The Accounting Quest of Steve Keen

Economics stands before a summit whose existence it is largely oblivious to. It can’t see the mountain because of a self-imposed fog. It sneers in the general direction of the summit, without really understanding what it is sneering at.

One great exception to this is the book:

‘Monetary Economics’ by Wynne Godley and Marc Lavoie

These authors have scaled the summit in question. They have demonstrated skills that are too rare in economic thinking – first, they are very knowledgeable in the detailed logic of financial and macro level accounting, and second, they understand the fundamental importance of accounting to economics. The idea of accounting “closure” resonates throughout the book. It is at its core the simple idea that balance sheets connect to each other through double-entry bookkeeping over space/time, with the facility of consistent income and flow of funds accounting as the required linkages for change. It acknowledges and uses the powerful logic that accounting is not just a rear view measurement device – it is also a constraint on all forward looking projections of economic outcomes – meaning that it is an important condition in the substance and shape of good economic analysis. If economics were to become fully congruent with real world observation, such a book would be viewed as ground breaking for the precision with which it weeds out some notable mainstream economic fallacies. And that might help pave the way for what should be a more cohesive fusion of accounting, finance, and economics as a connected and interdependent set of disciplines.

There is no short cut to head-on recognition of the importance of accounting to economics. There is no rationale for departing from the correct logic and method of double-entry bookkeeping as it exists – no reason to “spice it up” with creative but anomalous departures from what it actually is. Double-entry bookkeeping is not “a work in progress”. It is what it is. The neo-classical concepts of exogenous money and the money multiplier and loanable funds and ISLM and supply/demand equilibrium are part of the fog within which mainstream has constructed some economic imagery that is in fundamental conflict with the facts of accounting logic and real world financial measurement. This is all documented pretty cogently in ‘Monetary Economics’.

Another economist, Steve Keen, is attempting to scale the same summit from a different angle, although he writes reluctantly about having until recently “avoided the dry and dusty topic of double-entry bookkeeping”. Steve has written a post on quantitative easing, with an admirable bent towards the importance of accounting in that context:

Steve Keen: Is QE Quantitatively Irrelevant?

He says, “I’m open to correction that I’ve wrongly characterised what banks do or how QE works here.”

I think there is definitely a range of mischaracterizations in Keen’s post about how the banking system responds to QE in both operations and accounting.  So I’m going to offer some comments, as invited. While his construction departs from the actual case, I do have respect for the general direction of his pursuit.

As backdrop, there is a systematic tendency in the blogosphere and elsewhere to misrepresent the impact of QE in a particular way in terms of the related macroeconomic flow of funds. Keen has made this sort of error. Most descriptions will erroneously treat the macro flow as if banks were the original portfolio source of the bonds that are being sold to the Fed, obtaining reserves in exchange. This is not the case. A cursory scan of Fed flow of funds statistics will confirm that commercial banks are relatively small holders of bonds in their portfolios, especially Treasury bonds. The vast proportion of bonds that are sold to the Fed in QE originate from non-bank portfolios. The functional role of banking in this context is to act in effect as a broker of bonds between these portfolios and the Fed. Many descriptions of QE instead erroneously suggest the strong presence of a bank principal function in which bonds from bank portfolios are simply exchanged for reserves. In fact, for the most part, while the banking system has received reserve credit for bonds sold to the Fed, it has also passed on credits to the accounts of non-bank customers who have sold their bonds to the banks. This is integral to the overall QE flow of bonds.

It is also the case that many writers when explaining bank reserve operations have a propensity to overlook the banking system’s critical role as the original distributor of Treasury bonds to non-banks. Most bonds issued end up in the portfolios of non-banks, not banks. The QE function that is the subject here may be viewed as a “reverse distribution” version of that flow, or a gathering up of bonds that are sold to the Fed (although not back to Treasury) mostly from non-bank end holders. In both cases, commentators often overlook the end points of the flows that define the true macroeconomic distribution effect of such Treasury bond flows.

Given this acknowledgement of the full span of the QE bond transfer, it must accordingly be the case that the Fed’s activity in outright purchases of bonds in QE is most often associated with banking system reserve creation AND bank deposit liability creation, since for the most part it is non-bank portfolios that are ultimately affected by these bonds transactions.

