原创 2016-05-18 季天鹤央行观察

作者：季天鹤，方正中期期货利率汇率研究员、央行观察专栏作家

由于政府存款账户开在央行负债侧，政府对银行发行债券本身减少的是银行在央行存款，M2的下降恐怕另有隐情。

财政存款为什么不在M2里？

关于财政存款，熟悉中国流动性体系的读者一定知道两点：一是中国的财政存款账户开在央行负债侧，二是财政存款不在M2的统计口径里。上述两点，体现在央行的三张表上，就是财政存款在货币当局资产负债表负债侧，其他存款性公司资产负债表负债侧没有财政存款，以及存款性公司概览中财政存款被放在对政府债权（净）这个项目里面。

用公式总结，对政府债权（净）=货币当局对政府债权 + 其他存款性公司对政府债权 - 货币当局负债侧政府存款。该公式忽略了政府持有的纸币硬币，这一项目应该和货币当局负债侧政府存款一起被减去。但考虑到央行自己无法掌握政府持有的纸币硬币数量，因此只能使用前面这个不够严谨的公式。

从流动性的层级上来看，财政存款和货币发行以及准备金存款的等都是央行负债，而不像企业和居民存款一样是银行的负债。从M2的口径上看，M2包括流通中货币以及银行存款，而流通中货币乃是货币发行减去库存现金，也就是说银行手里的纸币硬币不算M2，而财政存款也不算M2。

M2这个口径和流动性的层级并不完全对应。M2口径是人指定的，但流动性层级为会计准则和清算结构决定。最简洁漂亮的结构，要么是没有纸币的，M2完全对应银行负债；要么是全纸币，M2完全对应央行负债。上述不对应所遗留的问题，就是财政存款、准备金存款和库存现金怎么办。

如前，对政府债权（净）忽视了政府持有的纸币。如果政府在央行完全没有存款，国库完全体现为政府自己持有的纸币，那么对央行来说，相当于政府存款被新增的货币发行完全替代，本身资产负债规模并无变化。但M2会增加，因为政府持有的纸币是流通中货币。准备金存款和库存现金也面临同样的情况。如果准备金存款也变成了库存现金，银行就和只持有纸币的财政处于相同的地位。

为什么财政存款、库存现金以及准备金存款都不被放在M2当中呢，我怀疑一个重要的原因，是因为他们在央行负债侧。准备金存款和库存现金被看作是对应了银行存款。储户取现会同时减少银行存款和库存现金，增加流通中货币，这个过程里面的M2是不变的，只不过流通中货币增加，银行存款减少而已。如果银行存款和库存现金都是M2，那么储户取现，银行存款和库存现金减少，流通中货币增加，M2会净减少，这恐怕是非常反直觉的，因此也没有被统计口径的制定者接受。

之所以“反直觉”，因为这种直觉乃是基于纸币是货币的前提。在那个只有纸币才能流通的时代，纸币是M0，存款等等都是M1-M0乃至M2-M1。个人存款被看作是不可流通的，需要取出纸币才能购买东西，于是被放在M2-M1里，这显然已经完全脱离了2016年的现实情况。

一旦意识到银行存款乃至非银行金融机构负债份额（例如余额宝份额）也可以流通之后，我们就不会觉得银行存款和库存现金同时都是货币很奇怪了，也就不会觉得余额宝在银行的存款与我在余额宝的份额同时都是货币很奇怪了。余额宝管理人就是可以拿着银行存款缴手续税，我就是可以拿着我在余额宝的份额买早餐，银行就是可以拿着库存现金发工资。

其中的奥秘，当然是准备金率的变动以及货币创造的机制。存入纸币时，银行创造了存款货币。取出纸币使其进入流通并减少存款，是减少了存款这种货币的规模。纸币转手不应该是纸币成为货币或者不成为货币的理由。注意，纸币转手不是纸币回购，也就是说纸币转手都是在央行以外的主体间进行。

或许会有读者认为，更改了M2的口径，就一劳永逸了。诚然，M2的口径会随着历史的发展而修订，但比口径更重要的乃是其背后的深层机制，而不是表层的统计数据。我们需要认清为什么M2口径曾经是合理的，为什么现在会不合理，未来会不会更不合理等一系列问题，而这些问题其实都是深层的货币银行问题。

财政存款变动如何影响M2?

