Is word count a ranking factor?
No. Absolutely not!
Any suggestion that word count is a ranking factor is bad SEO and completely wrong. In fact, it is an insult to Google and Bing to suggest such a thing.
There is no limit, short or long, or even a range that influences search except for the ridiculous. The advice to create high quality content stands. Short and concise content, if done well, is perfectly fine.
Looking at the question logically.
Google uses a rule in determining ranking factors. Can the factor be manipulated to artificially boost search performance?
Can word count be manipulated? In a word, Yes. Of course. If word count were a ranking factor, then content would be made long consistently across most sites to gain rank. This is not done. In fact, shorter content can rank very well. This is self evident to anyone who searches Google for any length of time.
Short of extremes, extremely short or extremely long content, the size of the content does not effect how well it performs. There are too many factors that go into indexing and evaluating content for content length to have any effect if it were to. The plain fact of the matter is that content length would, at least, be so insignificant a metric, if used, compared to others to matter.
Here is another rule Google uses to determining a ranking factor. What does the metric say about the content? For example, content quality or how the content should be found, etc. Content length literally says nothing about the content itself. Again, short of extremes, content length cannot logically be used to indicate the quality of the content, how the content should be found, the popularity of the content, etc.
Correlation aside, assuming that there is any correlation to be argued, content length as a metric offers no definitive rule to be made. Long content can be garbage as much as valuable. The same exists for shorter content. Ask yourself, Under what set of rules can content length definitively indicate content quality, value, popularity, etc? You will not be able to find one.
To nail it down.
Some years ago when researching trust factors for a security systems research project for NSF I was working on, Google shared some internal documents that detailed database schema, algorithms, business logic, etc. The reason for sharing the document was simple. There is a large overlap between domain trust factors that Google uses and ones we were developing to determine domain trust on the Internet and Google wanted to see if there were any new factors we were developing and what experiences statistically would help with how Google determines domain trust factors.
Based upon these documents, nowhere was there any metric or algorithm that addressed content length. This bears repeating. Content length is not a ranking factor or even a factor at all.
Where does the word count notion come from?
Primarily from SEO bloggers. SEO bloggers are, for the most part, not familiar with how technology is used in search. Much of what is posted is either parroted from others or anecdotal. Very few actually have researched how search works. They go from their own experience. However, they forget one simple fact. Short content can perform extremely well too.
What does Google say?
Google has a lot to say about satisfying content. Much of what is said is purely conceptual and designed for human consumption. It is not actually how search engines CAN see content. It is a machine afterall. It cannot evaluate content the way we do. The cognitive ability of any computer using artificial intelligence (AI) is extremely low. Instead, search engines use scientific measures many of which date back to the 70's. Semantic analysis of various types, trust networks, link maps, are older technology that search engines rely upon heavily. What is used are technologies and AI that are specific to evaluating and presenting large-scale ontologies of textual information some of which were built examining how humans use, process, and perceive language. Others are about processing information in a manner that allows information retrieval (IR).
Before I begin.
I thought how can I discuss how content performs without getting into the metrics? I looked at my list of metrics and it is far too overwhelming for this Q&A format. Then I thought how can I not get into the technologies such as mentioned above? Again, far too overwhelming. I could write a book! Instead, I will explain a few things you may not have thought about. I will keep it simple. I am working from memory.
How content is evaluated.
Content is evaluated for several things of which topical strength, expertise, reading level, target market, fact statements, semantic structure, citation analysis, etc., just to name a few.
Topical strength is the analysis of specific key terms found in a topical ontology to score what topics a given text is about. Topical strength is the score of all topics, especially related topics, and how focused and complete a given text is on a particular topic.
Expertise is the analysis of specific key terms found in a topical ontology and fact link analysis that indicates the level of expertise a given text presents. This is compared to other writings where expertise has been established. This requires comparison scores with known writings from experts on any given subject. For example, your writings can be compared to research papers from a known expert. In this case, proper use of terms using semantic analysis, topical strength scores, fact link analysis, and other forms of evaluation will indicate your expertise on a topic.
Reading level is the analysis of the required education level to understand a given text. While on some hand this is self explanatory, it is found that highly educated people will write at a targeted level. For example, people with PhDs will often write for their peers. This indicates a level of education and expertise not related specifically to the text. Instead, it will indicate who the text is targeted toward.
Target market is the analysis of markets that indicate who a given text is meant to appeal to. This may be obvious, however, market analysis is used to match a search query more precisely to intent. For example, queries that appear to be looking for SEO advice will lean heavily toward content that is SEO centric. Search queries that are not specific enough can skew results toward an entire market easily.
Fact statements is the analysis using semantics to extract fact statements from a given text and compare them with similar fact statements in other text. Fact statements that pass scrutiny and are accepted compares favorably in search. For example, George Washington's birthday is February 22, 1732. If your content states George Washington's birthday as something different, it will not be found in search. Fact statements are converted into fact links which comprises the knowledge graph. The knowledge graph is simply an ontology of facts seeded with trusted information. For any fact to be added to the knowledge graph, it must be corroborated by several trusted sources. When a fact statement within your content can be corroborated, the content receives a link within the knowledge graph (also known as a knowledge base) as a trusted source for that specific fact. The more fact statements that exist within a given text that can corroborated and linked within the knowledge graph the better that content will perform in search as being factual.
Semantic structure is analyzed and used in a variety of ways. One of the ways that semantic structure is used is to create facts. Facts are not always apparent. For example, "A dozen is 12." is apparent. Whereas, "Brian has an uncle named Pete. Pete has a daughter Diane.", semantic structure and fact analysis can determine that Diane is Brian's cousin and vice versa. Another way semantic analysis is used is to analyze a given text for linguistic expertise. Complex sentences, if structured correctly, can be properly understood and scored. This not only can indicate expertise on a given subject, but also allow complex relationships between facts to exist.
Citation analysis is the analysis where quotations and references are found to be used within a given text that refers to another text. This is often in the form of a quote, but can refer to a title or author of a given work that shows a level of expertise or not. Quoting from one or more authoritative works helps search.
Why did I mention all of this?
Because these all influence how content appears for a given search and applies evenly regardless of the content length. However, how your content scores using the analysis mentioned will influence how your content is found. Here are the two ends of the scale.
Shorter content tends to be fact based and concisely answers a single question. This content works well in the answer engine using the knowledge graph. Search queries that solicit a fact based response will often result in shorter content. This works best when search is trying to answer a search query with one correct factual answer.
Longer content tends to be heavily fact based with a high level of expertise. This often does not work well in the answer engine. Why? Because the answer is not as apparent. However, for more scholarly searches, longer content is often found.
There are exceptions of course. However, I suspect I made my point.
So why do SEO bloggers get this so wrong?
Because most SEO blog posts are relatively short and are shy of fact statements. What fact statements exist many times cannot be corroborated from highly authoritative sources such as research papers and most fact statements can be corroborated only among other SEO bloggers. Other analysis metrics come into play. Expertise analysis falls flat. Reading level is moderate. Topical scores are lacking. However, market analysis is strong. Based upon the experience of SEO bloggers, content within a relatively narrow window of context is created and therefore performs narrowly based upon what is created. SEO blogs are written for an audience that is looking for smaller bits of information. Most SEO queries are not in search of a single fact based answer and therefore rely more heavily on other search metrics. Longer posts tend to do well with more fact statements however, are very narrow in topical scope. It is a fait accompli. SEO bloggers tend to parrot each other in a large echo chamber and therefore few posts offer anything new or insightful. Few ever refer to real authoritative works. This is why SEOs drive me nuts. Most simply do not know how search works.