Thus, it is incorrect to suggest as Keen does that QE has no effect on the broad money supply. But the effect that it does have occurs through a mechanism that has nothing to do with the erroneous textbook “multiplier” explanation of deposit expansion. Rather, it reflects the more or less immediate bank deposit liability impact that occurs in conjunction with QE reserve expansion. Because most of the bond flow into the Fed does not originate ultimately from bank held bond portfolios, it must be the case that reserve expansion is offset elsewhere on bank balance sheets. The first instance of such an offset is on the liability side of banking, simply because ultimate sellers of bonds are receiving money credits for those bonds.

This dynamic may not be so obvious in the observation of macro statistics during the financial crisis and the overall period of QE. That is because the bank deposit liability effect of reserve expansion during QE has occurred against a backdrop of general deleveraging of bank balance sheets. Thus, cumulative QE has not necessarily shown up at all times in equal amounts of additionally robust net growth in bank deposit liabilities. Loan repayments in the process of deleveraging and other balance sheet processes such as bank recapitalizations, have had the effect of reducing some bank deposit balances in a way that might not have been the case in the absence of a financial crisis.

In fact, the effect of QE has been to provide the banking system with a considerable replenishment of deposit liabilities that would otherwise have been lost or at least grown more slowly in the counterfactual case without QE. The counterfactual case in the midst of the financial crisis would likely have involved an initial vicious deleveraging of the size of the banking system balance sheet, where absent the injection of QE reserves on the asset side, there might well have been a more damaging outright contraction of deposits on the liability side.

When depicting QE in his computer program ‘Minsky’, Keen introduces a strange form of repo accounting on bank balance sheets. This accounting is incorrect. It simply does not exist in the real world – not as a direct feature of QE transactions in the way he has described. The most detailed dissection of such a complex accounting distortion would be more complicated than is warranted here. But here are some of the basic points concerning what he seems to have constructed:

First, as noted, most of the QE bonds are sourced ultimately from non-banks, so repos appearing as bank liabilities would not be an issue even if the accounting were correct, which it is not.

Second, the Fed purchases bonds outright in QE. There is no repo activity as a direct feature of such purchases. There is no repo commitment on the part of the Fed to sell bonds back to those who provided them to the Fed, and such repos do not show up in either the micro or macro level accounting for QE bond purchases by the Fed. In fact, the Fed will decide if and when it wishes to sell back its QE bonds to the private sector as a function of a QE “exit” strategy. Moreover, many of the bonds may end up being matured on the Fed’s balance sheet instead of being sold back. So it will be the Fed’s option to sell in the case of both the occurrence and the timing of any such exit transactions. There is no legal or economic obligation, as is the case with actual repo transactions. Thus, the repo entries that are presumed in Keen’s presentation are quite erroneous as a reflection of reality. And he has compounded this accounting error with respect to the repo entry on commercial bank balance sheets by first creating a repo liability as the offset to an increase in QE created reserves (which is wrong), and then forgetting that under his own assumptions there is first a decline in bank holdings of bonds due to the QE sale of bonds from bank trading inventories into the Fed, which requires a single accounting entry as an asset reduction in its own right. Triple entry accounting doesn’t work.

In fact the accounting and operations that correspond in a correct way to what actually happens in QE are very straightforward, involving the transfer of outright ownership of bonds, the creation of bank reserve credits, and in most cases the creation of deposit liability credits, as described above. To insert a convoluted imaginary repo sequence in the midst of this quite normal bookkeeping is simply to distort the case of actual operations and accounting treatment for QE.

(It is worthwhile noting that the operations of the Fed in a normal, non-QE environment include the outright purchase of bonds in conjunction with the secular growth in banknote liabilities on the Fed’s balance sheet. Commercial bank customers who go to their banks to get banknotes pay for them with commercial bank debits to their deposit accounts. Banks pay for their purchase of those banknotes from the Fed with central bank debits to their reserve accounts. This typically creates a shortage of reserves in the system (i.e. in pre-crisis, pre-QE monetary mode). The Fed, in order to control interest rate levels, re-injects reserves into the banking system by purchasing bonds. The end result is that assets and liabilities on the Fed’s balance sheet expand together – newly acquired bonds along with newly issued banknotes. Given that the location of most of the end-sellers of bonds in the case of banknote expansion is the same as it is for QE (non-bank portfolios), it is also the case that the Fed’s purchase of bonds in the case of banknote liability expansion tends to replenish the deposit liabilities that were extinguished as a result of the original purchase of banknotes by the public.)

