Calorie deficit

What Is Energy Balance? Why Calories In vs Calories Out Feels Uneven

Energy balance is the energy you eat minus the energy you use, but both sides are estimates: calculators miss by 246 to 339 kcal a day, and one morning's scale reading can move about 0.5 kg.

Key takeaways

  • NASEM's 2023 equations have a root mean squared error of 246 kcal a day for women and 339 kcal for men, so the energy-out side of the equation is a range.
  • In the OPEN study, 24-hour recalls came in 12 to 14% below measured expenditure in men and 16 to 20% below in women, so a diary kept from memory may read low.
  • One healthy man's 9,521 days of weigh-ins show a spread of 0.69% across readings 7 days apart, which is 0.48 kg at 70 kg.
  • Averaging a week of weigh-ins cuts the noise to about 0.13 kg, and week 1 against week 2 gives a maintenance estimate good to roughly ±200 kcal a day from the scale alone.
  • Hall's 2011 model puts half of a lasting change in intake at about 1 year, so the straight 7,700 kcal per kg rule is a rough conversion even over 14 days.

Energy balance is the gap between the energy you eat and the energy you use, and the body stores that gap as tissue or draws it down. Neither side can be measured to better than a few hundred kcal a day at home, and the scale adds noise of its own. Averaging 14 days of weigh-ins and food logs gives a usable estimate of your own balance. I’ll show the error you carry with it.

What is energy balance, and how does it work?

The definition is intake minus expenditure. When you eat more than you use, the surplus is stored. When you eat less, the body draws on its stores, and that gap is a calorie deficit. When the two match, weight holds roughly steady.

Kevin Hall’s 2011 Lancet paper quantifies what an energy imbalance does to bodyweight, and its answer is that no fixed conversion holds. The usual bridge from energy to kilograms is the rule of 3,500 kcal per pound, or 32.2 MJ per kg. Hall prints the MJ figure. The 7,700 kcal per kg used below is my division of 32.2 MJ by 4.184, which gives 7,696, rounded. How many calories is a pound of body fat takes the rule apart, so I won’t repeat that here.

The imbalance that matters is small. Hall’s model puts the average US adult’s persistent gap between intake and expenditure at about 30 kJ a day, roughly 7 kcal, behind the observed average weight gain. That is far below the error of any home measurement in this article. Energy balance can be true and still be invisible on any one day.

I find it easier to treat the equation as three measurements instead of one: what went in, what went out, and what the scale says changed. Each has an error, and the errors don’t cancel.

Why does calories in vs calories out not feel equal?

Because every term in it is an estimate, and the third term, the scale, also moves for reasons that have nothing to do with energy. Start with the food side.

A packaged food can legally contain more than its label says. Under 21 CFR 101.9, a food is misbranded if its analyzed calories are more than 20 percent above the declared value. The rule runs one way. A bar labeled 200 kcal can come in at 240 without the rule applying. The regulation adds that no action is taken on a value above that level by less than the variability recognized for the lab method.

Your diary adds a second error. The OPEN study compared 24-hour dietary recalls with doubly labeled water in 484 adults aged 40 to 69. Men’s recalls ran 12 to 14% low, and women’s 16 to 20%. Recalls from memory aren’t weighed logs, and I found no matching figure for weighed logging, so read these as the high end of what a diary can miss. A log reading 2,000 kcal might mean 2,273 (2,000 ÷ 0.88) or 2,500 kcal (2,000 ÷ 0.80).

The expenditure side is no cleaner. The 2023 NASEM equations, fitted on 1,342 women and 1,016 men, have a root mean squared error of 246 kcal a day for women and 339 kcal for men. Both were measured on the data the equations were fitted to. Inside that number sit parts you don’t control. Digesting food costs about 10% of ingested energy on a mixed diet, or 200 kcal on a 2,000 kcal day. The same review lists 0 to 3% for fat, 5 to 10% for carbohydrate and 20 to 30% for protein, so your own share depends on what you eat. Everyday movement varies more. In a study of 20 sedentary volunteers, 10 lean and 10 mildly heavier, the heavier group sat about 2 hours longer a day. The authors estimate that if they adopted the lean group’s habits they might expend an additional 350 kcal a day. Hall’s paper adds that resting metabolic rate and the energy cost of activity shift as weight changes. The fixed rule ignores both. For the pieces of expenditure, see what TDEE is, and for how far calculators can drift, how accurate TDEE calculators are.

Term What you read Published error Direction
Food label kcal on the pack analyzed calories can exceed the label by up to 20% before the rule applies high only
Diary from 24-hour recall kcal a day logged 12–14% low in men, 16–20% low in women low
Calculator kcal a day used ±246 (women), ±339 (men) either
One weigh-in kg on the scale spread of about 0.69% across 7 days, 0.48 kg at 70 kg either

None of the rows is a flaw in you. I put the errors side by side for a 70 kg example. The surprise was that two weeks of scale noise and the calculator’s error come out about the same size in kcal. Before you decide a day went wrong, ask which of the three terms moved.

Why is one day on the scale mostly noise?

