Estimating the Effects of Minimum Wage

Estimating the Effects of Minimum Wage on Employment and Inequality: Evidence from Taiwan Lu, Chyi-Horng Economics, NTU 2018.6.14 Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 1 / 63

1. Abstract 2. Literature and Contribution 3. Backgrounds and Data Source 4. Methods and Data Processing 5. Estimated Results 6. Transition 7. Firm Adjustments 8. Primary Conclusion and Future Work Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 2 / 63

1. Abstract Abstract Methods Using monthly administrative data from Taiwanese mandatory employment insurance. Combining RDD and DD to estimate the effect of minimum wage Results Why? increase in July 2007. Positive wage effect but no employment effect to full-time workers. Decreasing the wage gap but no evidence to prove wage spillover effect. Influence was moderate. Firms bore the cost and the higher occupational injury rate represents that firms may cut down other expenditure or raise labor productivity. Turnover rate was decreased, so the shock could be mitigated. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 3 / 63

2. Literature and Contribution Literature The Employment Effects Would be Small The reasons may be monopsony, shock effect, two-sector change, heterogeneous workers and lagged adjustment (Brown, Gilroy, and Kohen 1982). Effects Close to 0 or Really Small Doucouliagos and Stanley (2009) did meta-analysis by using 64 US empirical papers (39 of them focus on teenagers and others focus on specific groups of people). Existing publication selection bias for adverse employment effects. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 4 / 63

2. Literature and Contribution Literature Figure: Meta-Analysis of Estimated Employment Effects of Minimum Wage in US Source: Doucouliagos and Stanley (2009) Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 5 / 63

2. Literature and Contribution Literature Still Controversy The Approach: Using approach of difference-in-difference with controlling the state-specific trends leads to no effect (Arindrajit, Dube, Reich, and Zipperer 2017). But it may exclude the real effect (Neumark and Wascher 2017). The Target: Why Teenagers? The importance to study the effect on teenagers is decreasing (Manning 2016). Beyond The Employment Effect Reduceing the wage gap with possibly fewer spillover effects than the findings of Lee (1999) (Autor, Manning and Smith 2016) A conclusive answer would require better wage data, ideally administrative payroll data Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 6 / 63

2. Literature and Contribution Contribution Using administrative data of insured wage Estimating the effects on people who were actually affected by the minimum wage increase. Observing the wage spillover effects. Generating a matched employer-employee dataset to find the firm s adjustments. Combining the approaches of regression discontinuity design and difference-in-difference Avoiding endogenous problems and the controversy question raised by using DD. To my knowledge, it s the first study that used the RD approach to estimate the minimum wage effect. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 7 / 63

3. Backgrounds and Data Source Minimum Wage Increase in Taiwan Figure: The Minimum Wage Increase in Taiwan Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 8 / 63

3. Backgrounds and Data Source Minimum Wage Increase in Taiwan Minimum wage increase in July 2007 The hourly minimum wage was increased from NT. 66 to 95, which is about 44%. The monthly minimum wage was increased from NT. 15,840 to 17,280, which is about 9%. The previous time of change is in 1997 and the next time is in 2011. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 9 / 63

3. Backgrounds and Data Source Data Source Employment Insurance It s operated by government. It is mandatory for employers to have their employees joined into the insurance. Monthly Insured wage set by levels Before July 2007: NT. (11,100, 12,300, 13,500), 15,840, 16,500, 17,400, 18,300, 19,200... After July 2007: NT. (11,100, 12,105, 12,300, 13,500), 17,280, 17,400, 18,300, 19,200... Insured wage is close to and not less to the actual wage. For example, if your actual wage is 17,000, then your insured wage should be 17,400. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 10 / 63

3. Backgrounds and Data Source Data Source Identify Full-Time Workers Were 15,840 or 16,500: Affected Were 17,400: Not Sure Were 18,300 or higher: Not Affected Difficulty for Identifying Part-Time Workers Can know who is part-time worker. But, no information of working hours or hourly wages. Risk Insured wage may be lower than the actual wage. Still have to pay more insurance fee and pension contribution for the full-time workers, which is about 15% of the increase amount ((17,280-15,840)*15%=216). Confirming by using the data from Survey of Family Income and Expenditure Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 11 / 63

3. Backgrounds and Data Source Data Source Monthly Insured File from 2004 to 2010: Variables include individual s insured wage, full-time or part-time, birth year, sex and the identification of their institution. However, there is no records of working hours or hourly wages. Institutional File: Variables include institution s category, industry, and area. Only keeping industrial workers, workers of commercial firms and shops, or employees in journalistic, cultural, nonprofit organizations or cooperative enterprises. Excluding employees in government agencies and schools, workers employed in fishing production, and the category of others. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 12 / 63

