Bioelectrical impedance is based on the fact that the lean tissue of the body is much more conductive due to its higher water content than fat tissue. Electrodes are attached to the body at the extremities and a small 500-800 micro-amp, 50 kilohertz signal measures the body's ability to conduct the current. The more lean tissue present in the body the greater the conductive potential, measured in ohms.
The first commercial bioimpedance units utilized "linear regression" formulas to predict body fat based on biological data input into a single equation. Units which utilize these linear regression equations tend to be somewhat valid for a "normal" population, but under-predict body fat for obese subjects and over-predict body fat of lean subjects. The standard errors of estimate for these equations are ?5% to ?6.4. in normal populations.
As far as calipers, the assumption is that substantial fat is proportional to overall body fat and thus by measuring several sites total body fat may be calculated. There are over 100 different equations available to estimate body fat with the use of skin fold calipers. This wide variety of equations is the problem with the accuracy of this methodology. The validity of skin fold measurements is at best ?6%. Because of the inaccuracy associated with skin fold calipers, many credible organizations such as the U.S. Army and the Los Angeles Police Department have abandoned the use of them.
In my opinion, and from experience, a well done test with bioelectrical impedance is more accurate, though it's a pain in the ass since the hydration status is more of a problem. I use it though. I do it in the morning after going to the bathroom and before eating or drinking anything. While it might not be 100% accurate, doing it under the same conditions will reflect the changes made. This rule will apply to pretty much any bodyfat testing.