In the overwhelming majority of criminal trials for felonies, and a very large share of misdemeanor and traffic trials, criminal defendants, following the advise of their criminal defense attorneys – who have a good knowledge of the system – overwhelmingly choose jury trials. (Even though criminal defense attorneys may personally have reasons to prefer bench trials which take less preparation as criminal defense attorneys are usually spread very thin with many clients and limited resources.)
This is primarily for one very good reason. A jury is six times more likely to acquit than a judge and, unlike a judge, can also deadlock (resulting in a mistrial followed by a retrial or more lenient plea bargain in most cases), if the facts are at all susceptible to more than one interpretation. Judges and juries agree on the verdict in roughly 78% of criminal jury trials where a verdict is entered, but judges are much less likely to acquit defendants than juries do, (3% v. 19%+) (see the sources cited in this previous Law.SE answer).
The National Center for State Courts conducted a survey of hung jury rates using felony case data from all federal courts and 30 state courts in 75 of the most populous counties. The NCSC project found that state courts in large urban areas had an average hung jury rate of 6.2%, with substantial variation across courts, ranging from a low of 0.1% in Pierce County, Washington to a high of 14.8% in Los Angeles County, California. Federal hung jury rates were found to be particularly low, averaging about 2% of all federal jury trials: federal civil trials had lower rates than federal criminal trials. (Source.)
On average a criminal defendant is eight times more likely to avoid a criminal conviction in a state court criminal case in a trial before a jury than in a bench trial.
In federal court, there is a 38% chance of bench trial acquittal (12% of all federal criminal trials are bench trials while 88% are jury trials), some of which are for misdemeanors and traffic cases, and a 14% chance of jury trial acquittal. But only 2% of federal criminal cases go to trial (roughly 99.6% of cases presented to federal grand juries result in indictments) with the remainder plea bargained (90% of cases charged) or dismissed (8% of cases charged, sometimes due to a state court conviction), and only about 2% of all U.S. criminal cases are filed in federal court. So, federal criminal bench trials are very rare and atypical.
Secondarily, there is lots of highly prejudicial and legally irrelevant information (e.g., the criminal history of the defendant, or evidence excluded for 4th Amendment violations) that a jury will not know, but a judge will know. Jurors almost never disregard evidence that they are told to disregard, and often give it undue weight, and judges aren't any better.
The real question should be: Who chooses a bench trial and why, notwithstanding these facts? Some situations in which someone might consider a bench trial rather than a jury trial include:
A bench trial can be easier for a pro se defendant, which is one not represented by counsel either because the defendant is not indigent and hence not entitled to counsel, or because the proceeding is one in which there is no right to counsel (usually traffic cases). In these, usually low stakes, matters the cost of counsel does not justify the expenditure and it can be easier to do a bench trial as a pro se party than a jury trial. The vast majority of criminal or quasi-criminal bench trials are in low stakes cases (which are segregated into separate courts in state court and mixed in with more serious cases in federal court) where this is a primary consideration.
A bench trial can be desirable for someone who needs a quicker trial date (e.g., someone who lives somewhere seasonally and needs to get home), especially in a low stakes case where the sentence is likely to be a fine or a sentence to "time served" if there is a sentence.
A bench trial can be desirable when the guilt-innocence determination is largely a foregone conclusion but the defendant would still like to at least try to be acquitted, as a way to more fully tell a story of how the incident was understandable even if not a legal justification for conduct, to prepare the judge for sentencing with a more full picture of what happened than a guilty plea would provide. For example, it might be a strategy that someone who assaults (ambush style) someone who raped his sister might take, even though a tape recorded confession and video tape of the incident makes the likelihood of acquittal small (unless the defendant thinks that "jury nullification", i.e. ruling in his favor notwithstanding the law, is likely). Another example would be to provide a forum in a drug possession case to show that someone in possession of a large quantity of controlled substances was really a "mule" and not a big time dealer. This kind of consideration drives a fair number of bench trials in federal criminal cases.
A bench trial can be desirable for a defendant who thinks that a judge will be favorably inclined towards him (e.g., some law enforcement officers or public officials).
A bench trial can be desirable for a defendant when a jury is likely to be biased against him (e.g., a Muslim defendant at a time when anti-Muslim sentiment is strong due to some recent news event, or a defendant in a community with lots of racist potential jurors who is before a black judge).
A bench trial can be desirable for a defendant whose primary defense requires above average sophistication to understand (e.g., a defendant in some complicated regulatory or white collar case).
Bench trials are more attractive in cases where the criminal defendant is a first time offender, has no arrest record, and is not relying on suppressing any evidence at trial, so that the availability of more information to the judge than the jury is not a problem.
A summary of jury trials in Colorado, which is typical, shows the revealed preference for jury trials. In 2016, defendants elected jury trials in 67% of traffic cases, 69% of misdemeanor cases, 97% of state court felony cases, and 59% of federal criminal cases (which are a mix of misdemeanors and felonies).