Conclusion: Body density from skin fold measurement has the advantage of simplicity, low cost and reasonable validity with predictions to within 34% for 70% of the population – IF done perfect, with a trained technician, using research quality equipment, and utilizing the appropriate equation for the person(s) studied. Skinfolds are more accurate in the nonobese population and are not valid in obese (BMI over 30) subjects. Most equations are accurate in the range of 45% and the actual reading can be plus or minus 10% of your real fat . The best way to use skinfold measurements is just to add the measures together and compare this number over time. IF, you are going to use an equation, then whenever possible use appropriate age, sex, and weight equations. Estimates of subcutaneous fat can range from 20 to 70% depending on the equation used. For accuracy you only need 34 sites.
There are hundreds of equations that predict percent body fat using skinfolds and the results are all over the place. Errors can be technician, equipment, assumption, or equation related.
TECHNICIAN errors – Test to test repeatability with the same technician can vary by nearly 1mm. The size of the skinfold grasp and time delay in reading measurements can cause significant variation. The improper selection of skinfold sites, experience of the technician, and subject fatness may account for a variation of up to 12 mm on one skinfold site. Skinfolds are hard to obtain in the very obese or very skinny persons.
EQUIPMENT errors – The calipers are supposed to exert a set pressure of 10 grams per square mm. Even among the expensive (over $200) “research and clinical” units such as Harpenden and Lange there can be a difference in readings of 12 mm/site between the two calipers. Most of the equations in the literature have been developed with either the Harpenden or Lange. The cheaper calipers can produce even greater errors.
ASSUMPTION errors – most equations that attempt to predict body fat from skin folds used underwater weighing (UWW) as the reference (criterion) method. UWW assumes that the fat free body has a constant density of 1.100 g/cm 3. However black and Hispanic women have a higher density of 1.106 and 1.105 respectively due to higher bone mineral content (other ethnic groups also have various densities other than 1.1). Hydration is another assumed constant. Yet, obese persons have more extracellular fluid than nonobese and this extra fluid persists even after losing weight. Persons on diuretics, after exercise, alcohol, caffeine, or dehydrated obviously have less water, again violating the basic assumption of the equation.
EQUATION errors – There are literally 100’s of equations to choose from. Equations prior to the mid1970’s tend to be population specific; groups of similar size, age, sex, ethnicity. Using these equations with different populations can overestimate lean groups and underestimate obese groups. Generalized equations are based on a large, heterogeneous sample that varies greatly in age and degree of body fatness. Using the wrong equation can result in errors of 510%.
Factors that influence the choice of equation include: gender, age, ethnicity, and various diseases (osteoporosis, kidney failure, congestive heart disease, edema, dehydration and so on).
Almost universally, the studies that try to identify how many and which skinfold sites to use come down to either 3 or 4 sites as optimal. Each additional site simply adds more measurement error without the benefit of significant new information. In choosing sites it appears that one from the upper body (biceps or triceps) one from the torso (usually subscapular) and one from the thigh (midthigh) works best. The fourth site will usually be the abdomen (suprailiac) or hip.
EQUATIONS
The two most widely used equations are those developed by Durnin and Womersley and Jackson and Pollock. Both of these equations were obtained by comparison to UWW, which measures total body density. Next, we use an equation developed by Siri to convert body density (BD) to percent body fat (%BF).
%BF = [(4.95/BD) – 4.5] X 100.
BD is wholebody density in g/cc. UWW is a 2C model which assumes that body hydration and bone mineral content are both stable. A recent study used data from the Fels Longitudinal Study at Wright University. The subjects are Americans primarily residing in Ohio and predominately Caucasian. This study compared the skinfold measurement against a 4C model instead of the 2C model. If you are going to use equations (and I believe you shouldn’t) then these might be the most accurate equations available for the Caucasian American population. Here are the equations from that study together with the results of the standard equations in the same population.

