There are several things to consider.
1) Not all horses were created equal. Are we talking a nag suitable only for pulling carts and destined in not too long for the knacker's block, a draft horse for the plow, a lady's riding palfrey, a good quality rounser, a courser for use in battle, or a knight's well breed drestier for use in tournaments? Prices could be very different depending on what we were talking about. Talking about the price of a horse even today is problematic, and it was even more so in the medieval era. It would be like assuming that a Honda Civic and a Ferrari have the same customer base and prices.
2) It's also worth noting that recorded prices of goods are fairly rare, and so most compiled lists you see have prices covering a span of centuries. Well, inflation and deflation aren't strictly modern phenomena, and in particular were effected by the local availability of 'coin' - silver and gold. The general trend though was perhaps 50% inflation per century, so prices always have to be compared to the local prevailing wages to determine prices in terms of constant units. Even this can be misleading, because industrialization throughout the period changed which jobs were considered 'high skill' or 'high demand', meaning that it's hard to look at any one profession and determine overall trend lines in wages.
3) To make matters worse, you are seeing all of this filtered through the lens of D&D, which has infamously maintained two separate price lists basically since the Gygax era - a 'realistic' price list for NPCs and a 'gamist' price list for PC's designed to drain off wealth and keep acquisitions in balance. The realistic price list - taxation, wages for non-expert hirlings, costs of non-adventuring gear - tends to be prices in silver pieces based off historical data. The gamist price list - adventuring gear, cost of lairs, costs of expert hirlings - tends to be based off of gold pieces and set by game considerations. So when you try to answer, "What should a horse cost?", you are trying to translate for one economy (the NPC economy) to the other (the PC economy).
We could go really deep into medieval economics to try to answer that question, but I personally have found that in the long run that's impractical. It also ignores that most D&D world's aren't remotely medieval and have access to highly advanced technology that would totally change medieval economics. This is true regardless of how medieval your trappings get - consider the available 'health care' technology of a D&D world, or the available technology implied by all those mechanical traps found in dungeons, or what a druid implies about agricultural technology, and so on and so forth. Essentially the D&D world is a world where all the medieval understanding of 'physics', medicine, chemistry and so forth is basically true.
Instead, what I find is a more useful method is to note that the price of produce, livestock, hand made goods, and really anything that isn't directly manufactured by machines is reasonably constant over time to within an acceptable degree (certainly within the margin created by the assumption magic is real and effective). So what you can do is take the price of a cow or a horse or a hand tooled leather saddle or jewelry on etsy or Amish made furniture whatever it that you are looking at, and divide that price by some assumption about wealth (ordinary people make the equivalent of $50 a day give or take depending on how wealthy you want your society to be), and produce a price in 'day wages' for the item. You can then make a simple assumption about what a day's wages are - 1 silver piece for example - to produce a price.
This works really well across the board IME any time you need a price quickly (and have Google to hand), and the only thing you have to do is introduce a multiplier of some sort (x3 works pretty well) if you suspect the price is suppressed by some industrial factor. For example, you can assume fight grade plate mail is worth perhaps 1/3rd of what it once costed because while most is still handmade, the cost of steel itself is greatly reduced because it can be made in an industrial fashion very cheaply. So, you might notice a handmade steel helmet runs $200-$1000 dollars on ebay, and translate this quickly to 12 s.p. - 60 s.p. And you might quickly compute that this means that it probably took a medieval craftsman 4-15 days to make it by assuming he's earning about 3 s.p. a day. Will you be wrong? Almost certainly! But you won't be that wrong, and you certainly won't be as wrong as I've seen some people be either relying on D&D prices or trying to reconstruct the whole of a medieval economy based on scant and often erroneous evidence.
Doing this for horses, I'm seeing horses in D&D's 'riding horse' to 'light warhorse' category running between $1000 and $6000. That translate to 40-120 days wages under my simple conversion, and which is 40-120 s.p. under my other simple conversion (or 40-120 g.p. for most tables, since most tables use "treasure is in gold" standards). Am I spot on correct? Probably not. Do I care? Not really. I'm close enough. By the same standards, I figure a PC will spend about 10-20 s.p. a month in upkeep for the horse, unless he owns his own good pasture land and doesn't do a lot of travelling.
