记者、经济学家和美国总统乔·拜登(Joe Biden)的竞选团队几个月来一直在困惑一个问题:当所有传统指标都是积极的时候，为什么美国人对经济如此不满?
根据最近的报告工作报告，就业增长稳定，失业率保持在低水平，工资也在上涨。国内生产总值是衡量经济健康状况的一个常用指标自2020年第二季度以来持续增长。甚至通货膨胀，在2022年达到了美国人几十年来最高水平，有终于被冷却了.
然而，消费者情绪悲观。2022年6月，密歇根大学消费者信心指数达到了自1966年开始跟踪以来的最低水平——甚至低于2008年大衰退期间的水平。无论你如何看待它，经济的表现和美国人的感受之间都有差距。
还是有?我们想了解在美国人对经济的看法中起作用的因素，以及为什么尽管基本面强劲，人们仍然对经济感觉不好。我们想知道也许不是那些人对经济状况的判断是错误的，但他们只是关心不同的问题在新冠肺炎疫情之后的剧变。
为了把这些拼凑起来，我们从网站下载了一系列与人们对经济的体验相关的经济指标圣路易斯美联储具体而言，与通货膨胀、利率、工作和工资、住房和基本家庭财务有关的指标。然后，我们使用这些指标拟合了一个线性回归模型，看看它们与1987年至2019年密歇根大学消费者情绪指数之间的关系(基本上,“时代之前").*
这个模型非常密切地跟踪了那个时期的消费者情绪指数。**作为趋势线的消费者情绪可能有点尖锐，因为重大新闻或经济冲击(如2008年房地产市场崩溃)可能会产生巨大影响，随着经济稳定，这些影响会很快消退。但除了这些峰值，我们的模型在追踪疫情之前32年的趋势方面相当一致。
然而，在疫情期间和之后，将同样的前疫情模型应用于消费者情绪是行不通的。在2020年之前，与人们对经济的感受相关的指标似乎不再同样重要。正如美国生活的许多领域一样，疫情几乎改变了人们对经济和他们关心的问题的所有看法。
在疫情之前，在我们的模型中，许多变量是消费者情绪的重要统计指标;特别是，最显著的变量似乎是汽车销售、汽油价格、中等家庭收入、联邦基金有效利率、个人储蓄和家庭支出(不包括食品和能源)。其中一些指标在过去四年中发生了巨大变化，这可能导致积极的经济消息与美国人持续的经济悲观主义之间的脱节。***
个人储蓄率
在疫情之前，我们的模型显示，当个人储蓄率(衡量家庭每月能够存多少钱而不是花多少钱)较高时，消费者对经济感觉更好。这是有道理的:当人们在银行有存款，并且能够为汽车和房子等重要的购买而储蓄时，他们会感觉更好。
在疫情期间个人储蓄率飙升。2020年4月，这一指标几乎是1975年5月创下的历史新高的两倍。当然，消费者能够在2020年存钱有很多原因;在企业关门、家庭被鼓励呆在家里的时候，人们无法像往常一样花钱。此外，联邦政府对疫情的反应给全国的家庭送去了很多额外的钱。2020年3月的《关怀法案》(CARES Act)，2020年12月的《冠状病毒应对和救济法案》(Coronavirus Response and Relief Act)，以及2021年3月的《美国救援计划法案》(new York Act)，都包括了对家庭的直接支付，以及将钱放进人们口袋的其他东西，如扩大的儿童税收抵免和增加的失业救济金。
所有这些加在一起意味着美国人现金充裕，但无处可花。因此，尽管储蓄率大幅上升，但消费者对经济仍然没有积极的感觉——这与我们在疫情之前几十年看到的这两个变量之间的关系相反。
快进到2024年，个人储蓄率已经下降到有史以来的最低水平之一(唯一一次储蓄率较低是在大衰退期间)。这可能部分是由于高通货膨胀率和重要的家庭购买成本增加，如汽车和房屋。即便如此，现在判断之前个人储蓄和消费者情绪之间的相关性是否会恢复还为时过早;虽然这两个指标目前都相对较低，但没有足够的数据来确定相关性是否回到了疫情之前的水平。
通货膨胀
