Balance is not achieved by putting one object on each side. Photo: Bärwinkel, Klaus, via Wikimedia Commons (CC BY 4.0).1
“Both sides are bad” is often true. It is also often the point at which thinking stops.
Two parties may both lie. Two governments may both misuse power. Two cultures may both constrain women. Two news sources may both make errors. None of those shared predicates establishes equivalence. It establishes only that the comparison has begun.
The mistake is subtle because it borrows the language of fairness. If we name a failure on one side, shouldn’t we name a failure on the other? Yes. But equal willingness to investigate is not an obligation to reach equal verdicts.
This is the difference between impartiality and symmetry. Impartiality applies the same test. Symmetry assumes the test must produce the same result.
The smallest common label
False equivalence works by moving upward to a category broad enough to contain both cases and then refusing to come back down.
One newspaper publishes a correction after getting a figure wrong. Another fabricates a source and quietly deletes the story. “Both spread misinformation” is technically true. It is also analytically thin. It erases intent, frequency, correction and institutional practice—the features we needed the comparison to reveal.
The same move appears in moral and political disputes:
- criticism and censorship both “constrain speech”;
- an inefficient agency and a captured agency both “fail citizens”;
- a beauty norm and a compulsory dress law both “regulate appearance”;
- corruption and the disabling of corruption investigators both “abuse power.”
The shared description is not necessarily false. The equivalence arrives when the speaker treats that description as sufficient.
Peeling the comparison
A better method is to peel away the common label one layer at a time. Five questions usually expose what the slogan conceals.
1. What exactly happened?
Describe the conduct at the lowest useful level. “Both violated norms” is too abstract. Did one official bypass a procurement rule? Did another rewrite the rule so bypassing scrutiny became legal? Did one outlet misstate a fact? Did another knowingly manufacture it?
Specific verbs prevent a vague noun from doing all the moral work.
2. How strong is the evidence?
The fact that each side has an accusation against it does not mean each accusation has equal support. Compare documents, independent witnesses, adjudication, replication and the quality of contrary evidence. Count neither allegations nor fact checks as if each were an identical unit.
Experimental research on journalistic “false balance” shows why presentation matters. Derek Koehler gave people numerical information about expert opinion and, in some conditions, paired it with comments from two disagreeing experts. The paired comments made participants perceive less consensus—even when they had the full count in front of them.2 A one-versus-one format can overpower the distribution of evidence it supposedly represents.
3. What is the scale and pattern?
An episode is not a system; a system is not an isolated episode. Compare frequency, duration, reach and whether the behaviour is exceptional, tolerated or organised.
This does not mean small wrongs should be ignored. It means “wrong” is a category, not a measurement. One stolen vote and an election-wide manipulation can both be electoral wrongdoing without being interchangeable.
4. Who has the power to impose the harm?
Identical words have different consequences when backed by different capacities. A citizen insulting a minister and a state agency intimidating a citizen are both hostile acts. Only one carries the machinery of detention, licensing, taxation and prosecution.
Power is not a permanent moral exemption for the weak. A powerless actor can still lie, abuse or become powerful. But present capacity belongs inside the comparison because it changes exposure, remedy and risk.
5. Can the system correct itself?
This is often the decisive layer. Can journalists investigate? Can courts review? Can opposition parties compete? Can citizens organise? Can officials lose office? Can an error be named without disabling the person who names it?
Political scientist Nancy Bermeo defines democratic backsliding as the state-led weakening or elimination of the institutions that sustain democracy. Her important observation is that contemporary backsliding often occurs gradually through executive aggrandisement and strategic manipulation, not only through spectacular coups.3 The attack is therefore not merely another policy failure. It targets the machinery by which policy failures can be exposed and reversed.
Comparative research on cases where democratic erosion stopped likewise identifies judicial and parliamentary oversight, civil society and media pressure, and electoral competition as different forms of accountability that helped halt it.4 A bad decision inside a functioning correction system and an attack on the correction system are not equivalent failures. The second can protect many future bad decisions from remedy.
The brake-light test
Imagine two cars.
One has a broken brake light. The other has had its brake lines cut. Both are unsafe. Both require action. But the first is a defect in the vehicle; the second disables the vehicle’s capacity to stop.
This is a useful political test. Ask whether the conduct is a failure within the system or a failure that degrades the system’s ability to identify, constrain or replace the people responsible.
A welfare programme can be badly administered while auditors, courts, opposition legislators and reporters remain able to expose it. That is serious. A government can also intimidate auditors, weaken courts, starve opposition scrutiny and capture the information environment. That is serious in a different, compounding way. It changes not only today’s outcome but tomorrow’s capacity for correction.
The test does not automatically vindicate the first government or condemn every institutional reform by the second. Correction mechanisms can themselves be biased, slow or unaccountable. The point is to examine whether they remain independent enough to impose costs on power.
Where the method can be abused
People who dislike a conclusion often call any unequal judgment “bias.” People who like a conclusion can declare their preferred side uniquely dangerous and skip the evidence. The peeling method prevents neither manoeuvre unless it is applied symmetrically.
Use the same axes for both cases:
| Axis | Question |
|---|---|
| Conduct | What specifically was done? |
| Evidence | How well is it established? |
| Scale | Is it isolated, repeated or systemic? |
| Capacity | What power backs it, and who bears the risk? |
| Correction | Can the actor be investigated, constrained and replaced? |
| Reversibility | How difficult is the damage to undo? |
Do not add an “intent” column unless the evidence supports it. Intent can aggravate a wrong, but it is also the easiest feature to invent for allies and enemies. Observable design and repeated response to correction are usually safer guides.
And do not let asymmetry become immunity. If one side poses the larger institutional danger, its opponent’s corruption or incompetence does not become harmless. A ranking of risks is not a pardon.
Balance is a property of the method
The purpose of comparison is not to make every scale level. It is to discover what each side weighs.
Sometimes the result will be close. Sometimes one failure will be larger but more reversible. Sometimes a smaller present harm will carry a greater future risk because it removes the guardrails. Sometimes two apparently different acts really will be morally equivalent.
We cannot know in advance. That is why impartiality belongs in the questions, not in a prearranged answer.
“Both sides” should open an inquiry. When it closes one, it is not balance. It is a way of hiding the weights.