In reference to the issue of repo in general, we should emphasize that repo is obviously an important institutional element in the macro flow of funds. But it is not central to the Fed’s balance sheet when viewed in quantitative (stock) terms – in either regular or QE monetary environments. In fact, Fed system repos – which provide funds to the system and which would be the natural location of any Fed repo activity directly associated with QE asset expansion, currently track at roughly $ zero, due to the massive QE reserve liquidity position of the banking system, and the corresponding absence of any material requirement for Fed repo activity (in quantitative stock terms) for purposes of fine tuning system reserve levels. The fact that the Fed now pays interest on outsized excess reserve balances that have been created by QE means that such repo activity is no longer a critical adjustment mechanism for excess reserve management as is the case in normal non-QE environments. But we should hasten to add that even in non-QE environments, the outstanding balance sheet stock of Fed repos is not a material presence when compared to the stock of outright held bonds and corresponding banknotes issued, as described above.

Third, we turn to what must now be noted as an accounting error of extreme proportions – which is that it is certainly not the case that the Fed draws on its equity account when it acquires assets. The Fed creates reserve liabilities as a result of the payments process that is used in the acquisition of assets. The phrase “loans create deposits”, which has become popularized in the endogenous money view of commercial banks, applies equally to the Fed in its own case of asset acquisition and reserve creation. The equity account is not touched in such a transaction, just as it is not touched when a commercial bank makes a loan and credits a deposit to the borrower’s account.

This type of accounting error is symptomatic of a misunderstanding of how an equity account works in the more universal context of double entry bookkeeping. The balance sheet equity account is where income accounting intersects with balance sheet and flow of funds accounting. At the margin, revenues and expenses on the income statement have incremental and decremental impacts on equity, respectively. These effects are summed up in periodic income accounting, resulting in a cumulative credit or debit to equity, depending on whether the result is a profit or loss for the period.

That said, the vast bulk of financial flows captured in flow of funds accounting (the purpose of which is to account for changes in the composition of balance sheets) has no direct effect on the equity account. Again, for example, the mere act of commercial bank lending and deposit creation, or Fed bond acquisition and reserve/deposit creation, is captured (as a flow) only in flow of funds accounting. This type of transaction does not touch the income statement directly and therefore does not touch the equity account directly. The blogosphere (including some economists), in seeking to understand the nature of macro flows, shows an occasional propensity to confuse these two basic, different accounting modes – income accounting and non-income flow of funds accounting. This is too lengthy a topic to pursue in more detail here, but some of us have pointed to this sort of error on a fairly regular basis in the past.

By way of contrast, the Fed does spend from equity (at the margin, via the income statement) when it pays for its own types of accounting expenses – salaries, purchase of regular goods and services in running its daily operations, and even payment of interest on reserves. These items are recorded as expenses on the income statement of the Fed.

And we should remember more generally of course that interest revenue and expense on the banking assets and liabilities that are first recorded in flow of funds and balance sheet accounting are then recorded as subsequent effects through income and equity accounting.

Steve Keen also notes his objective to establish a new definition for aggregate demand by equating it to income plus the change in debt. This entails embedded accounting confusion. Notwithstanding the evidence of impressive historic correlations and causal connections between changes in debt and economic outcomes, it is nevertheless incorrect to add income and flow of funds (i.e. a change in debt in this case) in any expression deemed to be an equation or an identity. Such an expression at best can serve as a regression function, in this case relating current period income to prior period income plus the change in debt. It includes accounting variables that are simply incompatible in an additive sense for an actual equation to hold, either during a single period of time or at a single point in continuous time. Income and changes in debt are orthogonal accounting measures. Flow of funds accounting where the balance sheet equity account is not involved (e.g. increases in debt) cannot intertwine indiscriminately with income accounting in any fashion that could be considered stock/flow consistent. Moreover, there is the problem that changes in debt may well be a source of finance used, not for spending as captured in NIPA, but for asset acquisition as captured only in the flow of funds accounts. This component of the use of funds in the economy has nothing directly to do with income accounting. Finally, the notion that there can be such a strict relationship in terms of income and the change in debt overlooks the fact that aggregate demand can fluctuate at times due to changes in money velocity, without necessarily involving changes in debt. In either case, it is potential and actual spending rather than a change in financial stocks or the utilization of existing stocks (i.e. money) that is inherent in the idea of aggregate demand.

