Baking Powder Is a Two-Stage Gas Engine

Baking powder is a pre-balanced leavening system: roughly 1 part baking soda, 1-2 parts dry acid (cream of tartar or sodium acid pyrophosphate), and cornstarch as a buffer. One teaspoon releases about 130 ml of carbon dioxide in two waves — once when wet, once when hot. The best swap is 1/4 tsp baking soda + 1/2 tsp cream of tartar per teaspoon. Buttermilk plus baking soda works at a 1/2 cup ratio if you cut other liquid.

What baking powder does

Baking powder exists because of a measurement problem. A cake batter needs to expand by roughly 60-80% in volume during baking, and that expansion has to happen on a clock. Too early, and the gas escapes before the starches gel. Too late, and the proteins set into a dense slab before the bubbles can lift them. The whole job of leavening is to put gas into the batter at the right moment, in the right quantity, and to do it without leaving an off-flavor behind.

Plain baking soda — sodium bicarbonate — does the gas-generation part. When it meets an acid in solution it splits into water, sodium salt, and carbon dioxide. But raw baking soda needs an acid in the recipe to react. If the recipe doesn't have one, the unreacted soda stays behind and tastes soapy and metallic, and turns the crumb a yellowish-brown. That's the problem baking powder solves.

A standard double-acting baking powder is a kit. The bicarbonate is there. The acid is also there, in dry form — usually monocalcium phosphate (which dissolves at room temperature, the moment it gets wet) plus sodium aluminum sulfate or sodium acid pyrophosphate (which only reacts above about 60°C / 140°F). The cornstarch in the can is a desiccant that keeps the two reactive ingredients from finding each other in the humidity of your pantry. This is an underappreciated design choice: the powder can sit on a shelf for eighteen months and still work because the acid and base are separated by a starch barrier.

This is where the measurement-and-data framing matters. One teaspoon of double-acting baking powder weighs about 4 grams and contains roughly 1.3 grams of bicarbonate, enough to release about 130 ml of CO₂ at full reaction. Roughly 30% of that gas comes off in the first reaction, when the batter is still cold and pourable. The other 70% comes off between 60°C and 90°C, when the batter has already started to set around the bubbles. That two-stage release is the entire point: the first wave makes the bubbles, the second wave inflates them at the moment the structure can hold them open.

Baking powder is also the leavening choice when your recipe has no acidic liquid. Pancakes made with whole milk, biscuits made with whole milk instead of buttermilk, vanilla cake with no sour cream — these all rely on baking powder because there's no recipe-side acid for plain soda to react with. (For the case where there is recipe-side acid, see the framing in the buttermilk biography.)

What breaks when you swap it

The first thing that breaks is texture. This is not a flavor problem and it's not a rise problem in the simple sense — it's a bubble structure problem. When you replace baking powder with something that releases all its gas at once, you get a batter that bubbles up enthusiastically on the counter, then collapses by the time it reaches the oven. The cake goes in flat and comes out denser than it should be, with a tight, gummy crumb instead of an open, tender one. Pancakes look like they're rising in the pan and then stop, leaving you with disks that are spongy at the edge and wet in the middle. This texture failure is the most common consequence of a poorly chosen substitution, and it's irreversible once baked. No amount of extra baking time will recover a crumb that never properly inflated.

The data block is blunt about which substitutes hit this wall. "Cornstarch has zero leavening power" — that's a structural warning attached to cornstarch as a candidate. People reach for it because they associate it with baked-goods chemistry, but cornstarch is a thickener; it has no bicarbonate and no acid, so it generates no gas. A cake made with cornstarch in place of baking powder doesn't rise at all.

The second structural warning: "Yogurt is not a direct leavener." Yogurt and Greek yogurt are acidic — pH around 4.5 — so they can react with baking soda. But on their own they cannot replace baking powder, because the soda still has to come from somewhere. People sometimes substitute Greek yogurt for baking powder in a 1:1 panic-swap and end up with a batter that's just thicker, not lifted.

The third: "Eggs add lift only when whipped thoroughly." Whipped eggs are the classic mechanical leavener — folded into a batter, the air pockets trapped in the foam expand in the oven and do most of the lifting work in genoise, angel food, and chiffon cakes. But the data says it plainly: unbeaten eggs do nothing. A pancake batter with an extra egg cracked in but not whipped is just a wetter pancake batter. The lift is in the foam, not in the protein.

There's also a ratio trap. The warnings flag "Reduce other liquid when using buttermilk" and "Replace some liquid with 1/2 cup yogurt" — meaning that when you use a wet acid as part of your leavening system, you've now added 1/2 cup of liquid to a recipe that didn't ask for it. If you don't pull an equivalent amount of milk or water back out, the batter is too thin and the cake comes out wet, dense, and pancake-flat. Texture failure again, by a different route.

