The title of this post is a well known reference to how difficult it is to make decisions in the face of uncertainty. This pertains to many important situations in life, including assessing risk in financial markets.
Here's the thing: If you have a decision tree (likely a collection of neurons) set up to handle uncertainty, then for any given tree, you have established what are the contingencies you are prepared to deal with. The "only" uncertainty is the values of your variables. But the tree is set up, and that gives you some comfort, some level of certainty, which all humans need. Having a tree set up and well defined, you are prepared for a set of known unknowns. If a new source of uncertainty enters your framework, e.g. an unfamiliar virus named ebola, and a few cases in your home country, you are now faced with adjusting, or perhaps re-creating your decision tree. This means coming to grips with these "unknown unknowns." This is a resource-intensive process, that accesses parts of the brain which calculate negative outcomes, and costs associated with them. And any time that part of the brain is utilized, it is possible for it to mis-apply information. The reason is that that part of the brain is not a fine-grained machine, it is coarse-grained, and in the same way that monetary policy cannot fix all ills in the economy, and may have side effects, so too the cost-calculation machinery in the human brain cannot redefine a decision tree without some additional costs along the way. Minimize that process, and you will go a long way in improving your ability to make decisions involving risk and uncertainty.
Wednesday, October 29, 2014
Wednesday, September 24, 2014
Jobs Fix
When the slow US economy is discussed, it's often noted that there is still weakness in the labor market, for structural reasons. Generally, this means that due to the combination of drag from the great recession and globalization, i.e. jobs moving overseas for lower prices, there are fewer people participating in the US labor force. However, because of the way we measure things, as well as the larger amount of overall wealth in the US, there are no long lines at the soup kitchen as there were in the 1930s. Most people with employment struggles find a way to downsize their lives, earn some money one way or another, and adjust their lifestyles accordingly. That is not a bad thing, as some may find a simpler life suits them better.
But if you look at the last great economic expansion, from 1940-2001, more or less, there were employers who innovated about what they needed done, how to get it done, and who to hire to do it. Ford and the assembly line comes to mind. Maybe that was earlier, and I'm not saying that repetitive and tedious work is a great way to spend a person's workday, but people will decide for themselves, is this a job I'm willing to do for this price? What appears to be lacking in today's economy are employers who define new roles, and then hire able citizens to do that work. Data and customer behavior prediction look to be the very large employment opportunity from my vantage point.
Having worked as a software developer and quantitative analyst in the financial and trading sector since 1996, I've heard colleagues say that creating a good model and using it to predict useful things, which all companies could do, is about 80% collection of data and 20% of actual modeling. So why are we paying expert modelers to spend 80% of their time doing tedious and repetitive work? Surely companies can define new roles and hire new workers to do that part of the process. Just because the work "looks the same," i.e. you're sitting at a computer and typing on a keyboard, doesn't mean it's not a completely different type of work. The first auto engineers may have done all the work themselves when building the first car, but eventually support roles developed, and many, many more people were employed in those roles.
The same should happen for data mining / predictive analytics / big data, whatever you want to call it. If you're good at modeling, you'd be happy to do it all day long, working with others who can collect, organize, and maintain the data needed. I've seen it in the financial world, it's only a matter of time before the rest of the economic sectors, and companies that make them up, apply that to their own benefit. Let's get to it!
[And Charlie Munger agrees, p. 406 of Poor Charlie's Almanack, re: task delegation: "Even if a manager can perform the full range of tasks better himself, it is still mutually advantageous to divide them up."]
But if you look at the last great economic expansion, from 1940-2001, more or less, there were employers who innovated about what they needed done, how to get it done, and who to hire to do it. Ford and the assembly line comes to mind. Maybe that was earlier, and I'm not saying that repetitive and tedious work is a great way to spend a person's workday, but people will decide for themselves, is this a job I'm willing to do for this price? What appears to be lacking in today's economy are employers who define new roles, and then hire able citizens to do that work. Data and customer behavior prediction look to be the very large employment opportunity from my vantage point.
Having worked as a software developer and quantitative analyst in the financial and trading sector since 1996, I've heard colleagues say that creating a good model and using it to predict useful things, which all companies could do, is about 80% collection of data and 20% of actual modeling. So why are we paying expert modelers to spend 80% of their time doing tedious and repetitive work? Surely companies can define new roles and hire new workers to do that part of the process. Just because the work "looks the same," i.e. you're sitting at a computer and typing on a keyboard, doesn't mean it's not a completely different type of work. The first auto engineers may have done all the work themselves when building the first car, but eventually support roles developed, and many, many more people were employed in those roles.