I should note again that ‘Monetary Economics” by Godley and Lavoie is crystal clear and comprehensive in its appreciation of the differences between these two accounting modes, and of the importance of that distinction to the understanding of stock/flow consistent projections or scenarios for future economic activity.

There are other errors in how Keen combines his interpretation of repo accounting with some odd suggestions about equity account effects. The repo margin he anticipates (incorrectly in this case) could not in any case become an ex ante entry in the equity account. As a general rule, neither the Fed nor commercial banks are permitted to capitalize net interest margins attributable to their balance sheet positions in assets and liabilities. Interest margins are accrued to the income statement over time, and thence to the equity account (after allowing for other expenses) as they are earned. This again is another dimension of standard micro financial and macro level NIPA accounting, in which it is important to distinguish between marked-to-market asset values and what is known as accrual accounting for income and expenditures. Repos even when accounted for correctly earn an interest margin on an accrual basis. And even if repos were applicable in this case, accrual of earnings to banking equity actually decreases the money supply. It does not increase it. This is true of all equity accounting in the banking system. Money effectively converts from deposit liabilities to bank equity as bank profits are generated and credited to the equity account. This is the result of interest being charged to borrowers (who pay the interest from their deposit accounts) being greater than interest being paid to depositors.

Next, whether or not banks increase the money supply by expanding their balance sheets through the acquisition of equity securities is a rather random take on bank balance sheet management as it may be affected by QE. We should understand the broader truth of the basic (post Keynesian) insight that banks don’t lend reserves. Neither does the banking system transform reserves in any technical way into holdings of equity securities when acquiring such securities from non-banks. Both lending and securities purchases are flow of funds decisions that in themselves do not require system reserves, but rather are driven by risk assessment combined with bank capital allocation commensurate with that risk. The reserves are for payment purposes, and remain in the banking system when used to make payments. So, to the extent that bank reserves are a non-issue for lending, they are also a non-issue for the purchase of equities, with or without QE. Either of these activities can expand bank balance sheets, including deposit expansion on the liability side, and both require capital management in order to justify them as risk taking activities.

Moreover, in fact the U.S. commercial banking system’s holding of equity securities is minimal – less than $ 100 billion – and there is no evidence to suggest that QE has played any meaningful role in a decision for banks to expand their holding of equities in a material way. This makes eminent sense, in that equity securities are the riskiest financial asset class. Therefore, the fact that bank reserves play no role in either the analysis of risk or the allocation of required capital in making asset mix decisions holds a fortiori when comparing the case of equity security acquisition to bank lending. And, similar to the case for bonds, nearly all of the QE effect on the distribution of equity security holdings (in this case an indirect effect) will have been reflected in the distribution as it is observed mostly in non-bank portfolios.

Also, a repo account is in no way some sort of quasi-reserve position for bank lending, as seems to be suggested in Keen’s construction. One of the puzzling things about Keen’s model is that he seems to have built in a capacity to view banking system lending or buying equities as an alternative to holding reserves. This is very odd, in that it conflicts notably with the normal post Keynesian attitude to the functional role of reserves. The functional requirement for any form of risk taking through balance sheet expansion is capital, not reserves.

There are a few other minor and mostly optical points of error. Double-entry bookkeeping doesn’t record liabilities and equity as negative amounts. That particular sign orientation is used in an “adding up constraint” under a Godley/Lavoie matrix approach. In effect, the balance sheet is represented as a column vector of positive and negative numbers in such an approach, rather than two columns of positive numbers adjacent to each other, which is what is done in actual financial accounting. The purpose of the Godley/Lavoie matrix approach is to capture the fundamental accounting constraint whereby balance sheets must balance, in a way that translates to a vector form that sums to zero in such an “adding up” visualization. That is a convenient mathematical transformation of an actual double-entry bookkeeping visualization. Double-entry bookkeeping incorporates positive signs for assets and liabilities, so that assets equal liabilities plus equity. In particular, increases in equity (such as retained earnings) do not show up as a negative number in double-entry bookkeeping.

