Analysing Related Data Using Shared Dimensions

Ravel can import data from numerous sources into one file, which enables easy analysis of the relationships between the data, using their shared dimensions.

Bank of International Settlements —imported into one Ravel file. The two Ravels share the dimensions of

Date , Country and Unit DebtHH__DC , for example, is household debt in domestic currency. All these variables inherit selected dimensions from the source data.

The entries on Units are different—Domestic currency, US$ and Percent of GDP for the debt data, Index of and change in house prices for the House Price data—but those on Country and Date are the same. This allows them—and any Ravels derived from them—to be linked, so that a selection of Country or Date made on one Ravel is also made on the other. This makes it straightforward to analyse relationships between data files.

DebtPriv__DC and DebtPriv__%GDP are already linked, because they are derived from the same data file. When variables are derived from different data files— DebtHH__%GDP is derived from the BIS file “WS_TC_csv_col.csv”, while _HPI_ data comes from “WS_SPP_csv_col.csv”, we need to link the two source Ravels.

You do this by attach those variables to Ravels, selecting the two Ravels using click and drag— hold the left mouse button down on the canvas and then drag the mouse until the desired Ravels are selected (they will be greyed out). Then choose “Link Selected Ravels” from the rightclick menu. If you do this without the mouse being over one of the Ravels, you will get the menu shown in Figure 43:

Figure 43: Linking two or more Ravels together

If your mouse is hovering over one of the Ravels, you will get the menu choice shown in Figure 44, where the menu item “Link specific handles” gives you a more detailed capacity to control which features of the Ravels are linked.

With these two Ravels linked, the household debt to GDP ratio can be plotted against the House Price Index, as shown below. Figure 45 shows household debt and house price data for the USA, but if the selector dot were moved to Spain, it would show the same two data series for Spain.

Figure 45: Comparing related variables from different data files.

Figure 45 illustrates a few more features of Ravel :

The data import procedures and the calculation of GDP have been put in Groups , to reduce clutter. A group is created by click-and-drag selecting of items, followed by using the right-click menu on the selected widgets, and choosing “Group”. Grouping is explained later under Organizing your Data on page 54;
A calliper has been applied to the Date axis. Callipers let you select a contiguous range of information on a Ravel axis. This is a right-click menu option which is accessible when the mouse is hovering over the axis;
The variable DebtHH__%GDP has been “Flipped” on its axis. This, as usual, is a right-click menu option for almost all objects in Ravel (you can also rotate any object via the Rotation field in its Edit form);
The wire connecting the HPI Ravel to the right-hand axis on the plot has been curved. This is done by clicking on a wire and then dragging. Only one curve has been added here, but multiple curves can be added by repeating the process;
A plot has been inserted; and
The plot has been given axis labels. The form to control the appearance of plots is, as usual, a right-click menu item “Options” (double-clicking, which is another way to

access the Edit menu for other entities in Ravel , brings up an enlarged version of the Plot in a separate window).

The comparison of debt to house price data isn’t promising: US house prices rose with rising household debt from 1965 till 2007, and fell with falling household debt from 2007 till 2013; but then house prices rose while household debt fell from 2013 onwards. Superficially, there’s no consistent relationship.

However, some researchers argued during the Global Financial Crisis that it wasn’t the level of household debt and house prices that are causally related, but change in household credit and change in house prices .[5]

The argument is simple. Most houses are bought primarily with mortgage debt, so that change in mortgage debt is the main source of monetary demand for housing. Given how inflexible supply of housing is, change in mortgage debt —which can be defined as “mortgage credit”—is therefore the primary determinant of the level of house prices. This implies a relationship between change in mortgage credit and change in house prices . Biggs, Meyer and Pick’s assertion is that the main factor causing change in house prices is change in mortgage credit .

The BIS doesn’t supply data on mortgage debt, but most of household debt is actually mortgage debt, and unsecured household debt (mainly credit cards and car loans) is a relatively stable share of GDP. Household debt can therefore be used as a proxy for mortgage debt here.

To check this theory out, we need to derive credit—the change in debt—from the debt data. This

uses the backward difference operator , which is on the Scan menu. The next figure shows the steps in using this operator.

Firstly, the variable DebtHH__DC is attached to the Delta-minus operator’s input port. Next, the Edit menu for the Delta operator is activated from the right-click menu, and the operator is applied to the Date dimension; the operator is given the argument 4, which tells the program to compare the value now to the value four quarters earlier (which is one year). Finally, the value of this calculation—for multiple countries and quarters—is assigned to the new variable CreditHH__DC .

divided by GDP to produce ∆ CreditHH%GDP : see Figure 46.

Figure 46: Using the Difference (“Delta”) operator

We can now compare this variable to change in house prices. The relationship between changes in household credit and changes in house prices is obvious in Figure 47.

There are currently two difference operators in Ravel : the backward difference operator

used here, and the forward difference operator . Future releases of Ravel will include time-based difference operators, where the change is specified in terms of date units—day, month, quarter, year—rather than the number of data intervals.

Figure 47: Finding relationships in your data using Ravel

Ravel has calculated these results for every country in the BIS database. These simple and easily read equations in Ravel take the place of tens of thousands of obscure and difficult to read cell reference formulas in a program like Excel.

Footnotes

5 Biggs, M., T. Mayer and A. Pick (2010). "Credit and Economic Recovery: Demystifying Phoenix Miracles": http://ssrn.com/abstract=1595980