And one flavor warning worth noting because people swap into it accidentally: molasses "adds dark caramel flavor to batter." Molasses is acidic and can pair with baking soda to leaven, but it brings color and a deep brown-sugar taste with it. Fine for gingerbread; ruinous for a vanilla cake. The flavor change is as disruptive as the textural one if the recipe wasn't designed for it.

The swaps that work and why

The cleanest substitute is 1/4 teaspoon baking soda plus 1/2 teaspoon cream of tartar per teaspoon of baking powder. Function-match 100/100. This is, almost literally, what's in the can — bicarbonate plus a dry acid in stoichiometric balance. The one difference is that you've made a single-acting powder: cream of tartar dissolves at room temperature and starts reacting the instant it hits liquid. Mix it fresh, mix it last, and get the batter into the oven inside about 10 minutes. The full 130 ml of CO₂ will release in one wave instead of two, but if your batter is already at the oven door, you'll catch most of the lift.

The second swap is buttermilk plus baking soda: 1/2 cup buttermilk and 1/4 teaspoon baking soda per 1 teaspoon baking powder, also function-match 100/100. This works because buttermilk's lactic acid (pH 4.4-4.6) reacts with the bicarbonate the same way cream of tartar does. The catch — and the data warns about it — is that 1/2 cup of buttermilk is real liquid. Pull 1/2 cup of milk or water out of the recipe. If you don't, the texture failure isn't even subtle. (For more on what buttermilk does as an acid, see the buttermilk piece.)

The third tier is 3 teaspoons of baking soda alone per teaspoon of baking powder, function-match 50/100. This is technically possible if the recipe already contains a strong acid — molasses, brown sugar with a lot of molasses content, citrus juice, cocoa, or sour cream — and that acid is in proportion to the soda. The function match drops to 50 because you're now relying on the recipe's existing acid balance, which most recipes don't actually have. Use it only when you've verified the math: roughly 1/4 cup of buttermilk-equivalent acid per 1/4 teaspoon of soda.

Cream of tartar alone, at 5/8 teaspoon per teaspoon of baking powder (function-match 50/100), is a marker swap — it tells you you're missing the bicarbonate. The note in the data is correct: 5/8 tsp cream of tartar + 1/4 tsp baking soda per 1 tsp baking powder. Without the soda, cream of tartar does nothing useful as a leavener. It's an acid waiting for a base.

Whipped eggs at roughly 1 egg per teaspoon of baking powder (function-match 16/100) are a real substitute for a specific category of recipe: pancakes, quick breads, and anything where mechanical aeration can carry the structure. The match score is low because eggs do not produce CO₂ — they trap air mechanically and let it expand in the heat. Separate the whites, whip them to medium-stiff peaks, and fold them in last. In a thin batter (crepes, pancakes), this works surprisingly well. In a dense batter (muffins, banana bread), the foam collapses against the weight and you get marginal lift. It's worth noting that a recipe already containing eggs can sometimes absorb a small reduction in baking powder by leaning harder on the eggs' existing foam — you're not adding a new leavener so much as redistributing the work.

Vinegar plus baking soda — 1/2 teaspoon apple cider or distilled vinegar plus 1/4 teaspoon baking soda per teaspoon of baking powder, function-match 0/100 in the database for reasons of completeness rather than effectiveness. The 0 reflects that this is a workaround, not a designed leavening system: vinegar reacts with soda almost instantly, so you get a violent first wave of gas and nothing in the oven. Use it only for flat baked goods (pancakes, cornbread) where you can pour and bake immediately.

Swap-by-use-case quick reference

The applicability scores tell you where baking powder substitutions actually land. Dessert and baking both score 3.13 (out of a possible 4), which is the highest in the table — these are the use cases the substitution data was built for. For both, the cream-of-tartar-plus-soda swap is your top pick (function-match 100), with buttermilk-plus-soda as a strong second when the recipe can absorb the extra 1/2 cup liquid; this is the mainline path for a cake or a quick bread.

Savory scores 2.67 — think biscuits, drop scones, savory muffins. Buttermilk-plus-soda often wins here because savory baked goods usually have room for buttermilk's tangy flavor, and many savory recipes already call for acidic ingredients that support the chemistry. Cooking at 2.53 covers things like crispy battered coatings; baking powder's role there is to puff the batter, and a cream-of-tartar swap holds up. Sauce (2.13), dressing (1.93), and marinade (1.87) are mostly cases where baking powder is borrowed as a tenderizer or pH adjuster, not a leavener — plain baking soda usually replaces it directly in these contexts without any acid needed. Drink (1.47), frying (1.27), and raw (1.07) are edge cases; the general substitute list is the right starting point if your use case isn't baking.

Related substitutions on SwapCook

For the full ranked list of swaps with ratios and function-match scores, see the baking powder substitutes page, or jump straight to substitutes for use in baking for the dessert and quick-bread case where most of the data lives.