另一大难题是通货膨胀。在疫情之前，在我们的模型中，通货膨胀率的变化在统计上并不显著(尽管我们确实看到了消费者情绪和个人消费支出之间的负相关关系)。****但在疫情期间和之后，美国人在很短的时间内经历了几十年来最高的通货膨胀率。这些突如其来的价格上涨自然震惊了许多人，他们在成年后一直幸福地享受着缓慢而稳定的价格增长。事实也的确如此震惊过了一会儿才消失尽管通货膨胀率已经接近典型水平。
“今天有很多人真的没有任何经历过任何形式的实质性通货膨胀，所以现在他们第一次看到它，”前白宫工作人员经济学家、斯坦福大学访问学生研究员瑞安·卡明斯说。"人们习惯于看到几乎为零的通货膨胀。"
卡明斯和他的研究顾问、前白宫同事尼尔·马奥尼使用回归模型为了证明在急剧的通货膨胀后消费者信心的缓慢恢复是我们以前见过的——而且是可以测量的。他们发现，如果通胀飙升至12%，或比美联储2%的目标高出10个百分点，消费者信心指数将下降约35个点。这种影响持续存在，衰减率约为每年50%。但由于美国人目前不仅受到今年3.2%的同比通胀率的影响，还受到去年7.8%的通胀率和2022年6.2%的通胀率的影响，这将对消费者情绪产生累积效应。
也就是说，卡明斯指出这不是一个完美的解释;积累只能解释消费者情绪和经济指标之间的部分差距，他们的模型只在两个真实世界事件(20世纪80年代初的通胀和大流行后的冲击)中得到训练。但这些数字符合我们对消费者如何处理突然增加的杂货店账单的直觉，也符合我们模型的发现。简而言之:即使通货膨胀正在好转，美国人也不会因为通货膨胀一开始就很糟糕而被责骂。
房屋
住房成本，无论是购买还是租赁，都是大多数家庭预算中最大的项目之一。但令人惊讶的是，我们的前疫情模型没有发现房价和消费者情绪之间的显著关系。*****我们还研究了房价和消费者情绪之间的直接关系，也没有发现有意义的关系。然而，在我们的大流行后数据中，当我们检查消费者情绪与我们考虑的每个指标的相关性时，消费者情绪和中值房价的相关性在所有******(负相关性，意味着较高的价格与较低的消费者情绪相关)中最强。
消费者的房子通常是他们最有价值的资产之一，因此随着时间的推移，房价的稳步上涨可以导致家庭财富的增加。但在疫情期间，低利率、高储蓄率和工作模式的变化——即许多工人新发现的在家工作的能力——帮助了购房市场过热，买家一头栽进了持续的供应短缺。根本没有足够的房子可以买，这就推高了待售房屋的价格。
标价并不是唯一让人们难以买得起房子的原因。通胀开始飙升后，美联储委员会提高了利率，并一直保持在高位。如今，30年期固定利率抵押贷款的平均利率约为7%。这意味着中低收入家庭很难获得抵押贷款资格，因为利率会让他们的月供特别高。美国全国房地产经纪人协会(National Association of Realtors)首席经济学家劳伦斯·云(Lawrence Yun)表示:“试图以当前的抵押贷款利率购买一套中等价位的住房，现在的月抵押贷款额是冠状病毒肺炎之前的两倍。”。
即使一个家庭已经攒够了首付，这也是事实，在房租居高不下的情况下，这已经是一项艰巨的任务。越来越少的人能够支付目前的住房成本，同时存够首付。低收入家庭仍然最有可能承受高房租，但他们不再是唯一受影响的人。高房租也开始影响中等收入人群。“年收入在3万美元到4.5万美元之间的家庭，成本负担率的上升速度比低收入或高收入家庭更快，”哈佛大学联合住房研究中心的高级研究员丹尼尔·麦库埃说。“我们已经看到越来越多的中等收入家庭感受到了负担。”
简而言之，在疫情之前就已经存在住房负担危机了。现在情况更糟了，将越来越多的高收入人群挡在了购房市场之外。穆迪分析公司(Moody's Analytics)助理董事兼经济学家马修·沃尔什(Matthew Walsh)表示:“美国市场严重的住房短缺确实对整体住房市场产生了持久的影响。”。他说，这种短缺的一个重要原因是，在本世纪早些时候的大衰退和房地产危机之后，很少建造新房或低价房，这使得首次购房者的情况更加糟糕。