The same should happen for data mining / predictive analytics / big data, whatever you want to call it. If you're good at modeling, you'd be happy to do it all day long, working with others who can collect, organize, and maintain the data needed. I've seen it in the financial world, it's only a matter of time before the rest of the economic sectors, and companies that make them up, apply that to their own benefit. Let's get to it!
[And Charlie Munger agrees, p. 406 of Poor Charlie's Almanack, re: task delegation: "Even if a manager can perform the full range of tasks better himself, it is still mutually advantageous to divide them up."]
Tuesday, August 19, 2014
Why Do Central Banks Matter?
Why does the Fed matter? Why is the financial community so obsessed with the rate that the Federal Reserve Open Market Committee (FOMC) sets? Here’s how it works, in theory. There is something taught at every business school in the country, called Net Present Value, or NPV. It is something companies do to determine whether to undertake a project. When a business decides yes on a project, then work gets done, people are hired, and Gross Domestic Product goes up.
Somehow or another, you forecast your cash flows, i.e. revenue minus costs. Let’s say you run a clothing company, and you want to determine whether you need to open a new factory. You estimate how many shirts you can sell over the next five years. You multiply by how much you charge for each shirt. That total is the revenue. You have that for each year, quarter, month, or week. You subtract your estimate of the cost to produce them. Now you have an estimate of profit for each month, or week. In order to compare these future cash flows to those of any other project, which might have a different timing, you project all of those cash flows back to the present. There is a standard way of doing this, and everyone does it that way. And in order to do that, you need to have an interest rate at all those times in the future. Each company will have their own interest rate to use, but the base rate, the place where everyone starts, is the interest rate set by the central bank.
So what? So this. When the central bank changes its rate, every company in the country will go into their Excel spreadsheets, change the cell for that rate, and then redo their Net Present Value calculation for all the projects in their hopper. You didn’t know they had a hopper? Oh, they have a hopper. A big one. Just a little comedic pause, here, for the laughter at the Seinfeld reference to Kramer. But laughter aside, this affects the decision of every company in the country, and whether they decide to take on a new project. New projects mean new purchases, also known as capital expenditure, and new labor, which means hiring workers, and income for them, and higher GDP for the country.
Therefore, in order to measure risk in an anticipatory way, you have to pay attention to the moves by the central banks. Period. Full stop.
Somehow or another, you forecast your cash flows, i.e. revenue minus costs. Let’s say you run a clothing company, and you want to determine whether you need to open a new factory. You estimate how many shirts you can sell over the next five years. You multiply by how much you charge for each shirt. That total is the revenue. You have that for each year, quarter, month, or week. You subtract your estimate of the cost to produce them. Now you have an estimate of profit for each month, or week. In order to compare these future cash flows to those of any other project, which might have a different timing, you project all of those cash flows back to the present. There is a standard way of doing this, and everyone does it that way. And in order to do that, you need to have an interest rate at all those times in the future. Each company will have their own interest rate to use, but the base rate, the place where everyone starts, is the interest rate set by the central bank.
So what? So this. When the central bank changes its rate, every company in the country will go into their Excel spreadsheets, change the cell for that rate, and then redo their Net Present Value calculation for all the projects in their hopper. You didn’t know they had a hopper? Oh, they have a hopper. A big one. Just a little comedic pause, here, for the laughter at the Seinfeld reference to Kramer. But laughter aside, this affects the decision of every company in the country, and whether they decide to take on a new project. New projects mean new purchases, also known as capital expenditure, and new labor, which means hiring workers, and income for them, and higher GDP for the country.
Therefore, in order to measure risk in an anticipatory way, you have to pay attention to the moves by the central banks. Period. Full stop.
Friday, July 18, 2014
How to grow GDP
It's not rocket science. This is the only country in the world, as far as I know (I've only visited 20 or so), where, at least for 5-10 major ethnicities, you can find a group of people with a similar background, and feel somewhat at home. Yes, sure, there will be plenty of little tit-for-tats based on some silly difference you have with others, but at least you can find a place and very likely do something productive for the little society you live in. So, all we have to do to grow GDP is to allow more immigrants into the country. If you've taken a serious roadtrip around the country, you'll notice we have lots of space. Some cities, I believe Dayton, Ohio is one, is making it easier for immigrants to move in and start businesses. New businesses will have new employees, with incomes to spend. Leaving your home country and starting fresh is a very motivating experience, and people will work very hard to make that work. After all, didn't your ancestors do that?