在过去，财政存款的主要来源是税收、非税收入、基金性收入、国企股利等，这些收入绝大多数都是从M2过来的，例如企业缴纳所得税，企业的银行存款减少，财政在央行的存款增加，同时，企业的开户银行在央行的存款也因转账而减少。这里面的两减一增（企业存款减少，银行存款减少，财政存款增加），和储户取现时的两减一增（储户存款减少，银行库存现金减少，储户持有纸币增加）是完全一样的，原因在于财政存款和纸币都是央行负债。

上面之所以是说绝大多数，原因就在于有两个特殊的业务需要考虑，一个是财政和央行间的业务，一个是财政和银行间的业务。央行持有国债，财政付息的时候减少财政存款，增加央行的留存收益，而财政收取央行上缴收益时财政存款增加，央行留存收益减少。而央行购买国债，其实不过是像银行放贷一样同时扩充资产负债，在增加国债资产的同时增加财政存款。如果2017年特别国债到期时，财政打算用财政存款还债，那么我们看到的一定是财政存款减少，国债减少。

财政和银行的业务也很特殊，例如银行缴税/上缴利润/购买国债，会使银行在央行存款减少，财政在央行存款增加。财政对银行持有的国债支付利息/偿付国债，会使财政在央行的存款减少，银行在央行的存款增加。财政与银行业务以及财政与央行业务的特点，在于上述操作都和M2无关。

当考虑财政存款对银行间流动性以及M2的扰动的时候，需要仔细考虑具体交易。央行购买国债增加财政存款不会影响银行间流动性，因为银行不参与这一交易。国债发行给银行会影响银行间流动性，但不会影响M2，因为非银行主体不参与这一交易。国债发行给非银行主体会影响M2，但同时也会影响银行间流动性，因为M2会消失，银行在央行存款会转到财政在央行的账户。

传统上我们一般直接考虑财政存款对M2的扰动，以及财政存款对银行间流动性的扰动。这种思考的习惯当然现实基础，即涉及到M2的财政收支主导了财政存款的变化。此时，财政存款的增加就对应了M2的减少，反之亦然。但我们对于国债发行和兑付如何影响M2考虑较少，我自己也是如此，在此前的分析中没有考虑过其中的复杂性。

银行买国债后财政支出资金给企业，和银行直接买企业债券并给企业创造M2在广义流动性上是等价的。具体而言，银行购买100亿债券，本身只是让财政存款增加，还不影响M2。但财政支出后，银行不但拿回了在央行的100亿存款，同时负债侧的企业存款也增加了。整个过程看上去，银行在央行的存款似乎就没动过，只是银行资产多了债券，负债多了存款，和一笔给企业的贷款有同样的效果。在这个问题上，我们不会认为财政存款和M2的变化导致了两者的不匹配，而是会认为银行扩张了信贷，扩张信贷当然是要创造存款的，因此我们就没有把财政存款不变和M2增加这一不匹配加以重视。

而如果财政发债之后乃是发新债偿旧债，例如3月期国债滚动发行，我们会发现财政存款的余额是不变的，而M2当然也没有变化，也就是说两者都保持了不变的状态。而如果财政发新债不是偿还自己的旧债，而是偿还平台企业的旧债，我们会发现如果发新债和偿旧债是连续进行的，那么财政存款余额仍然不变，对M2也不会有影响，对平台企业的影响乃是其债务，而不是其资产侧的M2。

通过存款性公司概览我们可以更好地理解上述操作。存款性公司概览给出了影响广义货币量的所有因素，包括国外净资产、对政府债权（净）、对非金融部门债权、对其他金融部门债权、不纳入广义货币的存款、债券、实收资本、其他（净）等。前面提到，对政府债权（净）=央行对政府债权+银行对政府债权-财政存款，因此财政存款的变化会影响M2。那么一个问题自然就来了：刚才提到的那些影响财政存款但不影响M2的操作，是如何在存款性公司概览里面体现的呢？