正在租房但又想买的人被套牢了。住在简易房并想搬到更大的房子的人被困住了。这种情况挫败了美国梦的一个基本要素。“大多数美国人认为他们的生命周期是:他们上大学……找到一份好工作，然后在工作几年后，试图买房子，”云说。“这种改善生活的正常序列现在已经被切断，因为即使一个人有一份好工作，即使一个人有良好的信用评分，也很难成为业主。所以我认为，这一冲击因素正在极大地影响人们对整体经济的情绪。”
汽车销售
我们想调查的另一个因素是汽车销售。在我们之前的疫情模型中，汽车总销量与消费者情绪之间存在着强烈的正相关关系:如果人们在购买汽车，你可以相当合理地打赌，他们对经济感觉良好。这感觉很直观——如果认为经济陷入困境，谁还会买车?但在疫情期间和之后，这个行业似乎发生了很多变化。
疫情在汽车市场造成的混乱已经被很好的证明了——在供应链问题、工厂停工、现金过剩的家庭，以及整个人口突然希望能够在不与数百名其他乘客共享空间的情况下旅行的情况下，疫情真的撼动了一切，汽车行业现在才开始恢复正常。
对于消费者和他们的汽车之间关系的转变，一个可能的解释是我们已经提到的:通货膨胀。国际汽车营销和研究集团考克斯汽车公司(Cox Automotive)的企业传播总监马克·斯奇默(Mark Schirmer)表示，我们的模型在疫情之前发现的汽车销售和消费者情绪之间的正相关关系在汽车行业得到了很好的证明。考克斯汽车公司拥有AutoTrader和凯利的蓝皮书。
“消费者情绪对购车非常重要，因为如果你对自己的财务状况和整体经济不乐观，你就不太可能购买大件商品。记住，汽车通常是你所购买的第二贵的东西，超过你的房子，”斯戈默说。
但自2020年以来，汽车总销量和消费者情绪之间就没有统计关系了——这可能是大流行后世界经济的复杂性和变化的结果(这将再次使模型难以从噪声中解析信号)。斯戈默说，这可能与汽车购买者面临的高利率和高价格有关。大多数汽车购买者会为购买提供资金平均新汽车贷款利息利率目前为9.7%，而平均二手汽车贷款利率为14.1%，略低于2月份达到的历史最高水平。
Cox Automotive还通过计算平均购买新车所需的中值收入的估计周数来跟踪车辆的可负担性，虽然该数字在过去两年中有所改善，但与疫情之前的水平相比仍然较高。四月，有数据的最近一个月，购买一辆汽车需要37.7周的中等收入，而2019年底不到35周。
“就在疫情之前，一辆新车的典型平均交易价格约为38，000美元。到2023年，这个数字是4.8万美元，”席尔默说。他指出，这些都可能导致汽车销售和情绪之间的关系破裂。基本上，人们可能会购买汽车，但他们不一定对此感到高兴。
* * *
受我们1987年至2019年经济指标和情绪模型的启发，我们试图在2021年至2024年的相同数据上训练一个类似的线性回归模型，以更直接地比较疫情之后的情况变化。虽然我们能够很好地拟合这个大流行后的模型，*******有趣的事情发生了:没有一个变量显示为消费者情绪的统计显著预测指标。
这表明幕后有更复杂的事情在发生:这些变量之间的相互作用可能推动了预测，而在这个大流行后的小数据集中有太多的噪音，模型无法理清它。这是有道理的——经济是一个复杂的有机体，事物以复杂的方式相互联系，疫情是对现状的百年一遇的动摇。与以前相比，这是一个显著的转变，当时我们对哪些变量与消费者的感受有关以及以何种方式有关有一个相对清晰的认识。
我们已经讨论过的购买类型的变化——房屋、汽车和杂货等日常用品——从根本上改变了美国人对他们的生活负担能力的看法以及他们衡量生活质量的方式。尽管实际工资赶上了通胀，就业市场和股市依然强劲，但消费者仍然面临着真正的挑战，这些挑战不仅仅局限于一两个指标。尽管一些指标可能正在改善，但美国人只是以不同于以往的方式衡量这些因素，这让人们有足够的理由对经济感到沮丧。
脚注
*我们在这里使用的模型不同于经济学家用来理解和预测消费者情绪的模型类型。我们使用该模型的目的是探索和理解与家庭财务相关的某些因素，以及它们在过去如何与消费者情绪相关联，而不是试图预测将来的情绪或对经济做出预测。我们限制了变量集的范围，只包括与家庭财务相关的变量，这样我们就可以分析这些指标与整体消费者情绪之间的关系。此外，我们使用了线性回归，而不是更复杂的建模技术，因此我们可以专注于这些变量和整体情绪之间的简单关系。我们还从回归中排除了任何高度相关的指标。