Finally, the Fed’s equity account in reality does not reflect “the value of its charter”. It is a standard financial account. It is the cumulative result of original capital injections and any retained income effects (which will be small due to remittance of Fed profits to Treasury). In fact, the equity account is where losses would be absorbed, should they occur, just as is the case with a commercial bank. That said, this “charter value” idea is a small point that is arguably not so important in the context of using a simplified model for the Fed balance sheet. Nevertheless, the factual operation of the equity account in the context of balance sheet, income and flow of funds accounting is an important general point of understanding.

I’ve already noted where I’d look for the highest standard of coherent accounting in the context of full macroeconomic modeling (Godley/Lavoie). Their book ‘Monetary Economics’ presents an integrated accounting and economic framework, featuring a rich supply of income and balance sheet simulations covering various types of economic behavior, where the outcomes are in full accordance with the actual accounting facts of the monetary system. I notice that Keen’s presentation includes something he calls the “Godley Matrix”. I tend to doubt that Godley would have employed the type of accounting that Keen is attempting to develop, an inference I come to naturally from reading the Godley/Lavoie book.

Steve did invite input regarding his method, and I’ve provided some here. More power to him in his pursuit of a robust post Keynesian oriented accounting emphasis, in the interest of advancing facts that need to prevail over neo-classical misconceptions that only obscure a factual view of the monetary system at work. But I suggest in the strongest way that the raw material to get such an effort on a solid footing already exists in an accounting framework that is well established in practice at both micro and macro levels.


A common reaction of mainstream economics to the suggestion that it has stumbled in the case of solid macro accounting requirements for some of its theory is for it to become rather snooty about the role of “the accountant” as a professional activity separate from economics. This is the sneering I referred to earlier. Such a purportedly “sophisticated”, self-elevating view avoids the substantive issue of how in various identified cases economic theory is simply not compatible with real world macro accounting closure requirements. Again, I refer the reader to Godley/Lavoie’s ‘Monetary Economics’ in order to get a sense of the full scale and meaning of this phenomenon.

Like Steve Keen, I am not, and never have been, an accountant.

Given a normal aversion to the excessive use of metaphors, I’ve refrained from expanding on the image of “the summit” with such bells and whistles as death zones, frozen bodies, reckless tourists, “into thin air”, and wise Sherpas. That would require an accounting extension of its own kind – which admittedly is quite tempting.



View all posts by

Leave a Reply

494 Comments on "The Accounting Quest of Steve Keen"

Notify of

3 years 11 months ago

Excellent, and very much appreciate the commentary on bank vs non-bank asset sales as an important detail often glossed over in discussions of QE. It seems to me that it is much easier to make the case that QE doesn’t impact aggregate demand in any fundamental sense when banks are the original sellers of bonds, given what we know from the ‘loans create deposits’ view of the world. And this is often the assumption made when discussing the impact of QE. The issue seems much more complex when we consider non-banks as the sellers of bonds. Scott Fullwiler and Cullen have broached this issue in the past (with respect to propensity to spend out of wealth, behavioral ramifications of having deposits vs bonds, portfolio shifts, etc.). I think Ramanan did a recent post on this as well. I’d be interested to hear your take, if you have more to add. I also feel like this is an issue where empirical analysis / behavioral finance might be able to provide useful information rather than hopelessly confounded analysis.

3 years 11 months ago

“And this is often the assumption made when discussing the impact of QE.”

Right – see my comment to Cullen above and then further below.

3 years 11 months ago

I would suggest that any effects from QE cannot be seen by looking at sellers of bonds.

They were selling anyway and therefore their behaviour has unlikely to have changed. Possibly there may be shifts in timing to gain advantage of the increased demand.

I would suggest that the effects of QE would come from the potential buyers of bonds who were outbid. What did they do instead?

QE increases the number of people bidding for Treasuries, because it adds a load of reserves to the primary banks – as offsets to the increase in liabilities to the secondary institutions. The primary banks will then try and swap them for Treasury bonds. You’ll notice that the balance sheets of the primary banks have more Treasuries in them than they did prior to QE. So you get even more buying pressure on Treasuries.

In the UK we have a lot of the primary banks as the market makers at the Gilt Auctions. So it’s fairly easy for them to push their own book and bid up the Gilts.

All the QE documentation seems to focus on sellers of bonds, and I don’t understand why.

3 years 11 months ago

What are the accounting entries when banks purchase bonds from entities other than the Fed?