答案是：通过影响M2以外的项目显现。财政发债给银行导致财政存款增加，但这个操作在存款性公司概览里面不体现为银行在央行存款的减少，而体现为银行对政府债权的增加。银行上缴利润给财政体现为财政存款的增加，同时体现为银行的所有者权益减少。财政替平台企业还债，财政存款减少并且银行对企业的债权减少。也就是说，在发生了重大的交易时，我们无法把财政存款的变化自身简单看作是影响M2的因素，因为财政存款的变化可能捆绑着M2以外项目的变化。

M2项下的非银金融机构存款

央行称，“4月份置换地方融资平台贷款规模不低于3500亿元，比2、3月份大幅增加，而去年4月份无置换贷款。地方政府存量贷款置换为债券，相应冲抵了贷款增量”，同时又称“4月份财政存款增加9318亿元，同比多增8217亿元，影响M2增速约0.64个百分点。这主要与地方政府债券大量发行以及“营改增”对税收进度产生的影响有关”。

需要注意的是，9318亿财政存款变动并不是存款性公司概览里面的政府存款口径，用来考虑M2的变动可能有所偏差，而央行资产负债表的政府存款变动6755亿。此外，4月份公共财政收支为净收入2400亿，国债和地方债余额增长1.15万亿，贷款置换规模3500亿。地方政府债券对银行大量发行本身并不影响M2，而是同时增加银行对政府债权以及政府存款，政府存款（净）保持不变。

因此简单认为财政存款增加影响了M2增速，在地方债券大量发行这个解释下恐怕是需要进一步分析的。如果地方政府债被银行原样持有而没有卖给非银金融机构，从而没有让非银金融机构在买债当中减持M2口径下的存款，那么M2就不会变化。而如果银行把国债卖给非银金融机构，导致M2变少，那么这个操作就是银行自己的操作，和财政又没有关系了。

如果观察存款类金融公司的存款，我们发现16年4月境内存款新增8660亿，而15年4月境内存款新增7818亿，两者差距并不悬殊。16年4月的住户存款减少9296亿，企业存款增加2206亿，而15年4月住户存款减少10510亿，企业存款增加3942亿，净减少分别是7090亿和6567亿，相差不大。

问题出在财政存款和非存款类金融机构存款上。16年4月财政性存款如前所述增加9318亿，而15年4月仅增加1101亿。同时，16年4月非存款类金融机构存款增加4115亿，而15年4月则是增加10610亿。考虑到15年4月地方政府债余额没有变动，而16年4月地方余额新增万亿，那么很有可能是银行把政府债券卖给了非银金融机构。很遗憾没有地方政府债券的持有者结构数据，但存款性金融机构债券投资在2016年4月份增加9472亿元，而2015年4月仅增加2849亿元。从上述信息来看，万亿地方债应该有不少还在银行的手里，只有一部分或许卖给了非银金融机构，使他们在银行的存款减少了。

4月份的贷款增速下降和货币增速下降，看上去虽然都和财政有关，但其实是发生了两件事情。一件事情是财政的债务置换使得贷款规模下降，一件是银行把国债卖给非银金融机构使得M2减少。财政发债给银行这一操作，虽然导致了财政存款增长，但既不影响银行贷款，也不影响银行存款，也不影响M2。

既然这样，M2的减少以及增速下滑就并没有那么可怕了。非银配置债券导致银行缩表，只是间接融资变成了直接融资而已。宏观层面上居民和企业的货币量同比依然是增长的，乃至快于此前几个月的同比增长率，实体经济的流动性依然是充裕的。M2的减少和增速下滑本身并不代表了货币政策的意图，而只是银行和非银金融机构间操作的自然结果。

以上便是对4月份货币量以及置换操作的一些思考。除此之外，还有一个异常的数据值得关注，即中资大型银行向央行的借款在4月增加1.4万亿。考虑到四大行向央行的借款仅增加632亿，可见超过1.3万亿的借款都是由开行、邮储和交行办理的。但是，央行资产负债表上并没有这样大规模的新增对银行信贷。因此这些借款是什么性质，耐人寻味。

Due to government deposit accounts being held on the liability side of the central bank, the reduction of bank issuance of bonds to the government results in a decrease in bank deposits at the central bank. There might be hidden factors behind the decline in M2.

Why isn't fiscal deposit included in M2?