3. Backgrounds and Data Source Proportion of Workers Affected by Monthly Minimum Wage Increase (a) Male (b) Female Figure: Prop. of Full Time Workers Affected by Next Minimum Wage Increase Note: Prop. of the affected = Prop. of workers who have wages below the next minimum wage Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 13 / 63

3. Backgrounds and Data Source Proportion of Workers Affected by Hourly Minimum Wage Increase (a) Male (b) Female Figure: Prop. of Part Time Workers to Total Workers Note: Since we don t have information of working hours nor hourly wages, every part-time worker could be affected by the hourly minimum wage increase. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 14 / 63

3. Backgrounds and Data Source Distribution Around July 2007 Full-Time Workers (a) Male (b) Female Figure: Number of Full-Time Workers with Insured Wages not above 17,400 Note: Since the policy change would only raise lower wages to 17,280, so the number of employees with insured wage equal to and below 17,400 should have no change if the policy has no effect. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 15 / 63

3. Backgrounds and Data Source Distribution Around July 2007 Part-Time Workers (a) Male (b) Female Figure: Number of Part Time Workers with Insured Wages not above 17,400 Note: Due to the working hours regulation, which was 168 hours for 4 weeks, it s rare that part-time workers who earned hourly minimum wage can have a monthly wage higher than 15,960 which is less than 17,400. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 16 / 63

4. Methods and Data Processing Research Methods Question: Did the minimum wage increase in July 2007 have any employment or wage effect on low-paid, full-time workers? Was there any spillover effect? The ratio of minimum wage increase is obvious. The effects would be isolated from other minimum wage adjustments. Setting Low-Paid Group (Treatment): People who had insured wage lower than 17,280 in Jan. 2004. Relatively High-Paid group (Control): People who had insured wage of Strategy 18,300 in Jan. 2004. 18,300 is the smallest level which would definitely not be affected by the minimum wage increase. First, using regression discontinuity design to see the employment change of the low-paid group by taking time as the running variable. Second, using the relatively high-paid group to fix unobservable trends. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 17 / 63

4. Methods and Data Processing Data Processing Individual Level Age limit for female is 22 to 54 and 24 to 54 for male in Jan. 2004. By the definition above, the treatment group has 926 thousand people in it and the control group has 144 thousand people. The sex, age, area, and industry ratio are as follows: Treatment: more than 65% is female; Control: only 53% Treatment: 15% is less than aged 25; Control: only 8% Treatment: 65% is out of Taipei or New Taipei City; Control: only 60% Treatment: 55% is in service sector; Control: only 50% Balanced Panel: Tracing the employment and wage changes of the same Firm Level people in the treatment and control groups from 2004 to 2010. Observing the whole existing firms. There are 370 to 440 thousands of firms during 2004 to 2010. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 18 / 63

4. Methods and Data Processing Model Regression Discontinuity Design y it = α 0 + α 1 D it + α 2 T ime t + α 3 (D it T ime t ) + θx i + ɛ it y it : Dummy variable of employment & nominal insured wage, which is equal to 0 if not being employed. D it : Dummy variable equal to 1 if observations is on the right hand side of 1/July/2007 and equal to 0 otherwise T ime t : Counted by the number of months before or after 1/July/2007. Since the data is from the end of the month, so the month of June 2007 is set as 0. X i : Other characteristics Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 19 / 63

4. Methods and Data Processing Model Regression Discontinuity Design & Difference-In-Difference y it = β 0 + β 1 (D it T reat i ) + β 2 D it + β 3 T reat i + β 4 T ime t + β 5 (D it T ime t ) + β 6 (T reat i T ime t ) + β 7 (D it T reat i T ime t ) + θx i + ɛ it T reat i : Dummy variable equal to 1 if observation is in the treatment group and equal to 0 otherwise Under the setting above, observations should include both treatment and control groups. Regression Discontinuity Design of Difference y control t y treat it = γ 0 + γ 1 D it + γ 2 T ime t + γ 3 (D it T ime t ) + θx i + ɛ it Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 20 / 63

4. Methods and Data Processing Expected Wage Effects The average nominal wage is equal to: w t = R 0 0 + R 1 15840 + R 2 16500 + R 3 wt h R 0 to R 3 are proportions of people with different insured wages. 0 means not being employed. wt h means the whole insured wages not less than the new minimum wage. Assume there s only time trend w t+1 w t = T rend = e T w t + (1 + e T ) w T e T : Employment changes caused by time trend; w T : Nominal wage changes caused by time trend. Assume the wage effects of time trend are proportionally distributed to every wage level, then w T = R 1 w T w t 15840 + R 2 w T w t 16500 + R 3 w T w t w h t Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 21 / 63

