What's missing

scose-no-doubt

Veteren member
Messages
4,630
Likes
954
Attempted model as implied by medium term inflation expectations commodity prices and dollar strength.

Any ideas as to how I can improve? I need more variables like Aloe Blacc needs that dollar, yo.
 

Attachments

  • model.JPG
    model.JPG
    35.2 KB · Views: 316
Bond yields are in as I work inflation expectation from spreads and I was first thinking along the lines of inverse of bond prices or something to account for flows into equity but wouldn't it be double counting of sorts? Oil is in which I think is more important than gold.

I think sussing what's missing out from the tech bubble period will be the key. It seems I'm missing out the mania. Been thinking about using economic confidence releases as a proxy but am unsure whether it would be a valid one.
 
For clarity, precisely what is it you are attempting to model: Medium term inflation?

What's your medium term?
 
asian currency crisis is a bit before my time but did it push money into stocks at all? seems to be when the bubble started. Looks like the model overvalues after the credit crunch thingymajig too.
I'm thinking I need to bring in something to do with flows of money or liquidity but I'm stumped.
 
contribute or gtfo zombie thread rapist

OK long term strategy. Ring the girls in the phone boxes, ask them their prices. Consider the people who use these services, people with money to burn.
Consider the reduction in the number of people with money to burn.
Consider this as a dry up of demand. Thus an increase in supply.
Supply overcomes demand therefore prices drop.

Ring the girls in the phone boxes and see what their prices are now.

:LOL::LOL::LOL::LOL::LOL::LOL::LOL::LOL::cool::cool::cool::cool::cool::whistle:whistling:whistling
 
okay I cant help unless you post the significance (t stats or F values) of your independent variables.

i understand if you feel protective over your model but i can guarantee its been done probably 100 times before

Im a part time econometrician if i can be of any use?
 
I don't feel protective at all. I deleted it cos it was giving me a headache but I remember it was based on a commodity basket, DXY, short to med term rates spreads and something else which I can't put my finger on now. It was very simple and comprised what I believed to be the fundamental economic driving forces so nothing ground-breaking. I did however find the dot-com divergence interesting in that it proved (to my layman's eye) that the activity was unhinged from fundamental economic reality - if you use the model's relative accuracy throughout the rest of the sample sans the missing shift factor post 08 - and was driven by something unknown. I think I need to encompass analysis of capital flows into it somewhere and possibly a weighted index of forecast expectation, maybe Fed Philly and the Global equivalents.

Anyway if I ever get around to making another one I'll hit ya up. I wish there were more people on here that find this kind of thing interesting and/or worth the bother :)
 
use think the answer is more analysis, if only I could understand a bit more about the market I could make it. you are clearly an intelligent chap (significantly above av anyway). the real issue (for you) is not understanding the market. the market is a whore and it is out to **** as many people as it can. find a quantifiable edge, test it and trade it.. anything else you are just punting, win lose or draw you are not moving forward. and dont listen to this advice whatever you do
 
This exercise will one day have applications for portfolio risk management... trading-wise I'm starting to get an idea of what I'm after :)
 
Hats off for what you're trying to do, but this looks like a lagging not leading indicator (it makes a right rickets of the peak turn in 2007). There are loads of (quite sophisticated) quant models that basically try and do the same. Only today I was reading the latest version of BAML's 'Global Wave'. That uses seven inputs, industrial and consumer confidence, capacity utilisation, unemployment, producer prices, credit spreads and global earnings revisions. It has something like an 86pc success ratio which isn't bad.
 
I think it was GammaJammer that gave me the little nugget advice that went along the lines of the only way you can get better at building models is to build models. Yes this one is lagging* atm but it's all baby steps as I'm just trying to build a working understanding and a prediction model is not what I'm trying to get out. This one was rather a reverse engineering exercise attempting to pinpoint fundamental drivers at any given time.
I'll be working on new models using forecasting inputs one day when I have the time to manually iterate all the data I need as I just can't afford any serious data subscriptions and when I have the time to attempt to build more models. I have so much going on that it's just not in my list of priorities at the moment. I'll be having a look at that BAML stuff too btw :)

*in a sense it is but if it was really lagging it would x shift
 
Last edited:
Top