Because a morning reading counts water, stored carbohydrate and the contents of your gut, and all three change faster than fat does. Glycogen is the stored carbohydrate in muscle and liver. Each gram is stored with at least 3 g of water. In a crossover study of 22 men, a moderate-carbohydrate diet after depleting exercise left muscle glycogen at 367 mmol per kg of dry weight. A high-carbohydrate diet left 525, and body mass was 0.7 kg lower on the moderate one.

The gut is the third pool. In 220 healthy UK adults who made careful collections, the median stool weighed 106 g a day and whole-gut transit took 60 hours. So part of today’s reading is food from two days ago. A drink goes on the scale the moment you swallow it.

How big is the swing? One long record is a healthy man weighed on 9,521 days. The standard deviation of the relative difference was 0.53% for readings one day apart and 0.69% for readings seven days apart. At 70 kg that is 0.37 kg and 0.48 kg. One man shows the size of the noise, and I use it for nothing more than that. How often to weigh yourself sets this swing against real weekly loss.

Set that against what a day of energy balance can do. A 500 kcal gap is 500 ÷ 7,700 = 0.065 kg of tissue by the rule. The scale’s daily noise is roughly six times larger. No single weigh-in can show you the gap, and the arithmetic of averaging is what rescues it. If each reading carries independent noise of 0.69% ÷ √2, about 0.34 kg at 70 kg, then the mean of seven readings carries 0.34 ÷ √7, about 0.13 kg. Real drift inside a week breaks the independence a little, so treat 0.13 as optimistic.

What does the dynamic model change about the rule?

It makes the answer depend on time. The Lancet model lets expenditure fall as weight falls, and it finds that the response of body weight to a change in intake is slow, with half-times of about 1 year. The authors boil it down to a rule of thumb for an average adult: every permanent change of 100 kJ a day leads to an eventual change of about 1 kg, and about half of it takes 1 year. Their figures are averages for a population. Your own curve will differ.

Run a 300 kcal gap through both. Three hundred kcal is 1,255 kJ, so the eventual change is near 12.6 kg and about half of that, 6.3 kg, shows up in a year. The straight line says 300 × 365 ÷ 7,700 = 14.2 kg in the same year.

Two weeks is a harder case for the straight rule than I expected. If I treat the response as one exponential with a one-year half-time, day 14 has produced about 0.33 kg. The straight line gives 300 × 14 ÷ 7,700 = 0.55 kg. That is my arithmetic on the paper’s half-time, not a figure the authors publish, and the paper’s model is more detailed than one curve. Still, the 7,700 rule looks about 0.2 kg generous even at two weeks. That gap is larger than the 0.13 kg noise in a weekly mean, so the conversion is a third source of error in any estimate below.

The same paper makes a related point about exercise. The energy cost of added activity is proportional to bodyweight itself, so an activity change and a food change that start at the same kcal can lead to different weight changes.

NIDDK’s Body Weight Planner is built on this work: the research page behind it credits Hall’s group and the 2011 paper. It asks for weight, sex, age, height and activity level, and its own page says people often underestimate how much they eat.

How do I estimate my own energy balance in 14 days?

Log everything you eat for 14 days, weigh yourself each morning, and apply the same 32.2 MJ per kg rule in this form:

maintenance = average intake − (Δweight × 7,700 ÷ days)

Δweight is the mean weight of week 2 minus the mean of week 1, in kg. Days is 7, the distance between the middle of the two weeks. If weight fell, Δweight is negative and the formula adds the gap to your intake.

Take a 70 kg adult. The log averages 2,000 kcal a day. Week 1 averages 70.4 kg and week 2 averages 70.1 kg, so Δweight is −0.3 kg. That is 0.3 × 7,700 = 2,310 kcal across 7 days, or 330 kcal a day. Maintenance comes out at 2,000 + 330 = 2,330 kcal.

The honest part is how far to trust 2,330. Using the 0.34 kg per-reading noise from above, the difference between two weekly means has a standard error of 0.34 × √(2 ÷ 7) = 0.18 kg. Over a 7-day gap that is 0.18 × 7,700 ÷ 7 = 200 kcal a day. Longer logs help.

Log design Days between the two averages Scale noise in your estimate (kcal a day)
Day 1 vs day 14, single weigh-ins 13 ±290
Week 1 vs week 2 averages, the 14-day log 7 ±200
Week 1 vs week 3 means 14 ±100
Week 1 vs week 4 means, from a log of four full weeks 21 ±67

These are one standard deviation, built on the single-man record and my independence assumption. The diary bias comes on top and doesn’t shrink with time. If your log runs 12% low, as the recalls did, true intake is 2,273 kcal and maintenance moves to 2,273 + 330 = 2,603 kcal. So 2,330 is the middle of a plausible band that runs from about 2,100 to 2,800.

That band is as wide as the calculator’s. So don’t swap one for the other. Start with the calorie calculator, which takes a minute and gives you a number to test. I treat a landing within about 250 kcal of it as agreement. If the two disagree by more, keep logging to week 4, where the scale term drops to ±67. How many calories to maintain weight walks through the same method over a longer window, with the NASEM equation as the starting point.

