Wednesday, February 3, 2010

How (Not) to Destroy American Jobs: Bad Analysis and Misleading Statistics

“The fundamental assumption behind [the Administration’s proposals to tax U.S. multinationals] is that U.S. multinationals expand abroad only to "export" jobs out of the country. Thus, taxing their foreign operations more would boost tax revenues here and create desperately needed U.S. jobs.
Academic research, including most recently by Harvard's Mihir Desai and Fritz Foley and University of Michigan's James Hines [here], has consistently found that expansion abroad by U.S. multinationals tends to support jobs based in the U.S. More investment and employment abroad is strongly associated with more investment and employment in American parent companies.”  Slaughter, WSJ 2010  
A group of international economists have been pushing two ideas for some time:  When multinational firms expand their overseas operations, jobs in the U.S. increase; when multinational firms increase their overseas investment, they also increase their investment in the United States.  Therefore, because these firms are also productive, taxing or inhibiting the growth of these firms harms the United States and slows domestic growth.  The seeds of the idea have merit: international trade should make the United States a wealthier country, at least when the driving force for the trade is not regulatory or tax avoidance.

Estimates from Desai et al. (2008) show that a 10 percent increase in foreign investment results in a 2.6 percent increase in domestic investment.  That is, for every dollar these multinational corporations spend abroad, they drop 25 cents in the United States.  They find similar results for sales, compensation, and employment.  They conclude, “These results do not support the popular notion that expansions abroad reduce a firm's domestic activity, instead suggesting the opposite.” 

First, and most important, their conclusion, as well as Slaughter’s in other work, is based on a flawed counterfactual.  Desai et al, in concluding the firms boost output in the United States, consider a counterfactual world in which the firms do not exist.  They do not consider a world in which the firms continue to exist but are prohibited, either through taxes or fiat, from expanding overseas.  The implicit assumption is that there growth would have been zero but for the expansion. 

I agree that their expansion would have been smaller had they not been able to take advantage of the cheaper, more lightly regulated, and lower tax foreign environment.  But do Desai, Foley, and Slaughter believe that Proctor & Gamble would not have grown at all between 1982 and 2004 if they had not been able to expand internationally.  So, to find direct evidence of the gains from outsourcing production, we need to compare the outcomes of the multinational firms with other U.S. firms.  I will do so in a moment.

Second, their statistical results rely on using the full sample.  Take a quick look at figure 1 in Desai et al (here  – scroll down to page 26).  The scatter plot shows foreign sales growth versus domestic sales growth over the sample period.  A positive relationship is easy to discern but is not particularly strong.  What the picture hides is the change in the relationship late (and early) in their sample.  Data post 1999 are very different than data before 1999, particularly the data between 1994 and 1999. 


The picture below uses aggregate data from the BEA (table 1 here).  The red bars show the annual average growth rate of employment in the foreign affiliates of U.S. multinationals (MNCs), the parents, and overall U.S. employment between 1982 and 2004.  Based on this aggregate data, the relationship between job creation at home and abroad is stronger than implied by the micro data, implying almost 80 jobs for every foreign job. 

However, job creation at multinationals in percentage terms pales in comparison to overall job creation in the United States.  Employment increased 1.8 percent per year on average between in 1982 and 2004 in the United States as a whole while growing a mere 0.6 percent at parent companies.  Indeed, the growth in employment of foreign affiliates is quite similar to the growth of overall U.S. employment. 

MCNs diminished net job creation in the United States.

But the picture is much worse for Slaughter’s argument if we simply examine the last 5 years of data.  Between 1999 and 2004, the growth rate of foreign affiliate employment was almost unchanged, near 1.8 percent, but the parents shed jobs.  Indeed, over this period, parents cut 2 jobs for every job created at a foreign affiliate.  Overall U.S. employment slowed over the period but remained positive.  Continuing the sample through 2007, does not change the picture.  U.S. and foreign affiliate employment remain positive while parent employment is still negative.  Although we do not have data beyond 2007, an increase in parent company employment during the recession is unimaginable, especially given the outsize decline in U.S. trade. 

What surprises most is that neither Desai et al. nor Slaughter acknowledge this point.  Nobody can work with the data as closely as these economists have and not notice such a basic fact.  In fact, the paper by Desai et al. never uses the words decline, slowing, reversal, or drop.  They have panel data and never include a time dummy.  They almost appear to be hiding the inconsistent fact in their data. 

Other Evidence

The two main other points brought up by Slaughter and Desai et al are the investments of MCNs and the value added (which leads to measured productivity).  The picture below shows gross value added for parents, foreign affiliates, and the nonfarm business sector in red and shows annual average growth rate of capital expenditures (gross investment for the non-farm business sector) in blue.

Parent’s growth is positive both in terms of investment and value added.  However, again the growth rates are well below that of both the foreign affiliates and the non-farm business sector. 

MCNs lowered investment and value added growth in the United States. 

By the way, the growth in value added combined with a decline in employment is what gives the parents their measured productivity edge. 


The expansion of international trade very likely has benefits for the United States.  But the expansion of trade has also had its costs.  The net benefit to the United State may be positive or negative (likely positive). 

With true facts before us, we can discuss the taxation of multinationals.  Taxing these companies will slow their growth, reduce their investment, and lower total U.S. employment.  That’s what taxes do; they reduce the profit incentives of firms.  But taxing U.S. companies and U.S. households has exactly the same effect.  If we make a public policy decision to have a large government, we must also make a public policy decision to tax heavily – either now or in the future. 

There is no data (which I consider credible) to support the idea that taxing multinationals is any worse than taxing any other company or for that matter than taxing the household sector. 

In any event, the overall employment, value added, or investment of these companies or any other is not the metric by which we decide which sectors to tax most heavily.  Optimal tax theory relies on relative elasticities.  The least elastic sector should face the highest tax burden, thereby minimizing the overall tax distortion. 

So, I don’t know whether we should tax the multinationals.  But there is zero evidence for taking them off the table before the discussion is even begun. 

No comments: