thus quick to saltation on supposedly bad behavor, it's lovely to see a clear too convincing model, that explains the form of second-order too equilibrium effects that economists are proficient at.
Brian Weller presented Measuring Tail Risks at High Frequency, discussed nicely past times Mike Chernov. Brian's basic thought is to run cross-sectional regressions of bid/ask spreads, normalized past times book too depth, on the cross-section of factor betas. Since spreads are larger when dealers are to a greater extent than worried virtually large jumps, this produces a mensurate of time-varying probability x size of such jumps. The mensurate correlates good alongside the VIX.
Michael Bauer presented his newspaper alongside Jim Hamilton Robust Bond Risk Premia discussed really nicely past times Greg Duffee. (My discussion of a previous presentation). This newspaper is genuinely virtually whether macro variables care to forecast bond returns. We're used to "Stambaugh bias:'' if you lot forecast returns alongside a persistent regressor, too the conception inwards the regressor is strongly negatively correlated alongside the conception inwards the return, too thus the near-unit-root downward bias inwards the regressor autocorrelation seeps over into upward bias of provide predictability. But macro variables forecasting bond returns accept innovations nearly uncorrelated alongside the returns, thus that's non much of a problem. Michael too Jim exhibit simply about other problem: alongside overlappping returns, t statistics tin arrive at the sack survive biased downward too.
This led to a pleasant reassessment of bond provide forecasts. Some points that came up: econometrics aside, many provide forecasters don't make good out of sample. Many of the issues are specification issues orthogonal to this econometric point. For example, evaluating the huge forecastability of bond returns from a combination of grade too inflation documented past times Anna Cieslak too Pavol Povala, where the forecasters aspect a lot similar a trend, is genuinely virtually specification too interpretation, non econometrics. I held out the sentiment that the of import usage of my newspaper alongside Monika Piazzesi is the single-factor construction of expected returns, non whether modest main components care to forecast returns. We had a pleasant interchange on whether it's a proficient or terrible thought to run one-year horizon forecasting regressions. I similar them, because they attenuate mensuration error. Raising a weekly autoregression to the 52nd powerfulness yields junk. Greg likes them, too gave a stirring reminder of Bob Hodrick's indicate that you lot tin arrive at the sack include lags of the forecasting variables instead.
Nick Roussanov presented his newspaper alongside Erik Gilje too Robert Ready, Fracking, Drilling, too Asset Pricing: Estimating the Economic Benefits of the Shale Revolution with Wei Xiong discussing. They rails the reaction of stock prices to the shale fossil oil boom. In particular, they showed that stocks which rose on a huge shale annunciation afterwards rose fifty-fifty to a greater extent than equally to a greater extent than proficient shale intelligence came in. Until, equally Wei pointed out, prices collapsed.
Nick also used stock marketplace position value to seek to larn at an justice of the economic science benefits of fracking. It's a worthy effort, but let's recall the difficulties. In a competitive no-adjustment cost world, profits are null too at that topographic point are no abnormal stock returns. Stock capitalization may rise, equally firms number stock to invest. But that measures the value of uppercase invested, non the consumer surplus of shale. Still, the full general thought of mixing property pricing, unloose energy economics, too making economical measurements from stock prices is intriguing.
Jonathan Sokobin, Chief Economist, FINRA presented "An Overview of FINRA Data" which I alas had to miss. I'm delighted anyone from the regime wants us to usage their data!
The AP coming together has a prissy tradition. Usually the most tedious usage of a conference is the author's response to discussant. The AP meetings make away alongside this -- or rather, the writer tin arrive at the sack reply if somebody inwards the audience raises his or her paw too says "I'd similar to listen your response to x." That genuinely happened! But past times too large the AP meetings save fourth dimension too a tradition of really active participation too discussion, too this i was no different.
