Market prediction based on macroeconomic indicators - page 47
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
...... I have a strong interest in finding out if crashes can be predicted.
If you were interested in economics AT ALL, then ... you would know that the main buyout of the dwindling and declining SP500 is now .... SP500 companies themselves(!) ...
Nah... Swiss bank of some sort there, according to one visitor to this thread :)
By the way, the companies have been buying their own shares for a long time. It's a PR compaign to tell investors that the stock is cheap and the management has no doubts that the future is bright. Buying shares also has another reason. Companies with big reserves of money (cash in the bank) invest this money like we invest our own money - we want to grow faster than the bank interest, which is now very low. These companies invest the money in other companies. But this is stupid. For example, the company where I work invested money in surgical instruments, internet service providers, and other industries unrelated to our specialty, without knowing anything about those industries. In the end my company was losing that money. Then one day the funds-investors came to us and said one simple thing that shocked me and made me understand the reasons for many things. They said: we-funds are investing in your company by buying your shares because we believe that your business will be successful and will bring us profit. If we wanted to invest our money in surgical instruments, we would find appropriate companies which specialise in that industry and buy their shares without middle-man. Don't waste our money investing in industries where you don't know shit. Invest that money in your own business. That's what we've been doing ever since: buying our own stocks, or buying companies whose business helps our business grow. How this affects recessions is still unclear to me.
About the establishment. I personally don't like governments. I believe that governments exist to limit our will. State borders exist to justify the existence of governments. It's like gangs dividing a city into zones of influence, collecting taxes from the "businessmen" (drug dealers, pimps,...) in those zones, and protecting those "businessmen" from the gangs of other zones. If one gang crosses zone boundaries, wars ensue. Sitting in such a zone and reasoning that my gang is better than yours makes no sense. Those who don't understand that our governments are gangs are worthy of their gang governments. Let us not allow borders between states to set boundaries in our thinking.
Nah... A Swiss bank of some kind, according to a visitor to this thread :)
By the way, the companies have been buying their shares for a long time. It's a PR compaign to tell investors that the shares are cheap and the management has no doubts that the future is bright. Buying shares also has another reason. Companies with big reserves of money (cash in the bank) invest this money like we invest our own money - we want to grow faster than the bank interest, which is now very low. These companies invest the money in other companies. But this is stupid. For example, the company where I work invested money in surgical instruments, internet service providers, and other industries unrelated to our specialty, without knowing anything about those industries. In the end my company was losing that money. Then one day the funds-investors came to us and said one simple thing that shocked me and made me understand the reasons for many things. They said: we-funds are investing in your company by buying your shares because we believe that your business will be successful and will bring us profit. If we wanted to invest our money in surgical instruments, we would find appropriate companies which specialise in that industry and buy their shares without a middle-man. Don't waste our money investing in industries where you don't know shit. Invest that money in your own business. That's what we've been doing ever since: buying our own stocks, or buying companies whose business helps our business grow. How this affects recessions is still unclear to me.
About the establishment. I personally don't like governments. I believe that governments exist to limit our will. State borders exist to justify the existence of governments. It's like gangs dividing a city into zones of influence, collecting taxes from the "businessmen" (drug dealers, pimps,...) in those zones, and protecting those "businessmen" from the gangs of other zones. If one gang crosses zone boundaries, wars ensue. Sitting in such a zone and reasoning that my gang is better than yours makes no sense. Those who don't understand that our governments are gangs are worthy of their gang governments. Let us not allow borders between states to set boundaries in our thinking.
By 1985, it was possible to automatically generate the structure of relational databases by describing economic processes in the language of economic indicators.
San Sanych, in other words, already in 1985 economists learned to store data in the form of a table? :)
I would elaborate on describing processes in the language of indicators and then selecting meaningful indicators, for example. Well, or something similar...
San Sanych, in other words, already in 1985 economists learned how to store data in table form? :)
I would elaborate on describing processes in the language of indicators and then selecting meaningful indicators, for example. Well, or something similar...
There was no table, the 3rd normal Codd form.