Regarding fiscal deposits, readers familiar with China's liquidity system are likely aware of two points: firstly, China's fiscal deposit accounts are held on the liability side of the central bank, and secondly, fiscal deposits are not included in the M2 statistical framework. These two points are reflected in the central bank's three balance sheets: fiscal deposits are on the liability side of the monetary authorities' balance sheet, other deposit-taking companies' liability side of the balance sheet does not include fiscal deposits, and fiscal deposits are placed within the item "Government Claims (Net)" in the overview of deposit-taking companies.

Summarizing with a formula, "Government Claims (Net)" = Central Bank Claims on Government + Other Deposit-taking Companies' Claims on Government - Central Bank Liabilities Side Fiscal Deposits. This formula overlooks the government-held currency and coins, which should be subtracted along with central bank liability side fiscal deposits. However, considering that the central bank cannot know the quantity of currency and coins held by the government, a less precise formula must be used.

From the perspective of liquidity hierarchy, fiscal deposits, currency issuance, and reserve deposits are all central bank liabilities, unlike corporate and household deposits, which are liabilities of banks. From the perspective of the M2 framework, M2 includes circulating currency and bank deposits. Circulating currency is the result of currency issuance minus cash in circulation, meaning bank-held currency and coins are not counted in M2, and fiscal deposits are also not counted.

The M2 framework and the hierarchy of liquidity do not entirely correspond. The M2 framework is a human-defined concept, while the hierarchy of liquidity is determined by accounting standards and settlement structures. The simplest and most elegant structure is either one without currency, where M2 corresponds entirely to bank liabilities, or one with all currency, where M2 corresponds entirely to central bank liabilities. The existing incongruity is how to handle fiscal deposits, reserve deposits, and cash in circulation.

Considering "Government Claims (Net)" ignores the government-held currency. If the government has no deposits at the central bank, and the treasury consists entirely of government-held currency, then for the central bank, it's as if government deposits have been entirely replaced by newly issued currency, and the overall balance sheet scale remains unchanged. However, M2 would increase because the government-held currency is part of circulating currency. The same situation applies to reserve deposits and cash in circulation. If reserve deposits become cash in circulation, banks would be in the same position as the treasury, holding only currency.

The reason fiscal deposits, cash in circulation, and reserve deposits are not included in M2 might be because they are on the liability side of the central bank. Reserve deposits and cash in circulation are seen as corresponding to bank deposits. When depositors withdraw cash, both bank deposits and cash in circulation decrease, increasing circulating currency. In this process, M2 remains unchanged; only circulating currency increases, and bank deposits decrease. If both bank deposits and cash in circulation were part of M2, when depositors withdraw cash, both bank deposits and cash in circulation would decrease, increasing circulating currency. This would lead to a net decrease in M2, which is counterintuitive and thus not accepted by the creators of the statistical framework.

This "counterintuitiveness" arises from the premise that paper currency is money. In the era when only paper currency could circulate, paper currency was M0, and deposits were classified as M1-M0 or even M2-M1. Personal deposits were considered non-circulating and required conversion to paper currency for purchases, thus they were categorized under M2-M1, which is quite detached from the reality of 2016.

Once it's realized that bank deposits and non-bank financial institution liabilities (e.g., holdings in money market funds) can also circulate, it's not strange anymore to consider both bank deposits and cash in circulation as money. The mystery lies in the changes in reserve ratios and the mechanism of money creation. When currency is deposited, banks create deposit money. Withdrawing currency into circulation and reducing deposits decreases the scale of that form of money. Currency changing hands should not be the reason for currency becoming or not becoming money. It's important to note that currency changing hands is not currency repurchase; it means currency changing hands between entities outside the central bank.

Some might think that changing the M2 framework would solve the issue once and for all. Indeed, the M2 framework will be revised as history develops, but what's more important than the framework itself is the underlying mechanism, not just the surface-level statistical data. We need to understand why the previous M2 framework was reasonable, why it might be unreasonable now, and whether it will become even more unreasonable in the future. All these questions are related to the deeper issues of monetary banking.

How does the fluctuation of fiscal deposits affect M2?