What doesn’t work: comparing two single weigh-ins, or filling in the log from memory at night. Changing how you eat halfway through fails too, because the first days of a new plan move glycogen and water, and that shows up as fake loss. Weigh the foods you eat most, and write down the oil, the sauces and the drinks, because those are the usual gaps. Hold the way you eat steady for the whole 14 days. A log that tracks your normal week tells you about your normal week.

When should you not use this?

Skip it if you’re pregnant, breastfeeding, under 18 or managing a medical condition, because a doctor or midwife sets those needs. A diary estimate won’t replace that advice. If your weight has changed and you can’t say why, see your doctor first.

If counting calories has started to feel hard to stop, or eating feels out of control, talk to someone: ANAD in the US (1-888-375-7767), Beat in England (0808 801 0677), NEDIC in Canada (1-866-633-4220) or the Butterfly Foundation in Australia (1800 33 4673). If you’re in danger right now, call your local emergency number. Other UK numbers and crisis lines are on our crisis support page.

Calories in vs calories out is true as accounting and weak as a daily check. Over a day, the scale tells you about water. Over two weeks it starts to tell you about energy, with an error you can now size. Tomorrow, write down your weight before breakfast and every meal you eat. On day 15, do the sum and write the range next to the number.

Questions and answers

Is calories in vs calories out true?

Yes as accounting, with large errors in practice. Stored energy changes by intake minus expenditure, but labels can run up to 20% above the declared calories before the rule applies, 24-hour recalls read 12 to 20% low in the OPEN study, and calculators miss by 246 to 339 kcal a day.

Why does my weight jump 1 kg overnight?

Mostly water, stored carbohydrate and gut contents rather than fat. Each gram of glycogen is stored with at least 3 g of water, and in one trial a moderate-carbohydrate diet left body mass 0.7 kg lower. A 500 kcal gap is only about 0.065 kg of tissue by the 7,700 rule. In the one long weight record I found, the typical day-to-day spread was 0.37 kg at 70 kg, so 1 kg is a large jump.

How long should I log to estimate my energy balance?

Fourteen days gives a usable first estimate, and 28 days is better. Scale noise alone is about ±200 kcal a day when you compare week 1 with week 2, and about ±67 kcal when you compare week 1 with week 4, using the single-man weight record I found.

Can a 14-day estimate replace a calorie calculator?

No, it works as a check on one. Both carry errors of a few hundred kcal a day. I treat a 14-day number within about 250 kcal of the calculator's as agreement, and if it is further off, log to week 4.

Sources

  1. Hall KD, et al. Quantification of the effect of energy imbalance on bodyweight. Lancet 2011, The Lancet
  2. Hall KD. What is the required energy deficit per unit weight loss? Int J Obes 2008, International Journal of Obesity
  3. 21 CFR 101.9, Nutrition labeling of food, paragraph (g)(5), Electronic Code of Federal Regulations
  4. Subar AF, et al. Using intake biomarkers to evaluate the extent of dietary misreporting in a large sample of adults: the OPEN study. Am J Epidemiol 2003, American Journal of Epidemiology
  5. Dietary Reference Intakes for Energy (2023), Chapter 5, Table 5-6: Evaluation of model fit for the general TEE prediction equations, National Academies of Sciences, Engineering, and Medicine
  6. Westerterp KR. Diet induced thermogenesis. Nutr Metab (Lond) 2004, Nutrition & Metabolism
  7. Levine JA, et al. Interindividual variation in posture allocation: possible role in human obesity. Science 2005, Science
  8. Murray B, Rosenbloom C. Fundamentals of glycogen metabolism for coaches and athletes. Nutr Rev 2018, Nutrition Reviews
  9. Schytz CT, et al. Lowered muscle glycogen reduces body mass with no effect on short-term exercise performance in men. Scand J Med Sci Sports 2023, Scandinavian Journal of Medicine & Science in Sports
  10. Cummings JH, et al. Fecal weight, colon cancer risk, and dietary intake of nonstarch polysaccharides (dietary fiber). Gastroenterology 1992, Gastroenterology
  11. Schneditz D, Hofmann P, Krenn S. Day-to-day variability in euvolemic body mass. Ren Fail 2023, Renal Failure
  12. Hall KD, et al. Quantification of the effect of energy imbalance on bodyweight (PubMed abstract: half-times of about 1 year, 30 kJ a day gap), PubMed
  13. NIDDK Body Weight Planner, National Institute of Diabetes and Digestive and Kidney Diseases
  14. Research Behind the Body Weight Planner, Integrative Physiology Section, National Institute of Diabetes and Digestive and Kidney Diseases

About the author

Claire Donovan

Editor, Calsa blog

Claire edits the Calsa blog. She writes about calorie counting, protein and the small questions that come up in your first weeks of tracking.

All articles by Claire

Put the numbers to work.

The guides show where each number comes from. In Calsa you log a meal by photo, voice or barcode, check the grams, and every calorie shows its source.

Log my first meal
A snack board of hummus, carrot and cucumber sticks, crackers, cheese, grapes and walnutsCalsa entry details: 71 kcal per 100 g times 250 g, with USDA as the source