4. Methods and Data Processing Expected Wage Effects Assume there s only policy effect w t+1 w t = R [ e 1 R mw R 1+R 2 15840+ e 2 R mw R 1+R 2 16500]+(R 1 + e 1 mw R 1+R 2 )(17280 R 15840) + (R 2 + e 2 mw R 1+R 2 )(17280 16500) + R 3 (wt h w h t ) e mw : Employment effect of minimum wage increase wt h : Wage of people not affected by minimum wage increase after July 2007 Combining the time trend and policy effect w t+1 w t = e T w t + (1 + e T ){[ e mw R 1 R 1+R 2 15840 + e mw R 2 R 1+R 2 16500]+(R 1 + e mw R 1 R 1+R 2 ) max{( w T 15840)} + (R 2 + e mw R 2 R 1+R 2 ) max{( w T 16500)} + R 3 max{( w T w t w t 15840), (17280 w t 16500), (17280 wt h ), (wt h w h t )}} Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 22 / 63

4. Methods and Data Processing Expected Wage Effects The expected policy effect w mw = (1 + e T ){[ e mw 17280] + [R 1 (17280 15840 w T w t 15840) + R 2 (17280 16500 w T w t 16500)]+ [R 3 (wt h w h t w T w t wt h )]} Spillover or not If there re positive spillover effects, then the estimated wage effects should be larger than the summation of the employment (red) and the wage increase effect (blue) on the low-paid workers. Otherwise, if the estimated results are not larger, then there s no positive spillover effect. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 23 / 63

4. Methods and Data Processing Expected Wage Effects (Control-Treatment) Difference of wage change Assume there s small difference of employment change and wage change caused by time trend between the two group of people, then the difference of wage effect caused by minimum wage increase could be simplified to: wmw c wmw t (1 + e T ){[( e c mw e t mw) 17280] + [(R1 c R1)(17280 t 15840 ratio T 15840)] + [(R2 c R2)(17280 t 16500 ratio T 16500)] + [(R3 c R3(w t t h w h t ratio T wt h )]} Spillover or not Under the assumption of no spillover effect, the estimated effect should be negative (the wage gap decreases). If there re spillover effects, then the estimated effects should be larger (less negative) than expected results. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 24 / 63

5. Estimated Results Description Figure: Description of Variables in Treatment Group Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 25 / 63

5. Estimated Results Description Figure: Description of Dependent variables (Control-Treatment) Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 26 / 63

5. Estimated Results Trends of Variables(Treatment Group) (a) Male (b) Female Figure: Prop. of People with Insured Wages Not Less Than 17,280 in Treatment Group Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 27 / 63

5. Estimated Results Trends of Variables(Treatment Group) (a) Male (b) Female Figure: Employment Ratio in Treatment Group Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 28 / 63

5. Estimated Results Trends of Variables(Treatment Group) (a) Male (b) Female Figure: Average Nominal Wage in Treatment Group Note: Nominal wage is set as 0 if not being employed. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 29 / 63

5. Estimated Results Trends of Variables(Control Group) (a) Male (b) Female Figure: Prop. of People with Insured Wages Not Less Than 17,280 in Treatment Group Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 30 / 63

5. Estimated Results Trends of Variables (Control-Treatment) (a) Male (b) Female Figure: Difference of Prop. of Non-Binding between Two Groups Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 33 / 63

5. Estimated Results Trends of Variables (Control-Treatment) (a) Male (b) Female Figure: Difference of Employment Ratio between Two Groups Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 34 / 63

5. Estimated Results Trends of Variables (Control-Treatment) (a) Male (b) Female Figure: Difference of Average Nominal Wage between Two Groups Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 35 / 63

5. Estimated Results Estimated Results Figure: Estimated results of Male in the Treatment Group Note: MSE = Mean Square Error optimal bandwidth; CER = Coverage Error-Rate optimal bandwidths Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 36 / 63

5. Estimated Results Estimated Results Figure: Estimated results of Female in the Treatment Group Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 37 / 63

5. Estimated Results Estimated Results (Control-Treatment) Figure: Estimated results of Male Between the Two Groups Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 38 / 63

5. Estimated Results Estimated Results (Control-Treatment) Figure: Estimated results of Female Between the Two Groups Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 39 / 63