http://www.nytimes.com/2013/10/07/us/ailing-cities-extend-hand-to-immigrants.html
http://www.nytimes.com/2013/10/07/us/ailing-cities-extend-hand-to-immigrants.html
Thursday, June 19, 2014
Market Outlook
Now that the June Fed meeting is done, focus will turn to geopolitical events and earnings. I see a choppy summer market, sp could drift up to 2k, or down to 1895; should be in that range unless a clear catalyst emerges. note that a 100 point range is only 5% now, vs 10% when the sp was at 1k. Interesting magazine covers this week - the economist has amazon.com going to mars, and time magazine has "the end of iraq."
Monday, May 26, 2014
How to measure risk?
Risk vs reward measures are typically used for performance evaluation of a strategy or portfolio. But what does risk really mean? Presumably we're talking about risk of loss. But risk of loss means that you actually exit a position at a loss. So really, then, we're talking about the risk of making a decision to take a loss. That means to measure risk using a statistical measure of past prices, like standard deviation (used in sharpe ratio) or negative semi-deviation (used in sortino ratio), you're ignoring any behavioral economic aspects in making that decision.
The implicit assumptions in using a price-derived statistical measure of volatility in a performance evaluation are (1) that we make decisions at fixed-width intervals (because the mathematical definitions of those measures use such intervals), (2) that measure is the same in the past as it will be in the future, and (3) average return compensates for having to tolerate the volatility in real time.
Assumption (1) depends on the freedom of investors (or portfolio manager) to withdraw funds or exit positions. (2) is definitely false, and generally just used for convenience. My main focus for a new measure of risk is on (3). The average return is the best estimate of future return, given that the future is unknown. But surely risk tolerance is path dependent. It's one thing for a portfolio to be up one day, down the next, for e.g. 30 straight days. And it's another for a portfolio to be down for 15 consecutive days, then up for 15 consecutive days; that may incur different behavioral economic aspects for investors. Many risk models assume that prices have no memory, but investors surely do have memory. And no matter what the average or expected return is, I would argue that it's the Probability(return will be x% greater than a volatility-free portfolio) that matters most. And that probability may depend on factors other than the statistics of past prices. Surely neuroeconomics will bring us to a new definition of risk, which does a better job of measuring how we anticipate future outcomes.
The managed macro portfolio is +5.4%, with its worst drawdown at -2.56% on May 15. The drawdown was 1 day after I moved to full leverage, so clearly I was not worried about the possible volatility. If I had outside investors, I would have told them that the probability of having a good return within 1-3 months was 100%. I said over a month ago that the russell will rally back to 1200, that the dax will get to 10k, and they are on their way. The stock market will continue higher this year, with a few bumps and detours along the way, to be sure. As Laszlo Birinyi says, it's like driving from NY to LA. You're sure to get there, and don't sweat the possible traffic or flat tire along the way.
The implicit assumptions in using a price-derived statistical measure of volatility in a performance evaluation are (1) that we make decisions at fixed-width intervals (because the mathematical definitions of those measures use such intervals), (2) that measure is the same in the past as it will be in the future, and (3) average return compensates for having to tolerate the volatility in real time.
Assumption (1) depends on the freedom of investors (or portfolio manager) to withdraw funds or exit positions. (2) is definitely false, and generally just used for convenience. My main focus for a new measure of risk is on (3). The average return is the best estimate of future return, given that the future is unknown. But surely risk tolerance is path dependent. It's one thing for a portfolio to be up one day, down the next, for e.g. 30 straight days. And it's another for a portfolio to be down for 15 consecutive days, then up for 15 consecutive days; that may incur different behavioral economic aspects for investors. Many risk models assume that prices have no memory, but investors surely do have memory. And no matter what the average or expected return is, I would argue that it's the Probability(return will be x% greater than a volatility-free portfolio) that matters most. And that probability may depend on factors other than the statistics of past prices. Surely neuroeconomics will bring us to a new definition of risk, which does a better job of measuring how we anticipate future outcomes.
The managed macro portfolio is +5.4%, with its worst drawdown at -2.56% on May 15. The drawdown was 1 day after I moved to full leverage, so clearly I was not worried about the possible volatility. If I had outside investors, I would have told them that the probability of having a good return within 1-3 months was 100%. I said over a month ago that the russell will rally back to 1200, that the dax will get to 10k, and they are on their way. The stock market will continue higher this year, with a few bumps and detours along the way, to be sure. As Laszlo Birinyi says, it's like driving from NY to LA. You're sure to get there, and don't sweat the possible traffic or flat tire along the way.
