Height Calculator – Predict Your Child's Adult Height

What an adult height prediction can and can't tell you

"How tall will my child be?" is one of the most common questions parents ask a pediatrician, and it has a surprisingly long scientific history. The calculator above answers it with three published methods: the Khamis-Roche equations, mid-parental (target) height, and a projection along the CDC growth curves. Each uses different information and each has known strengths and blind spots.

The honest summary is this: for a healthy child of average build and average timing of puberty, a good prediction is usually within about 5 cm of the eventual adult height, but not always. Predictions are statistical statements about groups of similar children, not promises about one child. This guide walks through where each method came from, what the numbers mean, how accurate they are in published studies, and how to get the best inputs.

A short history of height prediction

Height prediction grew out of the great longitudinal growth studies of the 20th century, in which the same children were measured every few months from infancy to adulthood.

• The Aberdeen Growth Study (1956). Healy, Lockhart, Mackenzie, Tanner and Whitehouse examined how well measurements taken each year from birth to age 5 predict adult body size. This work established that childhood height tracks adult height strongly but imperfectly, which is why our percentile projection uses a correlation of 0.75 rather than assuming a child stays on the same curve forever.
• Bone-age methods. The Bayley-Pinneau method and the Roche-Wainer-Thissen (RWT) method both use skeletal age, read from an X-ray of the hand and wrist, to judge how much growth remains. They're still used by pediatric endocrinologists, but they need an X-ray and a trained reader.
• Parental standards (1970). James Tanner, Harvey Goldstein and Reginald Whitehouse published standards for children's height at ages 2 to 9 that allowed for the heights of the parents, the origin of the "add or subtract 13 cm and average" rule.
• Khamis-Roche (1994). Harry Khamis and Alex Roche modified the RWT model to drop skeletal age entirely, using data from the Fels Longitudinal Study in southwest Ohio: 223 white boys and 210 white girls, measured every six months. Their paper in Pediatrics reported that errors were only slightly larger than RWT's, which made accurate prediction possible at home with a tape measure and a scale.

Khamis-Roche, step by step

Khamis and Roche fitted a separate linear equation for each sex at every half-year of age from 4 to 17.5:

Adult height (in) = β₀ + β₁ × child's height (in) + β₂ × child's weight (lb) + β₃ × mid-parent height (in)

The calculator converts your centimeters and kilograms to inches and pounds, linearly interpolates the coefficients between half-year ages, and converts the answer back to centimeters. It uses the coefficient table as corrected by the authors' 1995 erratum. Some copies of the table online carry the uncorrected weight column, which is roughly ten times too small; we cross-checked our table against two independent implementations.

The coefficients themselves tell an interesting story about how growth works:

Three patterns stand out. First, the weight coefficient is negative: at the same height, a heavier child is predicted to end up very slightly shorter, because extra weight tends to accompany earlier maturation and so less growth remains. Second, parents matter most around early puberty (boys about 12 to 14, girls about 10 to 12), when the child's own height is least informative because growth spurts start at different ages. Third, by the late teens the parents barely matter: for a 17.5-year-old boy the mid-parent coefficient is essentially zero, and for a 16-year-old girl it's 0.03, because the teenager's current height already contains nearly all the information.

Error bounds. The calculator shows 90 percent bounds of about ±5.3 cm for boys and ±4.3 cm for girls, based on the validation errors in the original paper, which was developed on 223 boys and 210 girls from the Fels Longitudinal Study. The paper's abstract describes those errors as only slightly larger than those of the Roche-Wainer-Thissen method, which requires a skeletal-age X-ray, and limits the method's applicability to white American children without growth-altering conditions.

Mid-parental height, and why extremes regress to the mean

Tanner's method averages the two parents after adjusting for the typical difference between men and women:

• Boys: (father + mother + 13 cm) ÷ 2
• Girls: (father + mother − 13 cm) ÷ 2

The result is a genetic "target," and the calculator shows a range of ±8.5 cm around it, roughly two standard deviations.

The method works well for families of average height, but it has a well-documented weakness at the extremes. Charlotte Wright and Tim Cheetham, writing in Archives of Disease in Childhood in 1999, compared 419 representatively sampled children aged 8 to 9 with their parents' heights. About 90 percent of the children were within 1.5 standard deviation scores of their mid-parental height, as expected. But when the parents were unusually tall or short, the children were less extreme than their parents, the statistical effect called regression to the mean. Among 13 children below the 2nd centile, eight were within the 90 percent range of their mid-parental height on the simple rule, yet only three remained so once regression to the mean was accounted for. The authors concluded that mid-parental height is a useful guide for average families but can be misleading when assessing short children.

Percentile projection on the CDC curves

When parents' heights aren't known, as in many adoptive and donor-conceived families, the calculator projects along the CDC 2000 stature-for-age reference. It converts the child's height to a z-score using the LMS method, then reads off the height at the same z-score at age 20. The CDC charts were built from US national examination surveys, as described in the 2002 methods report by Kuczmarski and colleagues.