5. Estimated Results Heterogeneity and Robustness Heterogeneity Estimation: Scale: In firms with employment 100 no effect; with employment 30 wage effect. Industry: Wholesale and Retail, Non-Service Sector wage effect; Restaurant no effect to male but wage effect to female. Age: Age (in 2004) 30 or 50 wage effect. Area: Taipei, Non-Taipei wage effect. Robustness Check: Sample selection: using individuals having low wages during Jan. 2004 to Mar. 2004. The results are similar. No treatment effect? Survey of Family Income and Expenditure: In 2007, the lowest 3 deciles of household had increased disposable income relative to other deciles of households except for the highest one. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 40 / 63

5. Estimated Results Household Income Figure: Disposable Income of Households Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 41 / 63

6. Transition Transition Matrix 1 (Control-Treatment) (a) Male (b) Female Figure: Full-Time to Full-Time (Difference) Note: Increasing means probabilities in the control group are larger than ones in the treatment group. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 42 / 63

6. Transition Transition Matrix 2 (Control-Treatment) (a) Male (b) Female Figure: Full-Time to Part-Time (Difference) Note: Decreasing means probabilities in the treatment group are larger than ones in the control group. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 43 / 63

6. Transition Transition Matrix 3 (Control-Treatment) (a) Male (b) Female Figure: Full-Time to Unemployed (Difference) Note: Decreasing means probabilities in the treatment group are larger than ones in the control group. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 44 / 63

6. Transition Transition Matrix 4 (Control-Treatment) (a) Male (b) Female Figure: Part-Time to Full-Time (Difference) Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 45 / 63

6. Transition Transition Matrix 5 (Control-Treatment) (a) Male (b) Female Figure: Part-Time to Part-Time (Difference) Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 46 / 63

6. Transition Transition Matrix 6 (Control-Treatment) (a) Male (b) Female Figure: Part-Time to Unemployed (Difference) Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 47 / 63

6. Transition Transition Matrix 7 (Control-Treatment) (a) Male (b) Female Figure: Unemployed to Full-Time (Difference) Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 48 / 63

6. Transition Transition Matrix 8 (Control-Treatment) (a) Male (b) Female Figure: Unemployed to Part-Time (Difference) Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 49 / 63

6. Transition Transition Matrix 9 (Control-Treatment) (a) Male (b) Female Figure: Unemployed to Unemployed (Difference) Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 50 / 63

7. Firm Adjustments Description Figure: Description of Variables of Firms Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 51 / 63

7. Firm Adjustments Figure: Description of Small Firms (with Employees < 30 in the First Place during Sample Period) Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 52 / 63 Description

7. Firm Adjustments Trends of Variables (Small Firms) Figure: Non-Binding Ratio of Firms Figure: Firms Density Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 53 / 63

7. Firm Adjustments Trends of Variables (Small Firms) Figure: Prop. of Firms in Service Sector Figure: Prop. of Firms in Taipei or New Taipei City Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 54 / 63

7. Firm Adjustments Trends of Variables (Small Firms) Figure: Full-Time Employment in A Firm Figure: Average Wage of Full-Time Workers in A Firm Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 55 / 63

7. Firm Adjustments Trends of Variables (Small Firms) Figure: Employment of Full-Time Figure: Average Wage of Full-Time Workers with Wage 18, 300 in A Firm Workers with Wage 18, 300 in A Firm Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 56 / 63

7. Firm Adjustments Trends of Variables (Small Firms) Figure: Total Insured Wages of Full-Time Workers in A Firm Figure: Total Insured Wages in A Firm Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 57 / 63

7. Firm Adjustments Trends of Variables (Small Firms) Figure: Separation Rate of Full-Time Workers in A Firm Figure: New Hiring Rate of Full-Time Workers in A Firm Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 58 / 63

7. Firm Adjustments Trends of Variables (Small Firms) Figure: Occupational Injury Rate in A Firm Figure: Ordinary Injury Rate in A Firm Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 59 / 63

7. Firm Adjustments Estimated Results (Small Firms without data in July & Aug.) Figure: Estimated Results of Small Firms Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 60 / 63

8. Primary Conclusion and Future Work Primary Conclusion and Future Work The effects of minimum wage increase in July 2007 Having positive wage effect but no significant employment effect to full-time workers. It decreased the wage gap, but no evidence proves spillover effects. Mechanism Low-paid, full-time, and female workers may lose jobs but the effect was too moderate. Firms bore the cost. Turnover rate was decreased. Firms may cut down other expenditure or push employees to work more which lead to higher occupational injury rate. Future Works The effects on part-time workers. Other firm s adjustments. Lu, Chyi-Horng (Economics, NTU) Estimating the Effects of Minimum Wage 2018.6.14 63 / 63