* *对于那些书呆子来说，调整后的R平方约为0.7。
* * *因为此分析是探索性的，我们保持模型相对简单以避免过度拟合，我们不想过多解读我们模型中观察到的关系的强度或功效，所以我们选择仅关注建模系数的方向。换句话说，如果所有其他指标保持不变，该指标与消费者情绪是正相关还是负相关?我们承认，各种指标之间的关系以及与消费者情绪的关系比我们线性回归中观察到的要复杂得多。
* * * *我们没有在模型中包括消费者价格指数，因为它与我们感兴趣的其他一些指标高度相关。(有关更多详细信息，请参见方法论。)然而，我们确实纳入了一个衡量上个月CPI变化幅度的指标，并未发现该指标与整体消费者情绪之间存在显著的统计关系。
* * * * *在我们的模型中，房价和疫情之前的消费者情绪之间缺乏统计意义上的显著关系，这让我们想知道是否有更复杂的关系掩盖了任何更直接的关系。所以我们尝试在我们的模型中包含一个相互作用(中值房价和新房开工率，或每月新建房屋数量之间的相互作用);基本上，我们检查了这两个变量是否以一种允许我们观察与消费者情绪的关系的方式“一起工作”。在这种情况下，我们看到了消费者情绪与中值房价和新屋开工率之间的负相关关系。就是这样是在疫情之前，房价和消费者情绪之间可能存在某种关系，但这很复杂。同样，因为这个分析是探索性的，我们不想在交互的细节上投入太多。
* * * * * *带皮尔逊相关系数的-0.77。
******调整后的R平方为0.7，与我们在疫情之前的模型相似。
方法学
为了探索消费者情绪和家庭财务基本指标之间的关系，我们分析了密歇根大学消费者情绪指数和21个可能的指标。我们从20个指标开始，这些指标选自圣路易斯美联储与五个主要类别相关:通货膨胀、利率、工作和工资、家庭财务和住房。因为消费者情绪指数是一个月度指标，对于不太频繁的指标，如季度或年度发布的指标，我们使用每个月的最新观测值。对于比每月更频繁的指标，我们取每个月观察到的所有值的平均值。我们还包括了第21个指标——每月一次标普500收盘价，可从以下渠道获得雅虎财经。然后将这些指标全部归一化，使平均值为0，标准偏差为1。
然后，我们计算了1987年至2019年所有21个指标之间的皮尔逊相关系数，以确保进入模型的指标不包括任何彼此高度相关的指标。我们删除了彼此之间皮尔逊相关系数超过0.85的任何指标(无论其相关方向如何)，保留了每组高度相关指标中的一个指标，同时还确保五个主要类别中的每一个类别都至少有一个指标。模型中包含的最后一组指标是失业率,联邦基金有效利率,个人储蓄率,实际中等家庭收入,新的住房单元开始建造，变化(月-月差异)消费者价格指数,原油价格,个人消费支出(不包括食品和能源)，实际周收入中位数,房屋销售价格中位数和汽车销售总额.
一旦我们选择了最终的一组指标，我们就对1987年至2019年的50%数据的随机子集进行了线性回归，并将消费者情绪指数作为因变量。为了评估疫情之前训练的回归是否能够使用大流行后的数据预测消费者情绪，我们使用疫情之前的模型根据2020年至2024年的数据预测消费者情绪。我们使用1000个重复样本的bootstrapped 95置信区间来评估统计显著性。
为了进一步比较消费者情绪从疫情大流行前到大流行后的变化，我们重新拟合了疫情大流行前的模型，但根据大流行后的数据对其进行了训练(即，使用相同的指标)，由于2020年经济的高度波动性，省略了2020年的数据。此外，我们检查了两个时间段内每个指标和消费者情绪之间的皮尔逊相关系数。
It's not just vibes. Americans' perception of the economy has completely changed.
