Examples 1-3 could be individually plotted using a heat map.
Yes, it may be reasonable if you have high resolution data. In this representation heat map is an invaluable tool (besides opinable choice of color pair):

However if you do not have high resolution data then an heat map is not the only available solution. Take for example this low-res map:

Which alternatives you have? For low-res data you can use other charts overlaid on map:

Watch as this simple chart conveys three information: location, Internet users and share of world population. You may (should) even transform those bubbles in an area chart (!) or - better - a bar chart to compare individually those two values.
If you use bubbles please also provide numbers, human beings are not good to compare areas and angles. Do you need a proof? In this (I admit widely abused) image:

Try to compare 19.5% slice with 21.2% slice. Which one seems bigger? That example is voluntary faked (more crude words: it honestly lies) but it introduces the point: do not represent data using areas, angles and 3D unless absolutely necessary (and probably 3D is never necessary).
However, what if we wanted to show all 3 data simultaneously?
You may use a bi-variate map but unless you're producing charts just for advertising and not for real usage then I wouldn't do that. Do you want an example of bad approach? Let's enjoy this:

It looks full of useful data but it's actually useless because you can't read it. Human beings eyes are not bad just to compare areas but also to distinguishing color shades; combining them is even worse (I do not even mention people with color blindness, 5~10% of male population). Heat maps are seldom a good solution (alone) for this kind of analysis.
You can pick last bubble graph and transform it into a column/bar chart (background map may even be simplified, as required):

Note that each chart in the grid may be an heat map and you can build a grid (often called Trellis chart or small multiple) for comparison:

AFAIK this kind of representation has been formalized (but it was in use even before, see next paragraphs) by Edward Tufte1:
At the heart of quantitative reasoning is a single question: Compared to what? Small multiple designs, multivariate and data bountiful, answer directly by visually enforcing comparisons of changes, of the differences among objects, of the scope of alternatives. For a wide range of problems in data presentation, small multiples are the best design solution.
What to do if you have a huge amount of data to compare?

And:

These kind of data representation is, for example, widely used in EEG to compare multiple related variables and it's often called cartooning and it may convey an huge amount of information (please ignore this bad example of kNN color interpolation, I guess they had to use a better kernel smoother or - at least - another distance function):

You should not add too much information on the same chart because it will make it harder to read. Each chart should tell one story, anything not immediately necessary should be left out ("A sentence should contain no unnecessary words ... for the same reason that a drawing should have no unnecessary lines and a machine no unnecessary parts"2)
Charts are an invaluable tool for data analysis but they must be effective and not (only) attractive. Information should be (easily) available when required. In your case (without more context about your users and kind of analysis they need to perform) I would use this setup (strip each feature is unnecessary in your case):
- One heat map where you visualize only one component (population, income, houses prices and so on) at time. No multivariate display for heat maps (IMO).
- One separate map (that may be tiled to heat map or replace it) to compare multiple components. Here you use bar charts.
- Possibility to switch between map view and tabular view (example taken from on-line demo of Tableau):

- When you present tabular data do not ever forget to sort them according to story you want to tell (or, at least, an arbitrary but reasonable order). Do not forget users may want to change this order (also dragging & dropping rows to easy comparison).
- If applicable for each visualization method you should have a slider (or equivalent control) to narrow time range.
- Multiple checkboxes (or equivalent control) to include/exclude each serie in comparison.
- If crime is categorized then you may provide a grouping feature to sum them all or to split them (in comparison charts).
- Basic interaction: pan & zoom but also selection and details view: if I have to analyze such data I would click one series in comparison chart and see it in heat map (if tiled). I may also want to pin one heat map to manually build a Trellis chart for single components I want to compare.
- Eventually I may need ability to apply some post-processing. For example I may want to normalize ([0...1] or [0...100%] scale) population and house prices and then combine them to calculate some sort of maps correlation or some spatial statistical analysis (you may want to read The Knox Method and Other Tests for Space-Time Interaction3 and Mantel test).
Note that a human-eyes-aided-visual-correlation may be simply done in this way:
- Make each map monochrome (use one single color).
- Make it transparent (let's say 25%).
- Overlay two (or more) of them.
- Give user ability to change transparency, color and visibility of each map.
Users will see correlations by combined intensity! It's easy and really effective, not much different from same technique already applied to dense scatter plots. Let's take a look, for example, to Image Overlay Using Transparency in MATLAB.
1 Edward Tufte. Envisioning information. Graphics Press Cheshire, CT, USA ©1990. ISBN:0-9613921-1-8.
2 William Strunk Jr. Elements of Style. Ithaca, N.Y.: Priv. print. [Geneva, N.Y.: Press of W. F. Humphrey], ©1918. ISBN:1-58734-060-7.
3 Martin Kulldorff, Ulf Hjalmars. The Knox Method and Other Tests for Space-Time Interaction. Biometrics, ©1999.