Because childhood and adult height are correlated at about 0.7 to 0.8 rather than 1.0, children who are far from average tend to drift partway back toward it. The calculator therefore keeps the child's own percentile for the point estimate (common clinical practice) but centers its approximate 80 percent range on the regression estimate, 0.75 times the current z-score, with a residual spread of √(1 − 0.75²) ≈ 0.66 standard deviations.

A worked example: three methods, one child

Take a 10-year-old boy who is 140 cm tall and weighs 32 kg, with a 180 cm father and a 165 cm mother.

At 140 cm he sits around the 58th percentile for his age; the CDC median for a 10-year-old boy is 138.6 cm. All three methods agree within 2 cm, which is typical for an average child in an average family. The ranges are the more important part of the answer: even the most precise method leaves a 10 cm window for 90 percent of children.

Now take an 8-year-old girl who is 128 cm and 26 kg, with a 178 cm father and a 164 cm mother. Khamis-Roche predicts 166.1 cm (5′5″) with 90 percent bounds of about ±4.3 cm, and mid-parental height gives 164.5 cm. The methods line up again because nothing about this child is unusual.

Why predictions move during puberty

Predictions made between about 9 and 15 years old are the least stable, because children enter their growth spurt at very different ages. The CDC reference curves show how differently boys and girls finish growing. At the median, a 14-year-old girl is already 160.4 cm, only 2.7 cm short of the 163.1 cm median at 18. A 14-year-old boy at the median is 163.8 cm and still has about 12.4 cm to go before reaching 176.2 cm at 18.

Boys and girls are almost the same height at 10. The adult gap of about 13 cm opens mostly during the teenage years, because boys' growth spurts start later and last longer. An early-maturing 12-year-old can look tall for their age and then stop growing sooner. A late maturer can look short and then catch up. A percentile projection made at 12 can't tell these two children apart, which is why its range is wide and why Khamis-Roche leans more on the parents' heights at that age.

How accurate are the methods for unusual children?

Agreement breaks down for children who are short, tall, or maturing early or late, which is precisely when families most want an answer. A 2010 study in Pediatrics by Laura Topor and colleagues applied the Bayley-Pinneau, Roche-Wainer-Thissen (RWT) and Khamis-Roche formulas to simulated samples of children with the ages, heights, weights, bone ages and parental heights typically seen in idiopathic short stature. They counted how many children each formula predicted would finish below 160 cm for boys or 150 cm for girls, roughly the 1.2nd adult percentile.

Overall agreement between the methods was poor for boys (kappa 0.21) and negative for girls, meaning the formulas agreed on which girls would be very short no better than chance. The authors concluded that parents, pediatricians and researchers need to recognize the considerable variation in height predictions. The lesson for parents: if your child is well outside the average, treat any single prediction, including ours, as a rough guide and discuss growth with a doctor.

Getting the best inputs

A prediction can only be as good as the measurements going into it.

• Measure the child properly. Use a flat wall and a hard floor, shoes off, heels together, and the back of the head, shoulder blades, buttocks and heels touching the wall. The US NHANES protocol places the head so the line from the ear canal to the lower edge of the eye socket is level, asks for a deep breath, and lowers a headpiece firmly enough to compress the hair. A hardback book held level works at home.
• Measure in the morning. Height falls by a centimeter or more over a day.
• Measure the parents too. Many adults round their height up, while the prediction equations were built on measured heights. A father who reports 180 cm but measures 177 cm moves the mid-parental target by 1.5 cm.
• Use biological parents. All parent-based methods assume genetic parents. For adopted or donor-conceived children, use the percentile projection.
• Weigh accurately. Weight enters Khamis-Roche with a small coefficient, so a kilogram or two makes little difference, but a wildly wrong weight does.

What else shapes adult height

Genes set the broad range, but environment determines how much of that potential is reached. Our World in Data summarizes the evidence: within a well-off population, most variation between individuals is genetic, while differences between populations and across generations largely reflect childhood nutrition and disease. That is why many countries have grown taller over the last century, as shown on the average height by country pages, and why chronic illness, poor nutrition or untreated hormone problems can keep a child below their predicted range.

When to talk to a doctor

No calculator can replace a growth assessment. In the US, CDC guidance flags measurements beyond about two standard deviations from the median (the 2.3rd and 97.7th percentiles) as potentially worth investigating. Talk to your child's doctor if:

• height is far below or above what the parents' heights suggest;
• your child crosses two or more major percentile lines between check-ups (plot several measurements on a growth chart to see the trend);
• growth seems to stall, or puberty starts much earlier or later than in peers.

To see where your child stands today, try the height percentile calculator. To picture a predicted adult height next to the family, add everyone to the height comparison chart.