There's a question journalists, economists and President Joe Biden's campaign have been puzzling over for months: Why, when every traditional indicator is positive, are Americans so displeased with the economy?
According to the most recentjobs report, job growth is steady, unemployment remains low and wages are rising. The gross domestic product, a common measure of the health of the economy, has beengrowing since the second quarter of 2020. Even inflation, which in 2022 reached thehighest levels Americans had seen in decades, hasfinally been cooling.
And yet,consumer sentiment is pessimistic. In June 2022, the University of Michigan Consumer Sentiment Index reached the lowest level ever recorded since tracking began in 1966 — lower even than during the Great Recession of 2008. No matter how you slice it, there's a gap between how the economy is doing and how Americans feel about it.
Or is there? We wanted to understand the factors at play in Americans' perception of the economy and why, despite strong fundamentals, people still aren't feeling good about it. And we wondered if maybe it's not that peopleare wrong about how the economy is doing, but that they simply care about different issues in the wake of the upheaval following the COVID-19 pandemic.
To piece this together, we downloaded a collection of economic indicators related to people's experiences of the economy from theSt. Louis Federal Reserve— specifically, indicators relating to inflation, interest rates, work and wages, housing, and basic household finances. We then fitted a linear regression model using those indicators to look at how they were related to the University of Michigan Consumer Sentiment Index from 1987 to 2019 (basically, the "Before Times").*
This model tracks the Consumer Sentiment Index over that time period pretty closely.** Consumer sentiment as a trendline can be a bit spiky, as big news or economic shocks (such as the collapse of the housing market in 2008) can have dramatic effects that fade fairly quickly as the economy stabilizes. But other than these spikes, our model is fairly consistent at tracking the trends over the 32-year period preceding the pandemic.
Applying the same pre-pandemic model to consumer sentiment during and after the pandemic, however, simply does not work. The indicators that correlated with people's feelings about the economy before 2020 no longer seem to matter in the same way. As with so many areas of American life, the pandemic has changed virtually everything about how people think about the economy and the issues that concern them.
Before the pandemic, a number of variables were statistically significant indicators for consumer sentiment in our model; in particular, the most salient variables appear to be vehicle sales, gas prices, median household income, the federal funds effective rate, personal savings and household expenditures (excluding food and energy). And some of these indicators have seen dramatic changes over the past four years, likely contributing to the disconnect between the positive economic news and Americans' continued economic pessimism.***
Personal savings rate
Prior to the pandemic, our model shows consumers felt better about the economy when the personal savings rate, a measure of how much money households are able to save rather than spend each month, was higher. This makes sense: People feel better when they have money in the bank and are able to save for important purchases like cars and houses.
During the pandemic, thepersonal savings rate soared. In April 2020, the metric was nearly double its previous high, recorded in May 1975. There are many reasons consumers were able to save in 2020, of course; people weren't able to spend their money the way they normally would have while businesses were closed and families were encouraged to stay home. Moreover, the pandemic response from the federal government sent a lot of extra money to families across the country. The CARES Act in March 2020, the Coronavirus Response and Relief Act in December 2020 and the American Rescue Plan Act in March 2021 all included direct payments to households, as well as other things that put money in people's pockets, like an expanded child tax credit and increased unemployment payments.
All this taken together meant Americans were flush with cash but had nowhere to spend it. So despite the fact that the savings rate went way up, consumers still weren't feeling positively about the economy — contrary to the relationship between these two variables we saw in the decades before the pandemic.
Fast forward to 2024, and the personal savings rate has dropped to one of its lowest levels ever (the only time the savings rate was lower was in the years surrounding the Great Recession). This may be due in part to high inflation and increased costs for important household purchases, like cars and homes. Even so, it's too early to tell if the previous correlation between personal savings and consumer sentiment will resume; while both measures are currently relatively low, there isn't enough data to determine if the correlation is back to its pre-pandemic levels.
