Macroeconomic implications of expected loss impairment in banking regulation

在一家专业服务公司担任经济学家期间，我支持一家大型英国银行开发宏观计量经济预测模型，以遵守新的国际财务报告准则9会计准则。反思我的硕士论文（《英国固定事件宏观经济预测的效率和无偏性分析》），这个项目引发了一些问题，即用于遵守这一规定的预测中的潜在偏差或低效率是否会对更广泛的金融体系和经济产生影响。金融危机导致银行评估损失的监管方式发生了变化。新的国际财务报告准则第9号要求采用预期信用损失法提前确认损失，这与以前的已发生损失法形成了对比。实际上，IFRS 9要求银行在一系列宏观经济情景下计算预期信贷损失（ECL）的无偏概率加权估计。这是一个重要的研究领域，因为经济对大型银行的依赖，正如2008年金融危机所表现出来的那样。管理人员如何执行监管规定以及做出预测的方法将涉及损失减值和可用信贷水平。关于贷款损失拨备和银行信贷水平对实体经济的影响存在争议：（当前）发生损失的方法被认为有助于顺周期性，因为它在萧条时期迫使资本大幅减少（只有在违约概率= 100%时才承认违约），同时在繁荣时期允许过度放贷。相反的观点包括，对不良贷款的快速识别为采取纠正措施提供了直接的压力。例如，有证据表明，已发生损失法并没有在很大程度上导致金融危机的严重性。ECL方法旨在抑制顺周期性效应，因为及早确认信贷损失的能力应能减少损失的累积。在这种情况下，监管旨在提高金融稳定性。由于ECL更好地代表了贷款的经济价值，因此可以说它对银行财务报表的主要用户（例如投资者、监管机构）最有用。然而，模型实施的要素和预测的固有性质意味着ECL方法可能导致损失评估中存在某些缺陷，例如：对预期损失的时间和计量的管理自由裁量权：IFRS 9有一个减值的三阶段模型，这需要管理层根据不同工具的风险水平判断要识别的ECL长度（即12个月或终身）。预测方法：许多银行没有完全具备监管要求的“公正”预测能力。例如，个别预测者（如内部经济学家）已被证明有偏见，并没有充分利用公共信息（效率低下），而城市与非城市预测者在使用信息方面存在差异。有待探索的领域银行在确定ECL时使用的预测中不完善的性质（例如方向、大小、起源）。根据经验，这可以使用银行在年报中发布的预测和信用风险摘要。管理自由裁量权是否只存在于某些类型的风险中？例如，它是特定于贷款的行业，还是随着经理的特点（例如经验）而变化。那么，这些缺陷是如何影响银行信贷水平，进而影响经济的呢？金融稳定与银行贷款损失准备之间的联系已被研究过，但新的IFRS 9规定为这一主题提供了新的背景。可以实施DSGE模型来分析更广泛的经济影响。

MotivationWhilst working as an Economist for a professional services firm I supported a large UK bank in developing a macro-econometric forecasting model for the purposes of complying with the new IFRS 9 accounting rules. Reflecting on my MSc dissertation ('An Analysis of the Efficiency and Unbiasedness of UK Fixed-Event Macroeconomic Forecasts'), this project prompted questions on whether the potential biases or inefficiencies in the forecasts being used to adhere to this regulation could be having an impact on the wider financial system and economy.Background and relevant literatureThe financial crisis led to a regulatory change in the way that banks assess losses. The new IFRS 9 standard requires early recognition of losses using the expected credit loss approach, which is in contrast to the previous incurred loss approach. Effectively, IFRS 9 requires banks to calculate an unbiased probability-weighted estimate of expected credit loss (ECL) over a range of macroeconomic scenarios.This is a critical area for study because of the reliance placed on large banks by the economy, as played out by the 2008 financial crisis. How managers implement the regulation and the approach to producing forecasts will implicate both loss impairment and the level of available credit.There is debate on the impact of loan loss provisioning and the level of bank credit on the real economy:The (current) incurred loss approach is argued to contribute to procyclicality by forcing a sharp reduction in capital in the bust (default only recognised when probability of default = 100%), whilst enabling excessive lending in the boom.Converse arguments include the idea that fast recognition of non-performing loans provides immediate pressure for corrective action. E.g. there is evidence that the incurred loss approach did not contribute in a major way to the severity of the financial crisis.The ECL approach is intended to dampen procyclicality effects, as the ability to recognise credit losses earlier should reduce the build-up of losses. In which case, the regulation is intended to increase financial stability. As the ECL better represents the economic value of the loan, it can be argued to be of most use to main users of bank financial statements (e.g. investors, regulators).However, elements of the models implementation and the inherent nature of forecasting means that the ECL approach could result in certain imperfections being built into loss assessment, for instance:Managerial discretion over the timing and measurement of expected losses: IFRS 9 has a three-stage model for impairment, which requires managerial judgement on the length of ECL to recognise (i.e. 12 month or lifetime) for different instruments depending on their level of risk.Approach to forecasting: Many banks are not fully equipped to make 'unbiased' forecasts, as required by the regulation. For instance, individual forecasters (e.g. in-house economists) have been shown to have biases and do not fully use public information (inefficiency), whilst there are differences in the use of information by city vs. non-city forecasters.Areas for explorationThe nature of imperfections in the forecasts used by banks in determining ECL (e.g. direction, magnitude, origin). Empirically, this could use forecasts and credit risk summaries published by banks in annual reports.Does managerial discretion exist only for certain types of risks? E.g. is it specific to the industry of the loan, or does it vary with manager characteristics (e.g. experience).Then, how do these imperfections impact the level of bank credit, and subsequently the economy? The link between financial stability and bank loan loss provisioning has been studied, but the new IFRS 9 regulation provides a new context for this subject. A DSGE model could be implemented to analyse the wider economic implications.