In the past, fiscal deposits primarily came from taxes, non-tax revenue, fund income, state-owned enterprise dividends, etc., most of which were part of M2. For instance, when companies pay income tax, their bank deposits decrease while fiscal deposits at the central bank increase. Also, the bank's deposits at the central bank decrease due to transfers. These two reductions and one increase (company deposits decrease, bank deposits decrease, fiscal deposits increase) are the same as two reductions and one increase (deposits decrease, bank cash in circulation decreases, currency holdings increase) when depositors withdraw cash. This is because both fiscal deposits and currency are liabilities of the central bank.

The term "most of the time" is used because there are two specific scenarios to consider: transactions between the government and the central bank, and transactions between the government and banks. When the central bank holds government bonds, when the government pays interest, fiscal deposits decrease and central bank retained earnings increase. Conversely, when the government receives earnings from the central bank, fiscal deposits increase, and central bank retained earnings decrease. When the central bank purchases government bonds, it's essentially an expansion of assets and liabilities, similar to bank lending, which increases government deposits. If the government had no deposits at the central bank, and the treasury was comprised solely of currency, it would be as if newly issued currency replaced government deposits, and the overall balance sheet scale wouldn't change. However, M2 would increase because government-held currency is part of circulating currency. The same applies to reserve deposits and cash in circulation.

For the effect of fiscal deposits on interbank liquidity and M2, specific transactions need careful consideration. When the central bank buys government bonds, fiscal deposits increase, but this operation doesn't impact interbank liquidity, as banks are not involved in this transaction. Issuing government bonds to banks affects interbank liquidity but doesn't affect M2 because non-bank entities are not involved. Issuing government bonds to non-bank entities affects M2 but also impacts interbank liquidity. M2 disappears, and bank deposits at the central bank shift to the government's account at the central bank.

Traditionally, we usually consider the disruption of M2 and interbank liquidity by fiscal deposits directly. This habit is based on the fact that fiscal flows dominated fiscal deposit changes related to M2. At this point, an increase in fiscal deposits corresponds to a decrease in M2 and vice versa. However, the effects of government bond issuance and redemption on M2 are often overlooked. In my previous analyses, I haven't given much thought to the complexity of these interactions.

When banks buy government bonds and the government spends funds on companies, it's equivalent in terms of broad liquidity. Specifically, when banks purchase 100 billion in bonds, fiscal deposits increase, but this doesn't impact M2. However, when the government spends, banks not only retrieve their 100 billion deposits at the central bank, but also the liability side of corporate deposits increases. Although it might appear that bank deposits at the central bank haven't changed, banks have more bonds as assets and more deposits as liabilities, similar to extending a loan to companies. In this case, we don't perceive the discrepancy between unchanged fiscal deposits and increased M2 as significant because banks' credit expansion, which inherently involves creating deposits, is the primary driver. Therefore, we don't emphasize the inconsistency between unchanged fiscal deposits and increased M2.

If we recognize that both bank deposits and non-bank financial institution liabilities (e.g., holdings in money market funds) can circulate, the situation becomes less puzzling. Non-bank financial institutions can use bank deposits to pay taxes, just as I can use my holdings in a money market fund to buy breakfast. The underlying principle is that these transactions occur between entities outside the central bank.

Perhaps some readers might think that revising the M2 framework would permanently resolve the issue. Indeed, the M2 framework will change over time, but more importantly, the underlying mechanism matters more than the statistical data. We need to understand why the previous M2 framework was reasonable, why it might not be now, and whether it will become even less reasonable in the future. These questions are all part of the deeper issues within monetary banking.

The fluctuation of M2 and its growth rate doesn't seem as worrisome when considering these factors. Non-bank allocation of bonds leads to a contraction of bank balance sheets, transitioning from indirect financing to direct financing. At the macro level, money supply for households and enterprises continues to grow on a year-on-year basis, even faster than previous months. Liquidity for the real economy remains sufficient. The decrease in M2 and its growth rate doesn't necessarily reflect the intention of monetary policy; it's a natural result of operations between banks and non-bank financial institutions.

These are some considerations regarding money supply in April and the replacement operations. There's an unusual data point worth noting: Chinese large commercial banks borrowing from the central bank increased by 1.4 trillion in April. Given that the four largest banks only increased borrowing by 63.2 billion, it's clear that more than 1.3 trillion in borrowing was handled by other banks. However, the central bank's balance sheet doesn't show a corresponding increase in bank credit on such a large scale. The nature of these borrowings is intriguing and warrants further investigation.