Inflation
Another big piece of the puzzle is inflation. Before the pandemic, changes in the inflation rate weren't statistically significant in our model (though we do see a negative relationship between consumer sentiment and personal consumption expenditures).**** But during and after the pandemic, Americans saw some of the highest rates of inflation the country has had in decades, and in a very short period of time. These sudden spikes naturally shocked many people who had been blissfully enjoying slow, steady price growth their entire adult lives. And it hastaken a while for that shock to wear off, even as inflation has crept closer to typical levels.
"You have a lot of people today that really don't have any experience with any sort of substantial inflation, and so now they're seeing it for the first time," said Ryan Cummings, a former staff economist at the White House and visiting student researcher at Stanford University. "People were used to seeing almost zero inflation."
Cummings and his research advisor and former White House colleague Neale Mahoneyused a regression modelto show that the slow recovery of consumer sentiment after dramatic inflation spikes was something we'd seen before — and it was measurable. They found that if inflation spikes to 12 percent — or 10 percentage points above the Federal Reserve's 2 percent target — the Consumer Sentiment Index drops by about 35 points. And that effect lingers, with a decay rate of about 50 percent per year. But since Americans are currently reeling from the aftershocks of not only this year's 3.2 percent year-over-year inflation but also last year's 7.8 percent inflation and 2022's 6.2 percent inflation, it creates a cumulative effect on consumer sentiment.
That said, Cummings noted this isn't a perfect explanation; accumulation accounts for only part of the gap between consumer sentiment and economic indicators, and their model was trained on only two real-world events (the inflation of the early 1980s and the post-pandemic shock). But the numbers align with our intuitive sense of how consumers process suddenly having their grocery store bill jump, as well as the findings from our model. In simple terms: Even if inflation is getting better, Americans aren't done being ticked off that it was bad to begin with.
Housing
The cost of housing, whether buying or renting, is one of the biggest items on most household budgets. But surprisingly, our pre-pandemic model didn't find a notable relationship between housing prices and consumer sentiment.***** We also looked at the direct correlation between housing prices and consumer sentiment and didn't see a meaningful relationship there either. However, in our post-pandemic data, when we examined how correlated consumer sentiment was with each indicator we considered, consumer sentiment and median housing prices had the strongest correlation of all****** (a negative one, meaning higher prices were associated with lower consumer sentiment).
Consumers' homes are often one of their most valuable assets, so a steady rise in housing prices can lead to an increase in a household's wealth over time. But during the pandemic, low interest rates, high savings rates and changes in working patterns — namely, many workers' newfound ability to work from home — helped overheat the homebuying market, and buyers ran headlong into an enduring supply shortage. There simply weren't enough houses to buy, which drove up the costs of the ones that were for sale.
The sticker price isn't the only thing making houses difficult to afford. After inflation began to soar, the Federal Reserve Board raised interest rates and has kept them high. Average mortgage rates for a 30-year fixed-rate mortgage are around 7 percent today. That means that lower- and middle-income families struggle to qualify for mortgages because the interest rates would make their monthly payments especially high. "Trying to buy a median-price home at the prevailing mortgage rates, it is twice as high a monthly mortgage payment now compared to pre-COVID," said Lawrence Yun, chief economist at the National Association of Realtors.
That's true even if a family has been able to save enough for a down payment, already a difficult task when rents remain high as well. Fewer people are able to cover their current housing costs while saving enough to make a down payment. Low-income households are still the most likely to be burdened with high rents, but they're not the only ones affected anymore. High rents have also begun to affect those at middle-income levels as well. "Households earning $30,000 to $45,000 per year, the rates of cost burden have been rising faster than for lower-income or higher-income households," says Daniel McCue, a senior research associate at the Harvard Joint Center for Housing Studies. "We've seen more and more moderate- and middle-income households feeling the burden."
In short, there was already a housing affordability crisis before the pandemic. Now it's worse, locking a wider array of people, at higher and higher income levels, out of the home-buying market. "The critical shortage of housing in the U.S. market is really having lasting effects on housing overall," said Matthew Walsh, assistant director and economist at Moody's Analytics. A big component of that shortage is also the fact that, after the Great Recession and housing crisis earlier this century, very few starter homes or low-cost units were built, making things even worse for first-time homebuyers, he said.