Saturday, May 3, 2014
1987 vs 2014
The market has moved roughly sideways, the PL in the model portfolio for the first 3 days has been (+63K, +15K, -20K). Most of the gain was due to entering mid-day before the FOMC announcement. I have no special knowledge about the Fed or any geopolitical events. I use my best judgement based on the aggregate experiences I have accumulated over the past 20 years.
Some market analysts / portfolio managers are looking for some sort of big market crash in the near future. It's possible, it's always possible, that's the nature of financial markets. I've listed all the known unknowns I can think of in a previous post. Today I will make some comments about 1987 vs today.
This is not 1987, this is not 2000, this is not 2008. This is 2014, and as far as I can see today, we are in a stable uptrend / bull market. There are selloffs in a bull market, and every one is an opportunity right now. Leverage must be managed, of course. Any incidence of over-leverage can destroy any strategy based on a correct analysis of the markets. Now, 1987 was the fifth year after a very strong rebound from the 1981-82 recession. The US economy was moving from fast growth and high inflation to slower growth and low inflation. This year, 2014, is the fifth year after a very weak rebound from the 2008 "Great Recession." We have had low inflation for some time, and obviously interest rates are much lower than in 1987. The Dow Jones moved from 1895 to 2722 in a matter of 7 months in 1987, peaking in August. The 10-yr Treasury rate moved from 7% to 10% between April and October 1987, after having moved down over a long stretch in the early and mid-1980s. It's interesting that the SP500 this year has pushed up near 1895 twice, with yet to break and stay above that level. It may soon, unless events unfold to derail the trend.
Additional events from 1987 were some geopolitical events in the middle east, with Iran firing a missile at a US built ship called Sungari. There were no casualties, and the ship did not carry a US flag. Therefore the US made no retaliation, only saying something to the effect that it was an attack on Kuwait. Of course, 3 years later, Gulf War I unfolded. Portfolio Insurance was a new fad back in 1987, which is said to have contributed to the size of the selloffs in October of that year. There was "program trading," which of course is always out there now, with the added uncertainty of "high frequency" and a "flash crash" as in May 2010.
Everyone knows the geopolitical events going on now, and as I've said in previous posts, there are lots of strategies and groups that will add to volatility on that score. Sometimes that allows the markets to climb a "wall of worry" and march higher. I stand firmly in that camp at present.
Some market analysts / portfolio managers are looking for some sort of big market crash in the near future. It's possible, it's always possible, that's the nature of financial markets. I've listed all the known unknowns I can think of in a previous post. Today I will make some comments about 1987 vs today.
This is not 1987, this is not 2000, this is not 2008. This is 2014, and as far as I can see today, we are in a stable uptrend / bull market. There are selloffs in a bull market, and every one is an opportunity right now. Leverage must be managed, of course. Any incidence of over-leverage can destroy any strategy based on a correct analysis of the markets. Now, 1987 was the fifth year after a very strong rebound from the 1981-82 recession. The US economy was moving from fast growth and high inflation to slower growth and low inflation. This year, 2014, is the fifth year after a very weak rebound from the 2008 "Great Recession." We have had low inflation for some time, and obviously interest rates are much lower than in 1987. The Dow Jones moved from 1895 to 2722 in a matter of 7 months in 1987, peaking in August. The 10-yr Treasury rate moved from 7% to 10% between April and October 1987, after having moved down over a long stretch in the early and mid-1980s. It's interesting that the SP500 this year has pushed up near 1895 twice, with yet to break and stay above that level. It may soon, unless events unfold to derail the trend.
Additional events from 1987 were some geopolitical events in the middle east, with Iran firing a missile at a US built ship called Sungari. There were no casualties, and the ship did not carry a US flag. Therefore the US made no retaliation, only saying something to the effect that it was an attack on Kuwait. Of course, 3 years later, Gulf War I unfolded. Portfolio Insurance was a new fad back in 1987, which is said to have contributed to the size of the selloffs in October of that year. There was "program trading," which of course is always out there now, with the added uncertainty of "high frequency" and a "flash crash" as in May 2010.
Everyone knows the geopolitical events going on now, and as I've said in previous posts, there are lots of strategies and groups that will add to volatility on that score. Sometimes that allows the markets to climb a "wall of worry" and march higher. I stand firmly in that camp at present.
Subscribe to:
Posts (Atom)