I was doing some light reading about the conflict and this question kept bothering me. When reading about individual battles, I noticed that historians seem somewhat skeptical about the Japanese casualties assessments. I also know that after defeats in the Pacific (like Midway), the Japanese went to great lengths to cover up the scale of their losses. Meanwhile, everytime I have seen someone write about Japanese losses overall in China, they seem to depend entirely on the Japanese claims. This strikes me as a bit suspect. Am I right to be suspicious? I know the Chinese were often tragically outmatched and that the data for the conflict is pretty lacking overall. But surely some historian has done a deeper assessment of this.
I’m not familiar with Japanese casualty lists, so you might need to wait a bit longer for a more direct answer, but I can respond to why historians tend to rely more on Japanese assessments in the China theatre. Casualty numbers come from two sources - well, the two opposing armies. In terms of the Second Sino-Japanese War, historians are reluctant to rely on Chinese (that is to say Guomindang and Communist) numbers.
For the Guomindang, casualty lists were in theory compiled by staff officers (參謀). They were required to submit battlefield reports to the General Staff (軍令部) once every day through telegram or phone, which would include friendly and enemy casualty numbers. However, the quality of staff officers were uneven, as a sizable minority came from regional military academies instead of the Guomindang’s Whampoa Military Academy. Many were simply incapable of carrying out battlefield reconnaissance or observation work, and it was not uncommon that Commanding Officers had to fill in intelligence reports themselves. In one of his many tirades, Chiang Kai-shek raged against the incompetency of staff officers (translation mine):
「现在各部队最大的恶习，就是情报不确，计划不周，文饰战况，虚报胜仗......我们过去失败的最大原因，就是侦探无能，情报不确，甚至没有材料，敌情无从判断……敌情不明，糊涂应战，每战失败，这真是我们参谋人员莫大的耻辱！」
The biggest bad habits of our troops right now are inaccurate intelligence, poor planning, made-up battlefield conditions, fake reports of victory...the biggest reasons for our past failures are useless reconnaissance, inaccurate intelligence, and sometimes even the very lack of intelligence material, leaving enemy conditions hard to determine...with uncertain enemy conditions, befuddled battlefield preparations, losses in every battle - this is an absolute humiliation for our staff officers!
Inaccurate or even fake reports was indeed a notorious element of Guomindang military command. Inflation of Japanese casualty numbers by officers eager to gain promotion was particularly common, with reported ‘victories’ often involving minimal contact with the enemy. For example, the Cantonese warlord Yu Han-mou declared his defence of Shaoguan in Guangdong in January 1940 to be a 'victory' - in actuality, it was a voluntary retreat by Japanese forces to shorten defensive lines, with no Guomindang interference whatsoever. Casualty calculations were even worse during times of retreat. With Guomindang generals and officers often the first to retreat from battle, the compilation of casualty lists were the furthest things from their minds.
The general inaccuracy of Guomindang military statistics has led to a heavy reliance on Japanese assessments to determine the extent of Japanese casualties. However inaccurate they might be, historians simply don't have anything more better to rely on other than Japanese internal casualty lists.
Sources:
葉銘。〈業務紐帶：淺析抗戰時期聯絡參謀與參謀視察制度〉。《民國檔案》，2020(02)：106-115。
彭玉龙。《抗日战争中国军队伤亡调查》。北京: 中共党史出版社，2016。