People who are renting but want to buy are stuck. People who live in starter homes and want to move to bigger homes are stuck. The conditions have frustrated a fundamental element of the American dream. "Most Americans view their life cycle as: They go to college … find a good job, then after having a job for several years, trying to buy a house," said Yun. "That kind of normal sequence of improving their life has been now cut off, because even if one has a good job, even if one has a good credit score, it's very difficult to become homeowners. So I think this shock factor is greatly influencing people's sentiment about the broader economy."
Vehicle sales
One other factor we wanted to investigate was car sales. In our pre-pandemic model, total vehicle sales had a strong positive relationship with consumer sentiment: If people were buying cars, you could pretty reasonably bet that they felt good about the economy. This feels intuitive — who buys a car if they think the economy is in the toilet? But during and after the pandemic, a lot seemed to change in the industry.
The tumult that the pandemic created in the auto markethas been well documented— between supply chain issues, factory shutdowns, households with excess cash and an entire population's sudden desire to be able to travel without sharing air space with hundreds of other passengers, the pandemic really shook things up, and the auto industry is only now starting to revert back to something like normal.
One possible explanation for a shifting relationship between consumers and their cars may be something we've already covered: inflation. The positive relationship between car sales and consumer sentiment that our model found before the pandemic has been well documented in the automotive industry, according to Mark Schirmer, the director of corporate communications for Cox Automotive, the international automotive marketing and research conglomerate that owns AutoTrader and Kelley's Blue Book.
"Consumer sentiment is really important for car buying in that, if you don't feel positive about your finances, the economy overall, you're not likely to be making big purchases. Remember, an automobile typically is like the second most expensive thing you'll ever buy, beyond your house," Schirmer said.
But since 2020, there hasn't been a statistical relationship between total vehicle sales and consumer sentiment — probably a result of the complexities and shifts in the economy in the post-pandemic world (which would again make it difficult for the model to parse signal from noise). Schirmer said it may have to do with the high interest rates and prices car buyers are facing. The majority of car buyers finance that purchase, andthe average new auto loan interestrate is currently 9.7 percent, while the average used auto loan rate is 14.1 percent, just below its all-time high, achieved in February.
Cox Automotive also tracks vehicle affordability by calculating the estimated number of weeks' worth of median income needed to purchase the average new vehicle, and while that number has improved over the last two years, it remains high compared to pre-pandemic levels. In April, the most recent month with data,it took 37.7 weeks of median income to purchase a car, compared with fewer than 35 weeks at the end of 2019.
"Right before the pandemic, the typical average transaction price was around $38,000 for a new car. By 2023, it was $48,000," Schirmer said. This could all be contributing to the break in the relationship between car sales and sentiment, he noted. Basically, people might be buying cars, but they aren't necessarily happy about it.
* * *
Inspired by our model of economic indicators and sentiment from 1987 to 2019, we tried to train a similar linear regression model on the same data from 2021 to 2024 to more directly compare how things changed after the pandemic. While we were able to get a pretty good fit for this post-pandemic model,******* something interesting happened: Not a single variable showed up as a statistically significant predictor of consumer sentiment.
This suggests there's something much more complicated going on behind the scenes: Interactions between these variables are probably driving the prediction, and there's too much noise in this small post-pandemic data set for the model to disentangle it. That makes sense — the economy is a complicated organism, things are connected to each other in complex ways and the pandemic was a once-in-a-century shakeup to the status quo. It's a notable shift from the Before Times, when we had a relatively clear sense of which variables were related to consumers' feelings and in what kinds of ways.
Changes in the kinds of purchases we've discussed — homes, cars and everyday items like groceries — have fundamentally shifted the way Americans view how affordable their lives are and how they measure their quality of life. And even as real wages catch up to inflation and the job market and stock market remain strong, there are still genuine challenges facing consumers that aren't isolated to just one or two indicators. Even though some indicators may be improving, Americans are simply weighing the factors differently than they used to, and that gives folks more than enough reason to have the economic blues.