Equation 
Results [accuracy] compared to 4C +/ 2 Standard deviations (95% confidence) and RMSE [precision] of %BF 
Men 
%BF = 20.94897 + (age x 0.1166) – (height x 0.11666) + (sum4 x 0.42696) – (sum4 2 x 0.00159)

0.18% +/ 9.5%. 4.6% 
Women 
%BF = 22.18945 + (age x 0.06368) – (BMI x 0.60404) – (height x 0.14520) + (sum4 x 0.30919) – (sum4 2 x 0.00099562)

0.25% +/ 9.5%. 5% 
Jackson and Pollock women 
Formula not specified 
6.6% +/ 3.3/16.5. 4.9% 
Durnin & Womersley Men 
Formula not specified 
2.8% +/ 7.2/12.9. 4.9% 
Durnin & Womersley Women 
Formula not specified 
1.8% +/ 8.0/11.4. 4.9% 
Measurement sites were triceps, subscapular, suprailiac, and midthigh, all measured in cm.
This article did not specify which formula from Jackson and Pollock or Durnin & Wormersley they used.
The exact equations from Durnin & Womersley are detailed with different equations for different sex and age groups complete with table lookups. Finally, the referenced paper for Jackson and Pollock has different equations for 3, 4, 5, 6, and 7 different skin fold sites. I have provided the complete equations plus the table as an excel spreadsheet rather than list it here
Some other generalized equations are:
Men, aged 1861
Jackson & Pollock 
D = 1.10938  (0.0008267 x sum of chest, abdominal, thigh) + (0.0000016 x square of the sum of chest, abdominal, thigh)  (0.0002574 x age), 
Jackson, A.S. & Pollock, M.L. (1978) Generalized equations for predicting body density of men. British J of Nutrition, 40: p497504.

Men, aged 1826.
Sloan 
D = 1.1043  (0.001327 x thigh)  (0.00131 x subscapular) 
Sloan AW: Estimation of body fat in young men., J Appl. Physiol. (1967);23:p311315. 
Men, college students
Yuhasz, M.S. 
%BF = (0.1051 x sum of triceps, subscapular, supraspinale, abdominal, thigh, calf) + 2.585, 
Yuhasz, M.S.: Physical Fitness Manual, London Ontario, University of Western Ontario, (1974). 
Female, aged 1855
Jackson, et al. 
D = 1.0994921  (0.0009929 x sum of triceps, suprailiac, thigh) + (0.0000023 x square of the sum of triceps, suprailiac, thigh)  (0.0001392 x age) 
Jackson, et al. (1980) Generalized equations for predicting body density of women. Medicine and Science in Sports and Exercise, 12:p175182. 
Male, aged 18.55
Jackson, et al. 
D = 1.10938  (0.0008267 x sum of chest, abdominal, thigh) + (0.0000016 x square of the sum of chest, abdominal, thigh)  (0.0002574 x age) 
Jackson, et al. (1980) Generalized equations for predicting body density of women. Medicine and Science in Sports and Exercise, 12:p175182. 
Female, aged 1725
Sloan, A.W. 
D = 1.0764  (0.0008 x iliac crest)  (0.00088 x triceps) 
Sloan, A.W., Burt A.J., Blyth C.S.: Estimating body fat in young women., J. Appl. Physiol. (1962);17:p967970. 
Female, college students
Yuhasz, M.S. 
%BF = (0.1548 x sum of triceps, subscapular, supraspinale, abdominal, thigh, calf) + 3.580 
Yuhasz, M.S.: Physical Fitness Manual, London Ontario, University of Western Ontario, (1974). 
%Fat = [4.57/D  4.142] * 100
COMBINATION EQUATIONS
Men, German aged 2667
Ada Garcia 
BFM New (kg) for men = –40.750 + {(0.397 x waist circumference) + [6.568 x (log triceps SF + log subscapular SF + log abdominal SF)]} 
Ada L. Garciaet al. Improved Prediction of Body Fat by Measuring Skinfold Thickness, Circumferences, and Bone Breadths Obesity Research 13:626634 (2005)

Women, German aged 2667
Ada Garcia 
BFM New (kg) for women = –75.231 + {(0.512 x hip circumference) + [8.889 x (log chin SF + log triceps SF + log subscapular SF)] + (1.905 x knee breadth)} 
Ada L. Garciaet al. Improved Prediction of Body Fat by Measuring Skinfold Thickness, Circumferences, and Bone Breadths Obesity Research 13:626634 (2005)

Obese women (BMI >27), age 2040 (study evaluating the equations 
= 1.1267 – 0.0626 x (log sum triceps + biceps + suprailiac) 
Durnin JVGA, Womersley J. Body fat assessed from total body density and its estimation from skinfold thickness: measurements on 481 men and women aged from 16 to 72 years. Br J Nutr 1974;32:77–97. 
Postpubertal boys. Error of 0.5 +/ 3.4 
BF% = 18.88 x (log10 sum biceps, triceps, subscapular and suprailiac) – 15.58

Deurenberg 1990. 
Note: Jackson Pollock formulas best for very lean people. Not recommended for men above 12% and women above 14% body fat.