That's all in the past. I posted it to demonstrate the problem of identifying meaningful relationships between economic concepts. In reality, the meaningful relationships are fundamental when using statistics. If we are unable to give economic content to the results of statistical calculations, then it is just a numbers game.
AUDNZD, D1.
NS
Training sample - 2009-2014.
FORWARD WITHOUT TRAINING - 2015-2016.
Total of two variables - AUD and NZD UnRate
I doubt it would be possible to build a practically valuable model on such a vague teacher.
Let's take ZZ as the teacher - the most primitive teacher. Up = 1, down = 0.
Learn a model in which these 0s and 1s are predicted. Getting a prediction error of about 40% is almost always possible.
But does such a result have any practical value?
No.
The point is that the teacher was not neatly defined. We have 0 and 1 marked a POINT on ZZ, not the entire ZZ shoulder. Therefore, our 40% error is not an error in the definition of the ZZ arm, but an error in the currents in that arm, and those points will be mixed up within the links. The result is a 100% error in the prediction of the NC link.
Therefore.
Teacher, prediction and usage must be strictly the same. And any inaccuracies in the teacher's formulation will instantly lead in practice to destructive depot results.
Continuing on about data conversion. As I said before, the best transformation in my opinion is the normalised increments:
y[i] = (x[i] - x[i-1])/n[i]
where x[] is the input series, n[] is the normalizing values. Since the incidence of many economic inputs increases with time, the normalizing values also need to increase with time to give us a more or less stationary series y[]. A simple method for calculating n[] is a running average, such as the EMA, of the absolute values of the increments. But the EMA is very non-smooth and makes it difficult to recalculate the predicted values of y[] back to x[]. For example, suppose we predict an unknown future value of y_gdp[k] based on some transformed input delayed by one quarter, y[k-1], using a linear model
y_gdp[k] = a + b*y[k-1]
y_gdp is the transformed series of the original series x_gdp, i.e.
y_gdp[k] = (x_gdp[k] - x_gdp[k-1])/n_gdp[k]
From this equation, we can find the unconverted future prediction of GDP at the k-th step
x_gdp[k] = y_gdp[k]*n_gdp[k] + x_gdp[k-1]
In this equation, y_gdp[k] we know (predictive model output) , x_gdp[k-1] we also know (previous known GDP value) and n_gdp[k] we do not know (future normalising value). But we know the previous normalizing value n_gdp[k-1], which is calculated as the running average of the known past values. To get n_gdp[k] we can for example extrapolate the known previous values n[k-1], n[k-2], ... But the EMA is rather difficult to extrapolate as it is highly non-smooth. We can compute n[i] as a smooth FIR filter of absolute increments |x[i] - x[i-1]|. But there will be a Period/2 group delay and also the first Period values of |x[i] - x[i-1]| will not be involved in the modelling. A third possibility is a Hodrick-Prescott or Savitsky-Golay filter. These filters are smooth, well extrapolated, and without delay, but look ahead, i.e. their last values will be redrawn as new data arrives. Below is the plot of absolute increments of GDP (blue line), the Hodrick-Prescott filter plotted over the entire history (red line), the ends of Hodrick-Prescott filters plotted over each past interval 1..k where k = 2..N (black line), EMA (purple line).
Although the Hodrick-Prescott filter is overdrawn (the tail wags), I believe it has little effect on the accuracy of the predictions of the normalized GDP increments, because those increments (the blue line) change much faster and more strongly than the red line. The red line essentially shows the historical trend on a large scale independent of short-term fluctuations in GDP growth, including recessions. The problem will arise when we choose a small lambda of this filter and it starts to track the magnitude of high frequency fluctuations. For example, in the extreme case where this filter n[i] tracks exactly the blue line |x[i] - x[i-1]|, i.e. n[i] = |x[i] - x[i-1]|, the normalization of GDP increments x[i] - x[i-1] by such a filter will lead to a binary series y[i] = +/-1 and all information about the increment size will be contained in the normalizing values of n[i]. In this case, we can also build a predictive model, but it will only predict the sign of GDP growth, which is good enough to predict recessions (two consecutive quarters of negative GDP growth).