Footnotes
*The model we're using here is different from the types of models economists use to understand and predict consumer sentiment. Our goal with this model is to explore and understand certain factors related to household finances and how they've correlated to consumer sentiment in the past, rather than to try to anticipatefuturesentiment or make predictions about the economy. We've limited the scope of our variable set to include only those related to household finances so that we can analyze changes in how those measures relate to overall consumer sentiment. Moreover, we used a linear regression, rather than a more sophisticated modeling technique, so that we can focus on simple relationships between these variables and overall sentiment. We also excluded any highly correlated indicators from our regression.
**For the wonks out there, the adjusted R-squared is about 0.7.
***Because this analysis is exploratory and we kept the model relatively simple to avoid overfitting, we didn't want to read too much into the strength or power of the observed relationships in our model, so we've opted to focus only on the direction of the modeled coefficients. In other words, does the indicator have a positive or a negative relationship with consumer sentiment, if all other indicators were held equal? We acknowledge that the relationship among the various indicators and with consumer sentiment are much more complex than observed in our linear regression.
****We didn't include the Consumer Price Index in our model because it was highly correlated with a few other indicators that we were interested in. (See the methodology for more details.) We did, however, include a measure of how much the CPI had changed in the last month and found no statistically significant relationship between that metric and overall consumer sentiment.
*****The lack of a statistically significant relationship between housing prices and pre-pandemic consumer sentiment in our model led us to wonder whether there might be a more complex relationship obscuring any more straightforward relationships. So we tried including an interaction (between median housing prices and new housing starts, or the number of new homes being built each month) in our model; basically, we checked whether the two variables were "working together" in a way that allowed us to observe a relationship with consumer sentiment. In this case, we saw a negative relationship between consumer sentiment and the interaction between median home prices and new housing starts. So therewaslikely a relationship between housing prices and consumer sentiment prior to the pandemic, but it was complicated. And again, because this analysis is exploratory, we don't want to put too much stock into the specifics of the interaction.
******With aPearson correlation coefficientof -0.77.
*******The adjusted R-squared was 0.7, which was similar to our pre-pandemic model.
Methodology
To explore the relationship between consumer sentiment and underlying measures of household finances, we analyzed the relationship between theUniversity of Michigan's Consumer Sentiment Indexand 21 possible indicators. We began with a set of 20 indicators chosen from those available from theSt. Louis Federal Reserverelated to five main categories: inflation, interest rates, work and wages, household finances, and housing. Because the Consumer Sentiment Index is a monthly indicator, for less frequent indicators, such as those that publish quarterly or annually, we used the most recent available observation for each month. For indicators that are more frequent than monthly, we took the average of all values observed in each month. We also included a 21st indicator — monthly S&P 500 closing prices, obtained fromYahoo Finance. These indicators were then all normalized to have a mean of 0 and standard deviation of 1.
We then calculated the Pearson correlation coefficients among all 21 indicators from 1987 to 2019 to ensure that the indicators going into the model did not include any that were highly correlated with each other. We removed any indicators that had a Pearson correlation coefficient of over 0.85 with each other (regardless of the direction of their correlation), keeping one measure from each group of highly correlated indicators while also ensuring we had at least one indicator from each of the five main categories. The final set of indicators included in the model were theunemployment rate,federal funds effective rate,personal savings rate,real median household income,new housing units started, change (month-over-month difference) inConsumer Price Index,crude oil prices,personal consumption expenditures(excluding food and energy),real median weekly earnings,median home sales pricesandtotal vehicle sales.
Once we had selected the final set of indicators, we trained a linear regression on a random subset of 50 percent of data from 1987 to 2019, with the Consumer Sentiment Index as the dependent variable. To assess whether the pre-pandemic-trained regression was able to predict consumer sentiment using post-pandemic data, we used the pre-pandemic model to predict consumer sentiment based on data from 2020 to 2024. We evaluated statistical significance using bootstrapped 95 percent confidence intervals with 1,000 repeated samples.
To further compare how consumer sentiment has changed from the pre-pandemic to post-pandemic data, we re-fitted the pre-pandemic model but trained it on post-pandemic data (i.e., used the same indicators), omitting data from 2020 due to the high level of volatility in the economy that year. In addition, we examined Pearson correlation coefficients between each indicator and consumer sentiment for both time periods.
