Analysis of china's currency flow rate from price levels, real GDP and monetary supply
I. Modelling
V = β1 + β2 m 1 + β3p + β4y + m
M1 is a demand deposit for cash and commercial banks and the data are derived from the moderation index = (nominal GDP growth/real GDP growth) x 100 per cent, which is the most macro-measured trend towards general price levels. Nominal GDP and real GDP data are derived from the macro-data base。
Y is real GDP and the data are derived from the macro-data base。
V = nominal GDP/m1 = real GDP x p/m1 received, by classic snow formula。
Spreadsheet:
When we get the data, we do the first equation regression calculation:
The first-time averaged:
Of c = >
The original assumption that the acceptance factor is zero, and the constant items are not relevant to the explained variable, indicates that the model was wrong。
There is no economic significance to the constant items here (no production, no prices, no money supply, no apparent currency circulation, there is no need or possibility for c to exist)
The assumption of a zero regression factor for the constant items is theoretically supported, meaning that the assumption will not bind the regression model and will not affect the results and reliability of the other parameters of the model。
A re-entry equation with no cut-off points can be created, removing constant items from the re-entry process, namely:
V = β2 m 1 + β3p + β4y + m
I'm sorry
The general equation after re-entry is:
M1 = m1 p = GDP reflator (price) y = real GDP v = y*p/m1
( )
(-) (-)
R2==t=20
Ii. Modelling tests and amendments
Purpose: the estimation model is to be consistent with the theoretical premise that, if the classic assumption is violated, the normal economic method of measurement will be ineffective or lead to the wrong conclusion, where the model will need to be set and the parameter pattern amended。
The models are then examined separately for normality, from the relevant tests to the heretical tests:
> , indicating that random disturbance items are assumed to be normal。
Custom test
=
There is not only one degree of correlation, but also two degrees of correlation。
Analysis: there is self-relevance, but this result is not surprising, since the three evis (p) and the monetary supply (m1) are time series data analyses, while economic behaviour in the economic system is time-consuming and volatile with the economic cycle, while increases in m1 and p do not immediately lead to v's rise, but do take several periods to achieve, and therefore there is a certain lag effect, in which case economic data are easily self-relevant。
Inhuman white test
No cross terms:
>obs*r-squared's prob>
The original assumption of acceptance of the difference is that there is no variation。
This model has been tested。
Based on the above, the model presents problems of its own, contrary to the assumption that random error items are irrelevant and require improvement。
Path 1: logarithmic variant (closing the gap between model variables without changing the positive-inverse relationship between variables)
Invt = β2 lnm + β3lnp + β4 ln + + β5 ln v t-1 + m ( constant is not economically meaningful and has been removed)
The result of setting the model to a logarithmic return is not ideal, and the first step of relevance still exists, and failure processes are omitted。
Path 2: broad differential method (the model is set correctly, and we can only use it to eliminate random error items)
Return by equation() gives a residual sequence resid01
In order to get involved, we have a one-time delay in resid01:
Enterls reid01 resid01(-1) in the command bar, get the equation
Known: the relevant coefficient is -
So we change the explanation variables of the original model and the explained variables separately
I didn't modify the y because the impact of y on v would not be too much of a delayed effect as m and p. This is one of the shortcomings of the model, and we do not have sufficient evidence to show how clearly m and p's delayed impact on v or how poorly y's delayed impact on v is)。
So we go back to the modified variables (gv gm gp y) and get:
This is our new return equation, and all the t tests of the regression factors are significant. The normal equation after return is:
Gm1 = m1t - old m1t-1 gp = pt - old pt-1 gv = vt - old vt-1
( )
(-) (-)
R2= = t= 19
This is how it works:
Step one:
Step 2:
Describes how the broad differential model is irrelevant。
And then to him
