Of course you get more EVPs on a cheap recordeer. Trouble is they aren't really EVPs, rather simply artifacts caused by low sample rates, conversion errors, aliasing, IM distortion, and often RF interference from the poor shielding such recorders have.
I personally won't even consider a recording for EVP analysis unless it meets the following criterea:
1. Sample arte of 96 KBPS or greater
2. 24 bit A to D conversion
3. Stereo recording for redundency.
4. WAV or other uncompressed format, not CELP based. No MP3s. These can only be used to post a reduced quality EVP so someone can hear it, they should never be used for actual research.
I'm curious... why do you consider only a digital recorder that records at a sample rate of 96 kbos or greater acceptable? According to the Nyquist-Shannon sampling theorem the perfect reconstruction of a signal is possible when the sampling frequency is greater than twice the maximum frequency of the sound being sampled. Since our audio range is from 12 hz to 20 khz, according to the theorem, perfect reconstruction of the sound is possible at a 40 kbps sample rate (44.1 khz is standard).
The reason I have set this as a standard is because we don't really know what or how an EVP is created. The Nyquist point must as you suggest be at least twice the highest frequency present. However this only applies to a sine wave. To reproduce a square wave without distortion requires the capability to reproduce all harmonics to infinity. Of course that is an impossibility, so for audio applications most accept what is refered to as the pseudo square wave. This is simply the first three harmonics, beyond that the level of distortion becomes acceptable.
Using this criterea the third harmonic of 8 kHz is 24 kHz. The Nyquist point to reproduce 24 kHz is 48 kHz, thus the limitation of 44.1 is somewhat less than 8 kHz. Acceptable for most applications to be sure, but remember we don't know whether EVP follows general audio rules. It may turn out that something beyond 8 kHz is a factor in EVP work. (Just a hypothesis). By raising the sample rate to 96 kHz, it becomes possible to capture audio up to 16 kHz using the same criterea. Thus the third harmonic is captured over much more of the audio spectrum without the high distortion levels occurring in the upper regions. And if in fact something is present there, we capture it too.
Admittedly some of this is theoretical, there may be no reason to push frequency respons above the lower normal voice band frequency. But since we are doing research it is better to capture everything including something you don't need than to capture only a portion and miss something significant. 96 KBPS was chosen as a reasonable compromise between the low sample rates which clearly are insufficient even to capture most voice, and a sample rate of several hundred KBPS which, while optimum, is likely beyond what most could afford. Research may at some point prove a need to go higher, but for now 96 KBPS seems sufficient.
another reason to go higher is that it allows more detailed analysis. While sampling at 2 times the highest frequency is sufficient for simply listening, for research purposes and display on spectrum analyzers a sample rate of 10 times is often advised.
In the end it is how seriously you want to study an EVP. If your only interested in What does it say?, then probably lower sample rates would suffice. But if you intend to do serious analysis, then you need more accurate conversions.