3 years 11 months ago

“You’ll notice that the balance sheets of the primary banks have more Treasuries in them than they did prior to QE.”

I did notice that and it’s an interesting point. It’s a gradual ramp up but it’s not huge as part of the balance sheet mix (much smaller than the reserve increase), and it’s leveled off.

It underlines the point that the commercial banks haven’t been the net portfolio source of QE bonds – quite the contrary. My guess on that is that with the system on the mend in terms of internally generated capital, and with the Fed steadily telegraphing low rates, they became a little more adventurous in taking some positions along the yield curve. Those positions do require capital to support interest rate risk, but that’s less than for credit risk.

3 years 11 months ago
wh10, “It seems to me that it is much easier to make the case that QE doesn’t impact aggregate demand in any fundamental sense when banks are the original sellers of bonds, given what we know from the ‘loans create deposits’ view of the world.” But that is true even when non-banks sell the bonds to the Fed. It doesn’t *directly* impact aggregate demand even though there is an effect on the money supply aggregates such as M1 etc. This is because monetary aggregates are not causes of anything – more of residuals. There is of course a portfolio balance effect as Bernanke himself argues. In my opinion it has a strong effect on equity prices. (Actually there is one more thing I ignored in my post on asset allocation – which is the reflux mechanism where economic agents extinguish their liabilities from the proceeds of the sale Here is from the Fed FAQ: “When the Fed buys an asset, the effect on the broad money supply depends on who sold the assets and what they do with the funds they receive. If the seller is a bank, reserves go up, but broad money only increases if the bank responds to the increase in its reserves by lending more to households and businesses. If the seller is an investor other than a bank, reserves go up, and broad money also goes up in the first instance as the seller’s bank puts a sum equal to the amount it receives from the Fed into the seller’s bank account. But if the seller uses the money to pay down debt, the broad money supply declines again by the amount of the debt repayment. As of early 2011, the behavior of the broad money supply, economic activity and inflation all suggested that recent… Read more »
3 years 11 months ago

Agree – see comment further below.

3 years 11 months ago

Ramanan, thank you for that NY Fed clip. The destruction of deposit money through debt repayment — principal, interest, and fees — is the story of our time. It supports my contention that deficit-funded stimulus is plutonomic, as the base money created is eternal as an asset of commercial banks, while the deposit money created is ephemeral as an asset of the public, subject to destruction on any pretext. Thus a $35 overdraft fee is a permanent subtraction from broad money.

JKH, thank you for a classic post.

3 years 11 months ago

thank you

3 years 11 months ago

The way I usually describe the non-bank seller situation is that it doesn’t change aggregate demand because it doesn’t alter any necessary component of the private sector’s balance sheet. If spending is a function of incomes relative to desired saving then QE via non-banks would have to alter some element of that equation. I always emphasize that QE is not enabling govt spending because that implies there’s a lack of demand for t-bonds in a non-QE world. We know that’s wrong because QE2 ended and yields fell. So QE can’t increase incomes in any real meaningful way aside from some of the residual effects of the portfolio rebalancing and the interest rate impact. QE has a strong psychological impact that potentially drives up asset prices, but I don’t think there’s strong evidence supporting the sustainability of such moves as the underlying fundamentals of the assets whose prices are in question appear to have no real transmission mechanism from which to be altered.

I’m not in the camp that says QE does nothing, but I do think it’s a lot less impactful than most others. I think fiscal policy has probably done most of the heavy lifting in recent years and QE has helped with varying degrees in the different ways its been implemented (for instance, QE1 was much more supportive of the economy than QE2 & QE3).

I’d love to see the USA in a weak credit environment, with a balance budget, a current account deficit and massive QE. Then we’d see where the rubber meets the road. My guess is that the economy would perform very poorly in such a situation assuming QE was impacted as it currently is being implemented…..

3 years 11 months ago

“I’m not in the camp that says QE does nothing”

Agree with that. One can take the “banks don’t lend reserves” meme too far in the case of QE – this is not necessarily the right question or the only question. Beyond that, I’ve probably spent more time pondering the nature of the right questions than considering the best answers at this stage.

Regarding aggregate demand, there is the wealth effect to consider, in addition to the income effect.

More general comment below.

Cullen Roche
3 years 11 months ago

Great post, JKH!

1 11 12 13