Summary
- In this post we conduct a structural equation modeling analysis, as a general case of a multiple regression analysis, to understand what has been moving the price of bitcoin since early 2019.
- To make our discussion more practical, it is based on financial instruments that can be easily traded with a typical brokerage account.
- Our results are consistent with the expectation that the price of bitcoin will continue going up as long as we are in a “nervous” bull market, where things seem to be improving but investors are still looking for ways to protect themselves against a possible economic collapse.
- That is, at least until the collapse happens, when bitcoin may simply go down like almost everything else.
The funds that we use in our analyses
In this post we conduct a structural equation modeling analysis, as a general case of a multiple regression analysis, to understand what has been moving the price of bitcoin since early 2019. To make our discussion more practical, it is based on financial instruments that can be easily traded with a typical brokerage account.
The financial instruments, for which daily prices from January 2019 to April 2020 were used, are: Grayscale Bitcoin Trust (GBTC), SPDR Gold Shares (GLD), iShares U.S. Regional Banks ETF (IAT), Invesco QQQ Trust (QQQ), iShares Silver Trust (SLV), SPDR S&P 500 ETF Trust (SPY), iShares 20+ Year Treasury Bond ETF (TLT), and Financial Select Sector SPDR Fund (XLF).
Bitcoin since 2019
The figure below shows the variation of the GBTC over time. Note the spike in price in late 2017. That was due to a number of factors, not the least of which being a general optimism about the use of bitcoin for payments. At that point some held the perception that bitcoin would make banks and other financial companies obsolete. That turned out to be an incorrect perception, and bitcoin's price dropped steadily until early 2019.
That situation changed around early 2019, when arguably the prevailing perception of bitcoin was that of a store of wealth. Something akin to digital gold, with the advantage of not having to be physically stored or transported. As such, individuals could also use bitcoin to easily move money around the world without paying any fees.
Digital gold, like gold, could presumably be used as a hedge against losses in equities. That is an appeal of bitcoin, which is shared by other instruments, notably silver and treasuries. Silver also has industrial applications. Treasures are essentially US government debt instruments with various maturities.
The results of our analyses
The model in the figure below was created with WarpPLS () as a basis for our analyses. This would be a simple multiple regression model if it were not for the aggregation of funds into three predictor “latent” variables.
The predictor latent variables can be seen as indices: FIN, for financials; HDG, for instruments that can be used for hedging; and MKT, for equities whose price is a general reflection of the stock market in the US.
WarpPLS creates the indices based on the component instruments by linearly matching weights with loadings. Weights are obtained by regressing the index on its components, and loadings by regressing the components on their index.
The figure below shows, among other things, the regressions of GBTC on the tree predictor indices. These are not the weights and loadings mentioned above; they are higher-level coefficients, or “structural” coefficients (in the technical jargon of this area of statistics). Also, they are standardized, which means that they can be directly compared against one another.
These results suggest that GBTC (standing in for the price of bitcoin) is significantly and positively associated with FIN, HDG and MKT. The statistical significance here is indicated by the P values lower than 0.01, meaning that the probability that those effects are not “real” is less than 1 percent.
So, if we assume that the hypothesized directions of causality are correct, then the price of bitcoin goes up when financials go up, and when the market goes up; but much more so when instruments that can be used for hedging (HDG) go up in value. The standardized regression coefficient for the latter association is a much higher 0.41 than those for the other predictors (both approximately 0.15).
There are intercorrelations among the variables, as we can see from the correlations table below. Even though some correlations are high, there is no multicollinearity in our model. We tested this via full-collinearity variance inflation factors.
Also, since the diagonal shows the square roots of the average variances extracted for each variable, one would normally see correlations higher than values on the diagonal if there was multicollinearity. Given that this is not the case, the table also suggests that the model has acceptable discriminant validity.
Conclusion
The fact that the price of bitcoin is positively associated with financials goes against the idea that there is a general perception among investors that bitcoin may make some of the functions performed by banks and financial organizations obsolete. Also, the price of bitcoin being positively associated with the broader market does not provide support for the idea of it being very effective when used as a hedge against market downturns.
Our results do seem consistent with the expectation that the price of bitcoin will continue going up as long as we are in a “nervous” bull market, where things seem to be improving but investors are still looking for ways to protect themselves against a possible economic collapse. That is, at least until the collapse happens, when bitcoin may simply go down like almost everything else.
