Why are templates used?
Pretty much exactly this sort of situation: changes that can be made to a variety of creatures, rather than specifically writing up a new creature/race for each combination.
What steps would I use to make a balanced template?
This is very tricky. It would depend a lot on what you are trying to do.
In my opinion, the first thing to do is to be careful about precedent: templates are balanced by their Level Adjustment (LA), but most templates and creatures have much more LA than they should. It’s very difficult to play with more than LA +1 or LA +2, and furthermore even if someone wants to do it, it can make life very difficult for the DM.
LA leads to skewed characters, where some of their stats and abilities are too good for their level, but other stats and abilities are too weak for their level, leading to weird situations where one kind of challenge, normally something of a challenge, can be solved automatically, but another, also usually only a moderate challenge, is suicidal. That makes it pretty difficult to design challenges.
So I would aim for LA +1 templates, most likely. LA +2 only if you’re very careful. Actually, I would consider LA +0 (giving drawbacks and benefits in roughly equal measure) before LA +2.
LA +0
This template effectively changes one LA +0 race into another LA +0 race. The problem is that, for example, LA +0 races rarely if ever give +4 to an ability score, or apply a −4 penalty – and doing so can be problematic when the +4 is to something that a character relies on exclusively, or the −4 is to something the character doesn’t care about. So I’d probably recommend avoiding ability score adjustments altogether, or doing something like
+2 racial bonus to X and −2 racial penalty to Y; does not stack with racial bonuses or racial penalties from one’s actual race. If one already has a racial bonus to X, one does not receive the racial penalty to Y, and if one already has the racial penalty to Y, one does not receive the racial bonus to X.
Which is verbose and annoying, but eliminates the +4 or −4 scenarios.
Other changes should probably lean slightly towards more drawback than benefit, since you must assume that players will try to apply benefits where most useful and drawbacks where least painful. It should only be slight, however. That’s a tricky line to draw.
Skill bonuses and penalties, in the −2 to +2 range, are very safe; you can probably just hand out bonuses and penalties equally and it will be fine. Bonuses up to +4 are possible, but I’d be leery about them.
Bonus martial weapon proficiencies don’t usually matter too much, though it’s hard to imagine an appropriate drawback. They’re probably safe enough to give in very small number.
Bonuses and penalties to saving throws are pretty good, if given out equally. Characters don’t get to choose when they’re making a Fort vs. a Will save.
Bonus feats (aside from Martial Weapon Proficiency) are probably too good for an LA +0 template.
LA +1
LA +1 means that you count as a level higher than you actually are for XP, leveling up, and in theory, what sorts of foes you can face. It’s very, very difficult to write something that is worth exactly one level regardless of how many levels you actually have (i.e. LA +1/Some Class 1 all the way to LA +1/Some Class 19, for all classes); most LA +1 creatures and templates are worth not even close to a level basically all the time.
The best things that LA +1 templates can give out are things that will stack with future classes. Maybe you don’t get all the features of the Rogue class, but you do get 1d6 Sneak Attack damage. Or you get Unarmed Strike damage like a Monk, that stacks with Monk levels, but not the other features. Most classes have these kinds of “partial features” that you can give out.
Feats, particularly good feats, often work out similarly. Bonus feats are quite nice. Maybe you don’t get a free pick, but you get two pre-selected feats: that’s better (but less flexible) than what the Fighter gets.
Most published creatures and templates do not do this, however. The most common thing you see are ability score changes. These are what lead to skewed characters. I recommend avoiding these, though a few small ones (a +4, a few more +2s, only a single −2 or maybe none at all) are appropriate since the character misses out on base HP, Constitution-based HP, base saving throw bonuses, base attack bonuses, and skill ranks. Bonuses to Constitution should be common, in fact, because HP increases with level so much.
Larger skill bonuses are also appropriate, but remember that an LA +1 character basically has a −1 penalty to all his favorite skills because his skill rank maximum is 1 less than it would otherwise be (and he has one level’s worth fewer skill points). I think it would actually be a really good idea to give “free ranks” in several skills (say, 4-6, or even 8-10 if you envision the race to be for Rogues and other skill-heavy classes) to mitigate that.
LA +2
These are very hard to balance. You have to give really good stuff, without overpowering the character. Being behind two spellcasting levels, in particular, is extremely painful: if you want spellcasters to consider this, you have to give them great stuff (including, perhaps, partially progressing their spellcasting, à la some prestige classes).
Really, I recommend against these.
Most of what I said for LA +1 just goes double here. You also want unique features, stuff they couldn’t just get by not taking the template and going into classes regularly. Spell Resistance is pretty common, but SR is a double-edged sword and not really all that good. Damage Reduction is definitely better, though DR/magic will soon be close to worthless. This is going to involve some creativity on your part, I imagine. Without more details on the races in question, it’s hard for me to suggest anything.
What’s the real difference between a race and a template as it relates to what I'm trying to create?
Templates go on top of races, so you get both. With a race, you have to choose between the two. That’s the only real difference between them.
Where can I find more information on these topics?
I will have to get back to you on that; I’m not sure.
Best Answer
This is going to be difficult to balance because bounded accuracy exists in 5e, and does not in 3.5.
The reasons you're going to run into serious problems with this is that 5th edition uses bounded accuracy. What this means is that there's a lower and upper limit on things like stats, saving throws, and skill checks.
For example: If you're using customizable stats, you can not have a score lower than 8 or higher than 15 before racial bonuses.
And except in extremely specific scenarios (like Barbarian or magic items), your maximum ability score is 20, even with racial and class bonuses.
If you're using random rolls to assign stats to your character's race, you're going to need to take into consideration the specific traits associated with each race and apply hard caps accordingly. Example, an Ogre has a trait called Legendary Stupidity. So even if you rolled 18 in INT, you'd still have trouble counting to 10 with your fingers in front of you, are prone to believing anything you're told, and are easily confused. This is traded off by their incredible Strength. So if I was to create a PC Ogre race, they'd get +4 to Strength as well as an increase maximum Strength comparable to a Barbarian (max of 24, and yes a Barbarian Ogre would then have a cap of 28 as it levelled), and a hard cap of 8 Intelligence to account for these. But this is what I mean when I say you have a lot of work ahead of you in trying to port these over effectively.
If you're going to try to make these fit, my recommendation is to use the stats from the Monster Manual as a direct translation of how strong that race should be when starting. Now, this is going to result in some disappointing races, as well as some incredibly overpowered ones.
The bottom line is that there is no easy way to translate 3.5 to 5. They don't run off the same foundation, so it's like trying to balance Warhammer with Scrabble.